IN
DEGREE PROJECT
ELECTRICAL ENGINEERING,
SECOND CYCLE, 30 CREDITS
,
STOCKHOLM SWEDEN 2018
Utilizing Privately Owned
Flexibilities in the German
Distribution System
Technical and Regulatory Framework
NAHAL TAMADON
KTH ROYAL INSTITUTE OF TECHNOLOGY
Master of Science Thesis 2018 Utilizing Privately Owned Flexibilities
in the German Distribution System Nahal Tamadon
Approved Examiner Supervisor
2018- Patrik Hilber Ebrahim Shayesteh
Thorsten Gross
Commissioner Contact Person
Abstract
This Master’s thesis project aims to define the technical and regulatory framework of a system flexibility service utilizing distributed flexibilities connected to the low-voltage (LV) level. The flexibility service outlined here is local balancing, with the purpose of increasing the load-generation balance in LV networks using thermostatically controlled loads and Electrical Energy Storage (EES) devices. The thesis is performed at Avacon Netz GmbH, a German DSO, as part of the German demo for project InterFlex.
The first part of the thesis encompasses the technical framework of local balancing. For this purpose, day-ahead scheduling algorithms for heat pumps and EES devices are proposed to increase the local consumption of distributed generation, and thereby decrease the reverse power flow and load peaks. Two scheduling algorithms are proposed in this thesis, one based on numerical methods and one on optimization. The proposed algorithms are simulated for nine LV-networks in Avacon Netz’ grid area. The simulation results show that by utilizing all the residential heat pumps and EES devices, the load peaks and the peak reverse flow can be reduced by up to 21.30% and 37.30% respectively.
Sammanfattning
Syftet med detta examensarbete är att definiera det tekniska och regulatoriska ramverket för en flexibilitetstjänst som utnyttjar distribuerad flexibilitet kopplade till distributionssyste-mets lågspänningsnät (LV). Den flexibilitetstjänst som beskrivs här kallas lokal balansering, med syftet att öka balansen mellan den producerade och den utnyttjade effekten i enskilda lågspänningsnät med hjälp av termostatstyrda laster och elektrisk energilagring (EES). Ex-amensarbetet utförs i samarbete med det tyska elnätsbolaget Avacon Netz GmbH som ett delprojekt av det europeiska projektet InterFlex.
Rapporten består av två delar. Den första delen beskriver den tekniska processen för lokal balansering. I denna rapport föreslås två algoritmer, en baserad på numeriska- och en ba-serad på optimeringsmetoder, för schemaläggning av värmepumpar och EES med syftet att öka den lokala användningen av den lokalt producerade eleffekten. Algoritmerna tillämpas på nio lågspänningsnät, och simuleringsresultaten visar att de föreslagna algoritmerna kan mins-ka effekttopparna och effektflödet till högre spänningsnivåer med upp till 21,30%, respektive 37,30%.
Contents
1 Introduction and Background 5
1.1 Introduction. . . 5
1.2 Project Background . . . 5
1.3 Ethical Aspects . . . 6
1.4 Scope and Structure of the Report . . . 6
I Technical Framework
7
2 Literature Review 7 2.1 Distributed Flexibility Sources . . . 72.1.1 Energy Storage System . . . 7
2.1.2 Demand Side Management . . . 8
2.1.2.1 Demand Side Management Objectives . . . 8
2.2 Methods for Utilizing Distributed Flexibilities . . . 9
2.3 Estimating Capital Expenditures . . . 10
2.3.1 Cost of Grid Reinforcement . . . 10
2.3.2 Cost of Smart Grid Infrastructure . . . 10
3 Methodology 12 3.1 Operational Principle of Electrical Energy Storage Systems . . . 12
3.2 Operational Principle of Heat Pumps. . . 12
3.3 Scheduling Model . . . 12 3.3.1 Numerical Approach . . . 13 3.3.1.1 EES Discharging . . . 14 3.3.1.2 Heat-Pump Activation. . . 16 3.3.2 Optimization Approach . . . 16 4 Model Implementation 18 4.1 Technical Specifications of the Electrical Energy Storage System . . . 18
4.2 Thermal Parameters of the Heating System . . . 18
4.3 Test Network and InterFlex Customers. . . 19
4.4 Load and Generation Data . . . 19
4.5 Smart Grid Hub Architecture . . . 20
4.6 Simulation Cases . . . 21
4.7 Simulation Software . . . 21
4.8 Key Performance Indicators . . . 21
4.9 Capital Expenditure . . . 22 5 Results 23 5.1 Base Case . . . 23 5.2 Case A. . . 23 5.2.1 Numerical Approach . . . 24 5.2.2 Optimization Approach . . . 25 5.3 Case B. . . 27 5.3.1 Numerical Approach . . . 27 5.3.2 Optimization Approach . . . 29
5.4 Comparison of the Proposed Methods . . . 31
5.5 Reverse Power Flow . . . 31
5.6 Loading of the Medium-Voltage Feeder. . . 32
5.7 Sensitivity Analysis. . . 33
5.8 Economical Analysis . . . 34
6 Flexibility Services 36
6.1 System Flexibility Services. . . 36
6.2 Commercial Flexibility Services . . . 37
7 Regulatory Framework for Direct Load Control 37
8 Flexibility Markets 38
8.1 Potential Market Structures . . . 38
8.2 Proposed Market Structure . . . 38
1 Introduction and Background
1.1 Introduction
The share of renewable energy sources (RES) has increased noticeably during recent years. In Ger-many, the share of RES-based electric power consumption has increased from only 9.3% in 2004 to 31.6% in 2015 [1]. To comply with the German Renewable Energy Sources Act (EEG 2017), the share of renewable-based generation will increase to 40-45% and 55-60% by 2025 and 2035 respec-tively [2]. The increase in electricity demand is also expected to continue in the coming decades. Due to the uncontrollable nature of RES on one hand, and the increasing electricity demand on the other hand, maintaining the load-generation balance while preserving a high level of power quality and system reliability has become increasingly challenging for power system operators. In a conventional power system, the power is generated centrally, in schedulable, synchronous gen-erators. The most frequent cause of unbalance in a conventional power system is thus the uncer-tainty of load forecast, which is balanced out using the power reserve provided by the conventional generators. In a power system with a large share of RES-based generation, the uncertainty is caused by both demand and generation. With less and less conventional generators in the power mix, developing solutions to replace the synchronous generators’ ability in providing power reserve and voltage support, has been the focus of a significant amount of research. Technologies such as Demand Side Management (DSM), Electrical Energy Storage (EES) systems, and feed-in man-agement are the potential sources of balancing power in a modern, actively managed power system. In the context of active distribution system management and local energy systems, Avacon Netz GmbH is preparing a demo for the European Commission’s project InterFlex. As part of the said demo, this thesis proposes a flexibility service called local balancing, which utilizes distributed flex-ibility sources to increase the load-generation balance at low-voltage (LV) level. The work carried out in this thesis consists of a technical part, proposing two algorithms for local balancing, and a regulatory part, defining the commercial and regulatory frameworks for implementing this service. The following subsections provide an overview of the InterFlex project, the main contributions and the ethical aspects of this thesis, and an overview of the report.
1.2 Project Background
As part of the Horizon 2020 program, the InterFlex project aims at enhancing the performance of the European Distribution System Operators (DSOs) by taking advantage of the opportunities and overcoming the challenges associated to more flexible local energy communities. The objective is to run 18 use-case demonstrations in five European countries, each utilizing one or more of the following technologies:
• Energy storage • Demand response • Electric vehicles • Grid automation
Avacon Netz GmbH is responsible for the German demo, and has defined three use cases, focusing on grid automation and demand response. The use cases are as follows:
1. Feed-in management, where the objective is to include small scale RES in the congestion management process, in order to decrease RES curtailment
2. Demand response, where the objective is to utilize residential load in order to solve grid congestions or other critical events
The subject of this thesis is developing functions for use case 3. The German DSO is fully unbun-dled from all trading actions, and is hence prohibited from procuring flexibility services. However, the need for active management of the distribution system has been realized and is being evaluated, among others by the European Commission [3], and the European Distribution System Operators’ Association (EDSO) [4]. As part of an R&D program, use case 3 aims at providing a technical framework for, and evaluating the potential of, utilizing distributed flexibilities in order to actively manage the distribution system and improve its performance. This thesis focuses on utilizing the flexibility on residential customers’ premises to support the local grid. However, an analysis of the regulatory framework is performed to identify the modifications that would enable the DSO to take full advantage of privately owned flexibility.
For the German demo, the distribution Supervisory Control and Data Acquisition (SCADA) system is extended with a Smart Grid Hub (SGH) in a local distribution area. The chosen area has seen an increase in RES penetration in recent years, which is increasing the risk for voltage instability and congestion. More than 370 residential customers in the test area have volunteered to take part in the InterFlex project and delegate control of their flexible devices to the DSO. Each of these households possesses one or several of the devices central to the Avacon demo, which are distributed generation (DG), flexible loads, and EES. Each household is equipped with a Smart Meter (SM) and a control box connected to the SGH to enable central control of the devices.
1.3 Ethical Aspects
The contributions of this thesis are in line with the H2020 objectives for emission reduction. This work aims at supporting system operators in locations with high RES penetration, thus facilitating the integration of renewable based distributed generation. In addition to the apparent environ-mental benefits, the improved utilization of distributed flexibility also empowers the customers and provides them with more market power and lower electricity costs. The communication in-frastructure of the SGH complies with the technical guidelines of the German Federal Office for Information Security (BSI TR-03109), thus ensuring the privacy of end-users. A downside to the model proposed here, is that it enables the DSO to influence the consumption pattern. High level of transparency in load dispatch is thus required to ensure impartiality.
1.4 Scope and Structure of the Report
The work performed in this thesis, and consequently the report at hand, consists of two main parts: a technical part, and a regulatory part. Chapter I, consisting of Sections 2-5, contains the technical part of the report. Section2 provides a literature review of distributed flexibilities, methods for utilizing distributed flexibilities, and the basics for an economical analysis of the pro-posed application. Section3 provides a detailed description of the proposed models. The basics of the numerical simulation are described in Section4, the results of which are provided in Section5. Chapter II, consisting of Sections 6- 8, covers the regulatory framework of utilizing distributed flexibilities. Section6, provides a literature review of flexibility services. In Section7, the current German regulations for utilizing distributed flexibilities through the DSO are reviewed. Section8
Part I
Technical Framework
2 Literature Review
The German government has taken effective measures to promote installation of Photovoltaic cells (PVs)[5], which has resulted in a high level of distributed RES penetration. As a result of the German Renewable Energy Act, the integration of distributed wind and solar power, and biomass-fueled CHP is expected to further increase in the upcoming three decades [6]. The steady transision towards distributed generation has introduced a high level of stress on the distribution system. Ar-eas with low energy consumption and high generation, and highly clustered arAr-eas are faced with increasing congestion and voltage limit violations [7,8]. Appropriate control measures can, how-ever, increase the stability of the system and delay the need for expanding grid capacity [7,9,10]. Local balancing of the distribution system is proposed in [11] as a means to decrease the stress that DG puts on the distribution system. That is, [11] proposes an active distribution management approach based on economic dispatch by using an objective function that minimizes system oper-ation costs, taking into account the voltage and power constraints of the network. The German Renewable Energy Sources Act (EEG), enables distributed RES-based generators to inject power into the grid without being included in an active dispatch mechanism to promote the integration of RES-based generation. To comply with the energy act, and to further promote RES, an alternative approach for local balancing is proposed and evaluated in this thesis. The proposed model aims at balancing the system on the LV level by adjusting the load profile of each LV network to as far as possible follow the generation profile of that network.
Section2.1 provides a background and literature review on the distributed flexibility sources uti-lized in this thesis. Section2.2Provides a literature review on flexibility-utilization methods. To evaluate the potential of flexibility services in decreasing grid expansion costs, an estimation of the distribution system capital expenditures is provided in Section2.3. The potential flexibility services are reviewed in Section6.
2.1 Distributed Flexibility Sources
The balance between production and consumption in a power system has long been provided by large controllable generators [12–14], but a large share of these units are expected to be switched off and replaced by RES-based generation with lower marginal costs and emission levels [15,16]. Besides the uncontrollable nature of RES, many of these units are connected to the distribution grid, causing reverse power flows and increasing the need for voltage control. The combination of increased RES-based distributed generation and decreased availability of conventional generators, creates a greater need for flexibility in the power system [17,18] known as a "flexibility gap" [3]. Flexibility in a power system is defined as "the ability of the power system to respond to changes in demand and supply" [19] or more generally, as "the ability to adapt to dynamic and changing conditions" [20]. Storage units, DSM programs, improved feed-in management (RES curtailment), electric vehicles (EVs), and interconnection are some of the alternative sources for flexibility in a smart power system with large penetration of RES [12,13,21,22]. As previously mentioned, this project focuses on the flexibility provided by EES devices and DSM programs.
2.1.1 Energy Storage System
The operational principle of EES devices are provided in Section3.1
2.1.2 Demand Side Management
DSM programs are defined as modifications in consumption caused by variations in electricity price or incentives provided to customers [18, 28–30]. The majority of the available literature classify DSM programs into two main types: price-based programs and incentive-based programs [18,28,31–33]. In price-based programs, the electricity price is higher during peak hours to encour-age the customers to shift the load to off-peak hours. In incentive-based programs, the customer is either awarded for decreasing the consumption when required, or penalized when not doing so. According to [34], DSM programs can also be classified into market-based and Direct Load Control (DLC) programs. In market-based programs, the electricity price or the financial incentives are adjusted to obtain the desired load profile. There is a high level of uncertainty associated with this kind of programs since they depend on anticipating consumer behavior [35]. Another drawback is that these types of DSM programs are not practical in unpredicted situations such as component failure. In DLC programs the customer delegates the control of flexible loads to a third-party agent which is authorized to utilize the demand side flexibility whenever required. It is, however, crucial to maintain high quality of service and not compromise the comfort of the customer[34].
The DSM program applied in this project is incentive-based DLC, where the DSO has full con-trol of the distributed flexibility resources. The incentives are paid to all participating customers, regardless of the level of available or utilized flexibility. Hence, there are no market mechanisms involved and the algorithm for controlling the privately owned flexibilities are optimized from the system-performance perspective, independent of the electricity price.
Since the responsible agent in DLC programs has direct control over the flexible loads, it is crucial that the comfort of the customers is not compromised and that a high quality of service is kept at all time. The flexible loads utilized in this thesis are heat pumps. The comfort provided by a heat pump is measured in units of temperature and humidity. Here, we use the model proposed in [36] and disregard the humidity. The thermal comfort zones for indoor spaces, specified by the ASHRAE standard [37] are used as constraints for the operation of heat pumps. The operational principle of heat pumps and the thermal comfort zones are described in Section3.2 and Table4
respectively.
2.1.2.1 Demand Side Management Objectives
As previously stated, DSM programs aim at modifying load patterns to increase system perfor-mance. There are six main objectives of load shaping [38–40]. The three basic load shape objectives are
• Peak clipping: reduces the system peak load and is of interest for systems with insufficient generation or grid capacity. It also decreases the power-based price component of the DSO expenses
• Valley filling: increases the load during off-peak hours and could reduce the average fuel cost. Shifting the load to hours with high generation also decreases the amount of power export and reverse power flows
• Load shifting: shifts the load from peak hours to off-peak hours. This method can be regarded as a combination of peak clipping and valley filling, without necessarily changing the average consumption
In addition to these basic objectives, there are also three advanced-level objectives. The advanced objectives are
• Strategic load growth: also known as electrification, encourages the application of electrical technologies such as electric vehicles either to decrease fossil fuel consumption or to increase the overall quality of life
• Flexible load shape: uses flexible loads to achieve the desired changes in load profile in order to increase system reliability
The above mentioned load shaping approaches are illustrated in Figure1. More generally, DSM programs can be defined as means to adjust the consumption in order to match the desired load profile.
Figure 1: Demand Side Management Objectives [40]
2.2 Methods for Utilizing Distributed Flexibilities
The available literature include numerous methods for DSM and EES allocation. These methods can be classified into real-time allocation, and day-ahead schedules. A real-time dispatch algo-rithm for EES is used in [41], where the difference between load and generation is checked at the beginning of each control period, and either storage capacity or stored energy is allocated when needed. One real-time approach is demonstrated in [42], where the control center keeps a real-time list of available flexibilities and activates them when voltage or power limits are exceeded. Another real-time approach is demonstrated in [43], where thermostatically controlled loads are switched on or off to match the RES generation. The simulation results in [43] show that the algorithm could lead to an increase in local electricity consumption. To avoid this side effect, and to make optimal use of the available resources, a day-ahead scheduling approach is adopted in this thesis. The majority of the available literature focus on day-ahead scheduling approaches using optimiza-tion. Among these, the most frequent objective is either cost minimization or profit maximization depending on the point of view of the decision maker, [25–28, 32, 40, 44–50]. Models for using optimization to improve system performance, e.g. through peak clipping and valley filling [39,51], minimizing load variance [52], or increasing RES power injection [16], have also been demonstrated. As previously described, the objective in this thesis is increasing the local load-generation balance. For this purpose, two models are proposed and analyzed. The first method, described in Section
on a continuous range of values and the model optimizes both the duration and the magnitude of power consumption.
2.3 Estimating Capital Expenditures
Utilizing local flexibilities in grid operation is expected to change the cost structure of DSOs. A study conducted by [54] suggest that taking advantage of flexible loads decreases the investment cost of grid reinforcement. This study, however, does not take into account the cost of DSM. The trade off between contracting flexibility versus expanding the distribution grid assuming that the DSO purchases flexibility at the same price as electricity in the wholesale market was analyzed in [55]. Reference [31] compares the TSO costs of congestion management when flexibility is pur-chased at the reserves market, versus when an incentive-based DR program is used. According to [3], active management of DR can increase operating expenditures (OPEX) of DSOs, but argues that the impact of active network management on capital expenditures (CAPEX) is not obvious. Active network management can be used to decrease grid expansion costs, but in order to evaluate the net impact on CAPEX, the investment cost for smart grid infrastructure should be taken into account [56]. According to [57], and to the best of the author’s knowledge, the analysis of net impact of active network management on CAPEX is lacking in the literature. This thesis attempt to fill this gap by quantifying the affect of the proposed network operation approach on distribution system CAPEX. The impact of local balancing on CAPEX includes the avoided cost of grid expansion due to decreased net load, and the investment cost of advanced metering and communication infrastructure. The methodologies used to quantify each of these expenditures are described in the next sections.
2.3.1 Cost of Grid Reinforcement
The cost analysis used in this section is based on [58]. The cost components of distribution grid reinforcement that can be deferred using distributed resources can be grouped into transformer and substation investments, and lines and feeders investment. A more detailed overview of the cost components are given in Table1. The study has analyzed the investment costs of 124 US utilities during a five-year period and estimated the average cost of investment per MW of system peak growth. The numerical values obtained in this study are provided in Section4.9.
Table 1: Grid Reinforcement Cost Components Associated Costs
Transformer and Substation Investment Land and Land RightsStructures and Improvements Station Equipment
Line and Feeder Investment
Land and Land Rights Structures and Improvements storage Battery Equipment Poles, Towers and Fixtures Overhead Conductors and Devices Underground Conduit
Underground Conductors and Devices Line Transformers
2.3.2 Cost of Smart Grid Infrastructure
The prerequisite for DSM is a communication and metering infrastructure that enables the re-mote switching of behind-the-meter appliances. The building blocks of an Advanced Metering Infrastructure (AMI) are [59]
• Smart meters with two-way communication capabilities, and remotely controllable switching relay
• Communication system to exchange metering data and control signals between the AMI and the power system SCADA
3 Methodology
This section outlines the technical framework for local balancing. The operational principles of EES and heat pumps are provided in Sections3.1 and3.2 respectively. The two flexibility scheduling models proposed in this thesis are described in Section3.3.
3.1 Operational Principle of Electrical Energy Storage Systems
The operational principle of EES, as defined in [25,27], is given by
EtEES= Et 1EES+ ⌘EESPtEES t (1)
EEES
min EEESt EmaxEES (2)
PtEES PmaxEES (3)
where EEES
t is the energy stored in the battery at the end of time-step t,
⌘EES is the charging/discharging efficiency of the battery,
PEES
t is the power input of the ESS during time-step t,
tis the duration of each time-step, PEES
max is the maximum charging/discharging power of the EES, and
EEESmin and EmaxEES are the recommended lower and upper limits of EES stored energy respectively
[27].
Depending on whether the ESS is being charged or discharged, PEES
k has a positive or negative
value respectively. The values for these parameters are given in Section4.1.
3.2 Operational Principle of Heat Pumps
In some literature, thermal specifications of buildings, such as the area and heat capacity of walls and windows, are used to model the indoor temperature as a function of the electricity consumption of the heater[60,61]. Since the physical specifications of the buildings with DR capabilities in this project are not available, we use the model proposed in [36]. Based on this model, the space temperature in the next time interval is given by
✓t= ✏✓t 1+ (1 ✏)(✓tA+ ⌘tCOP
PHP t
A ) (4)
where
✓t[ C]is the indoor temperature at the end of time-step t,
✓A
t [ C]is the ambient temperature during time-step t,
PHP
t [kW ]is the active power consumed by the heater during time-step t,
⌘COP
t is the coefficient of performance (COP) of the heater during time-step t,
A [kW/ C]is the overall thermal conductivity of the building, and ✏is the factor of inertia given by
✏ = exp[ t/T C] (5)
where t is the duration of one time step used in the study, and T C is the time constant of the system [25]. The only thermodynamic parameters needed for this model are ✏, ⌘COP
t , and A. The
values of these parameters are given in Section4.2.
3.3 Scheduling Model
description of the proposed models. The devices in each local area are treated in an aggregated manner, i.e. for each area and device type, the model result in a single schedule or switching signal, sent out to all devices in respective category.
3.3.1 Numerical Approach
To increase the local load-generation balance, the algorithm seeks to charge the EES during the time periods when local net load
Pnet= PLoad PDG (6)
is at its lowest. Here, PLoadand PDG are the local load and generation respectively. Charging all
EES capacity in one or two time periods could potentially create fluctuations in the load profile. In order to avoid this, an iterative method is proposed. With this method, during each iteration, the time period with lowest net load is identified and a fraction of the available storage capacity is schedule to be charged during that time period, t. The forecasted net load of that time period is updated, taking into account the scheduled EES charging, and a new minimum is identified for allocation of the next fraction of storage capacity. The number of iterations are chosen such that the EES-charging schedule creates as small fluctuations as possible.
Algorithm 1 Algorithm for EES charging Input: PDG, PLoad, CEES, tol
u, tolex
Output: EES charging schedule EEES
Initialisation :
Define tolf luct, e, Iterate = 1, isgood = 0 1: while isgood = 0 (Iteration-accession loop) do
2: Create an empty array to save the indices of the modified time-periods, IndCH=[ ] 3: Create an array to save the EES charging schedule, EEES = [0, ..., 0]
4: while min(CEES, sum(PDG)) > tol
c (Capacity-allocation loop) do 5: At each iteration, allocate a fraction of the available EES capacity
CEES
f raction= C
EES
Iterate 6: PEES= EEES/ t
7: Pnet= PLoad+ PEES PDG
8: Find the net load for the time-periods with nonzero generation, and available EES power
capacity Pnet
search={Pnet|PDG 6= 0&PtEES < PmaxEES} 9: Find the time period with lowest net load Pmin
t = min(Psearchnet ) 10: IndCH=[IndCH, t]
11: EES = min{CEES
f raction, PtDG t, (PtEES PmaxEES) t}
12: Update PEES
t , PtDG, and CEES
13: end while
14: Update Pnet, and EEES 15: Remove doubles from IndCH 16: Define a counter Fluct=0
17: for (t 2 IndCH) (Fluctuation-counting loop) do
18: if (Ptnet Pt 1net> tolf luct) or (Ptnet Pt+1net> tolf luct) then
19: Fluct=Fluct+1
20: end if
21: end for
22: if (Fluct > 0) then
23: if iterate is too large then 24: tolf luct= tolf luct+ e
The algorithm for EES charging consists of several nested loops. The main loop is the Iteration-accession loop, which creates an empty array to store the indices of the modified time periods, and passes it on to the Capacity-allocation loop. This while loop starts by calculating a fraction of EES capacity, CEES
f raction, to be allocated during each iteration. An array containing the local
net-load, Pnet= PLoad+ PEES PDGis then created/updated. The time period with lowest net
load, t, is then identified and stored in the array containing the modified indices, Indmodif ied.
The allocated EES capacity, EES is chosen such that the following three conditions are met:
• EES CEES
f raction
• the EES does not store more than is locally produced • PEES
t PmaxEES
The Capacity-allocation loop then updates CEES, PDG, and PEES. PDG is updated to make sure
that the algorithm does not store more energy than is locally produced and the loop is repeated until either the unallocated part of CEES or the generated power that has not been stored is
smaller than a tolerance value tolc. This value is given by
tolc= ErEES EmaxEES (7)
When the Capacity-allocation loop has finished running, the net load, Pnet is updated, taking
into account the EES charging schedule. The Fluctuation-counting loop is then used to count the number of fluctuations in Pnet created during the modified time periods. To ensure that the
algorithm always converges, a tolerance value, tolf luct, is used, i.e. if the power fluctuation is less
than tolf luct, the fluctuation is neglected. If the EES charging schedule causes power fluctuations,
the algorithm checks to see whether the number of iterations has exceeded a preset value, and in that case increases the tolerance for fluctuations. The number of iterations is increased and the Iteration-accession loop is repeated. If the fluctuations caused by EES charging are negligible, the algorithm exits the Iteration-accession loop.
3.3.1.1 EES Discharging
The EES discharging algorithm ,similar to the charging algorithm, seeks to increase the local load-generation balance by discharging the EES when local net load, Pnet, is at its highest. In order
to avoid causing fluctuations in the power profile, this algorithm also adopts an iterative method, during which, a fraction of the available stored capacity is scheduled to be discharged during each iteration. The number of iterations are then successively increased until the obtained discharging schedule causes as small fluctuations as possible.
The discharging algorithm also contains an Iteration-accession loop, which calculates the EES power input, PEES, the total available stored energy, Estored, and creates an empty array to
store the indices of the modified time-periods, Indmodif ied. The second main loop is the
Energy-allocation loop which starts by calculating a fraction of the stored energy, Estored
f raction, to be allocated
during each iteration. The array, Pnet
search is then created to limit the search to time periods during
which the state of charge of the EES, SOCEES, is non-zero and the discharging power of the
EES has not reached its limit. The time-period, t, with highest net-load is identified and saved to IndDC. The amount of stored energy to be allocated, EES is then chosen such that
• EES Estored f raction
• the EES does not discharge more than is locally consumed • |Pnet
t | PmaxEES
The Energy-allocation loop then updates Pnet, PEES
t , Estored, and SOCEES, and repeats until
either the unallocated part of Estored, or the net load is smaller than the tolerance value tol dc.The
tolerance is given by
toldc= ErEES EminEES (8)
Similar to the charging algorithm, when the Energy-allocation loop has finished running, the Fluctuation-counting loop is then used to count the number of fluctuations in Pnetcreated during
the modified time periods. Similarly, a tolerance value, tolf luct, is used to exclude the oscillations
that are smaller than tolf luct. If the fluctuations are negligible, the algorithm exits the
Algorithm 2 Algorithm for EES discharging Input: SOCEES, Pnet, EEES, tol
u
Output: EES discharging schedule Initialisation :
Define tolf luct, e, Iterate = 1, isgood = 0, calculate SOCEES 1: while isgood = 0 (Iteration-accession loop) do
2: IndDC=[ ]
3: PEES= EEES/ t 4: EStored= sum(EEES)
5: while min(EStored, Pnet) > tol
dc (Energy-allocation loop) do 6: At each iteration, allocate a fraction of the stored energy
Estored f raction=
Estored
Iterate 7: Pnet
search={Pnet|SOCEES 6= 0&|PtEES| PmaxEES}
8: Find Pmax
t = max(Psearchnet ) 9: IndDCd=[IndDC, t]
10: EES = min{Estored
f raction, Ptnet t, (|PtEES| PmaxEES) t}
11: Update Pnet
t , PtEES, Estored, and SOCEES
12: end while
13: Remove doubles from IndDC 14: Define a counter Fluct=0
15: for t 2 IndDC(Fluctuation-counting loop) do 16: if ((Pnet
t 1 Ptnet) > tolf luct) or ((Pt+1net Ptnet) > tolf luct) then
17: Fluct=Fluct+1
18: end if
19: end for
20: if Fluct>0 then
21: if Iterate is too large then 22: tolf luct= tolf luct+ e
3.3.1.2 Heat-Pump Activation
One of the challenges with thermostatically controlled loads is the load rebound. Assuming that in the absence of DSM, the indoor temperature is kept at the reference temperature ✓ref, if the
DSM program decreases the temperature for peak hours, the heating system corrects the low indoor temperature at the end of the peak hours by increasing the heat pump power consumption, causing a rebound. To avoid this, we propose using PHP as the controlled variable, instead of the indoor
temperature. For this purpose, three arrays are created: • PHP
min: the power needed to keep the indoor temperature at ✓min
• PHP
ref: the power needed to keep the indoor temperature at ✓ref =✓min+✓2 max
• PHP
max: the power needed to keep the indoor temperature at ✓max
Using Equation4, we obtain
PrefHP = T X t=1 ( A ⌘COP t )(✓ref ✓At) (9) PHP
min and PmaxHP are obtained by replacing ✓ref in Equation 9 with ✓min and ✓max respectively.
Using PHP
ref as the default setting, the algorithm changes PtHP for peak hours to the corresponding
values in PHP
min, and for the hours with low net load to their corresponding values in PmaxHP. These
hours are selected using IndCH and IndDCcreated by the EES charging and discharging algorithms.
The heat-pump activation algorithm then calculates the indoor temperature ✓tfor each hour using
Equation4, and corrects the heat pump schedule for the time periods when the indoor temperature is outside the comfort zone.
Algorithm 3 Algorithm for heat-pump activation Input: Pnet, ✓
ref, ✓max, ✓min
Output: Heat pump activation schedule Initialisation :
Calculate PHP
min, PmaxHP, PrefHP
set PHP = PHP ref 1: for (i in IndCH) do 2: PHP t = Pt,maxHP 3: end for 4: for (i in IndDH) do 5: PHP t = Pt,minHP 6: end for 7: for t in T do 8: Calculate ✓t
9: if ✓t< ✓min or ✓t> ✓max then
10: Modify PHP
t 11: end if
12: end for
3.3.2 Optimization Approach
As previously mentioned, the DSM program proposed here seeks to adjust the load profile so that the local load-generation mismatch is minimized. A general objective function for local balancing [53] can be formulated as min T X t=1 (Ptnet)2= T X t=1 (PtLoad+ PtHP + PtEES PtDG)2 (10) where PLoad
t is the share of the load that is not available to the system operator for controlling,
and PDG
Based on Equation4, the power consumed by the heat pump during time-step t is given by PtHP = nHP( ✓t ✏✓t 1 1 ✏ ✓ A t) A ⌘COP (11)
where nHP is the number of controllable heat pumps in the area. Using Equation 1, the power
injected into the EES during time-step t can be formulated as a function of the energy stored in the EES, given by
PtEES=
1 t ⌘EES(E
EES
t Et 1EES) (12)
Based on Equations10,11, and12the optimization problem is formulated as follows min T X t=1 ✓ Ptuncont+ nHP( ✓t ✏✓t 1 1 ✏ ✓ A t) A ⌘COP + 1 t ⌘EES(E EES t Et 1EES) PtDG ◆2 (13) Subject to: ✓min ✓t ✓max (14) PtHP 0 (15) 0.9 PrefHP T X t=1 (✓t ✏✓t 1 1 ✏ ✓ A t) A ⌘COP t 1.1 P HP ref (16)
EminEES EEESt EmaxEES (17)
PEES
t PmaxEES (18)
PtEES PtDG (19)
PtHP PmaxHP (20)
The variables of the optimization problem are
✓t f or i = 1, ..., T
EEES
t f or t = 1, ..., T
The indoor temperature at the beginning of current control period, ✓0, is equal to the temperature
at the end of the previous control period. Hence, ✓0 is not a decision variable in the optimization
problem and is assumed to be given. The inequality constraints in14ensures thermal comfort of the customers, where ✓min and ✓maxare the lower and upper limits of the thermal comfort zone given
in Table 4. The inequality constraint in 16is imposed to ensure that the total power consumed by each heat pump stays within ±10% of a reference value PHP
ref. The reference value corresponds
to the power consumption of a single heat pump in the absence of the DSM program. Assuming that in the absence of DSM, the inside temperature is kept at a constant value corresponding to the middle of the comfort zone,
✓ref = 0.5(✓min+ ✓max) (21)
the reference value for power consumption is given by PrefHP = T X t=1 ( A ⌘COP t )(✓ref ✓At) (22)
4 Model Implementation
This section provides the numerical data used for the case study. The technical specifications of the EES and heat pumps are provided in Sections4.1 and 4.2 respectively. Sections 4.3 and
4.6 describe the distribution system and the architecture of the smart grid hub in the InterFlex test area. The simulation cases, simulation software and key performance indicators are given in Sections 4.6, 4.7, and 4.8 respectively. The numerical data used for the economical analysis are provided in Section4.9.
4.1 Technical Specifications of the Electrical Energy Storage System
The operation principle of ESS was described in 2.1.1. The efficiency of ESS depends on the technology used, and varies between 0.7 and 0.98 [23]. The storage capacity and the rated power of the EES strongly depends on the manufacturing. The technical data of three commercial home battery systems are presented in Table2. For the simulations in this thesis, the averages of the parameters of these three models are used, presented in the last row of Table2.
Table 2: EES Technical Data EEES
r PmaxEES ⌘EES
[kWh] [kW] [%]
E.ON Aura 4.4 2.5 92
Tesla Powerwall 13.5 7 90
Sonnenbatterie 8.0 3.3 98
Test model 8.63 4.27 93.3
4.2 Thermal Parameters of the Heating System
The model used to calculate the indoor temperature as a function of the electricity consumption of the heater is given in Equation4. The time constant of the system, T C, and the overall thermal conductivity of the buildings, A, can be estimated by observing the thermal behavior of each building. Performing these tests, or obtaining the average value of these parameters for a large number of building falls outside the scope of this thesis. Instead, the thermodynamic parameters of one "reasonable" [36] building design are used. The values are taken from [36] and presented in Table3.
Table 3: Thermodynamic Parameters of Buildings
Parameter Symbol Value Unit
System time constant T C 25 h
Overall thermal conductivity A 0.14 kW/ C
In order to ensure thermal comfort of the customers, the indoor temperature must be kept within specified thermal comfort zones, given in Table4.
Table 4: Thermal comfort zones
Summer Winter
Indoor Air Temperature 23-26 C 20-24 C
The COP of a heat pump depends on the temperature difference between the two working environ-ments and is lower at lower outdoor temperatures [62]. The Swedish Energy Agency, Energimyn-digheten, performed a series of tests between 2009 and 2013 on 19 different heat pumps [63]. For each heat pump, COP was calculated for four different outdoor temperatures. The mean values of COP at each temperature for these 19 cases are presented in Table5. Figure2shows the COP for temperatures between -20 C and 12 C, obtained through linear interpolation and the data in Table5. To obtain the COP for a wider range of temperatures, it is assumed that the polynomial has constant slope between -20 C and -7 C, and between 2 C and 15 C.
Table 5: Average COP of Tested Heat Pumps
Outdoor Temperature 15 C 7 C 2 C 7 C
COP 2.27 2.53 2.79 3.48
Figure 2: COP of an Average Heat Pump as a Function of Outside Temperature
4.3 Test Network and InterFlex Customers
As described in Section 1.2, over 370 customers have volunteered to take part in the InterFlex project. The project consists of three use cases, two of which are applicable within the current regulatory framework. For these use cases, only 200 customers out of the available volunteers have been chosen. The reason for this is that InterFlex is a research project, where the goal was set to include 200 control points. However, the objective of Use-case 3 is investigating the potential of distributed flexiblities in providing ancillary services for the DSO. Since the incentives and the ethical aspects of this project have gained the interest of more than 370 customers, the simulation in this use case is based on the total number of volunteers.
To evaluate the effects of the proposed model, nine LV networks with relatively high density of InterFlex customers connected to an MV feeder have been chosen. The models described in Section 3.3 are applied to each low-voltage bus connected to the MV feeder, i.e. the DSM and EES scheduling is performed with the objective of decreasing the load-generation mismatch within each LV network. Table6provides an overview of the Installed PV capacity, number of household connections, and number of heat pumps and EES owned by InterFlex customers in each LV network.
4.4 Load and Generation Data
Table 6: InterFlex Low-Voltage Networks
Bus number PV Capacity[MW] Householdconnections HeatPumps EES
1 0.011 125 2 1 2 0.08 105 3 1 3 0.026 104 2 1 4 0.044 93 2 1 5 0.153 59 1 1 6 0.080 54 4 2 7 0.070 187 3 1 8 0.056 133 2 0 9 0.079 126 2 1
network. The power output of PV generators is modeled using historical data over PV generation in Germany [65], and scaled by the installed capacity in each LV network.
4.5 Smart Grid Hub Architecture
The infrastructure to enable the control of distributed flexibility consists of a Smart Grid Hub (SGH) and a set of metering and control devices on customer premises. This section provides a detailed description of the SGH architecture and its functionalities.
The SGH is an extension to the DS SCADA and is connected to the central grid control. The grid control center combines online measurements with static data, e.g. thermal rating of grid equipment, to create state estimates and identify potential imbalances and voltage or active power violations. If required, the control center sends requests to the SGH.
The SGH itself consists of two main parts; a process unit (PU) and a data unit (DU). Based on the requests of the control center and data provided by the DU, the PU generates the optimal set of control signals and transmits them to the control boxes on customer premises. The secu-rity, reliability and processing speed requirements for the PU are as high as for the control center itself. The functionalities of the DU are providing the PU with data from other sources within the company, such as equipment specification and weather data. The DU also receives and saves data from the PU. The security and speed requirements for the DU are lower than those for the PU. The customers are equipped with a Smart Meter (SM), a Smart Meter Gateway (SMGw), and a control box. The SM provides the SGH with online measurements of the controllable devices on-site, and also sends out confirmation signals to the SGH indicating the success or failure of a control command. The control box translates the SGH control signals into applicable commands for the DG or flexible load. Finally, the SMGw ensures the safety and reliability of the communi-cation between these devices.
The main functionalities of the SGH are as follows:
• Create group: This function aggregates the control points into groups. Each group is then assigned to a superordinate group, creating a hierarchical structure of the controllable devices. The grouping is performed dynamically, i.e. for each request that the control center sends out, the appropriate aggregation of the devices is determined and carried out. A measurement or switching signal that is received by a certain group, involves all the devices and groups subordinate to the receiving group.
• Measurement: The SMGw specifies what measurements are possible, e.g. voltage, active power, phase angle, and, as the name indicates, the measurement function carries out the measuring procedure. The measurements are collected periodically and/or at the request of the PU.
4.6 Simulation Cases
The system is simulated for three different cases, as described below:
• Base Case: this corresponds to the status quo of the system. It is assumed that all heat pumps are set to keep the indoor temperature at the reference value specified in Equation21
• Case A: the proposed model is applied to the devices that are currently available to InterFlex, as quantified in Table6
• Case B: based on an estimation performed by Avacon Netz, the number of heat pumps available to InterFlex corresponds to 8% of the total number of residential heat pumps in the test area. For Case B, it is assumed that all PV generators owned by residential customers are equipped with EES, and the proposed model is applied to all the heat pumps and EES devices available in the area. To estimate the number of heat pumps in each LV network, the total number of heat pumps in the test area is scaled with a factor of 12.5. The number of heat pumps is then distributed among LV networks with respect to the number of residential customers in each network. The total number of devices in each LV network is given in Table
7
Table 7: Devices Utilized in Case B
Bus number Heat Pumps EES
1 38 3 2 28 6 3 31 4 4 21 6 5 13 3 6 17 2 7 52 13 8 40 10 9 23 6
For all three cases, simulations are performed for the following three days: • Day 1: a typical weekday in spring
• Day 2: a typical weekday in winter
• Day 3: the day with highest PV generation. This occurred on a weekday in summer Each simulation is performed over a 24h period with 15-minute intervals:
t = 0.25h T = 96
4.7 Simulation Software
The simulations in this thesis have been performed using two different software. The nonlinear programming solver in GAMS modeling software is used to solve the optimization problem for the model described in Section3.3.2. MATLAB is used to simulate the model described in Section
3.3.1, and for presenting the results obtained with GAMS.
4.8 Key Performance Indicators
In order to evaluate the affects of the proposed algorithms, some key performance indicators (KPI) are chosen. The main expected result of the models is decreased peak power in the network. To evaluate this result, the net load Pnet of each network is chosen as KPI. The proposed model is
also expected to improve the load factor, given by [66]
Load F actor = Average Load
P eak Load (23)
4.9 Capital Expenditure
The investment cost of grid reinforcement has been estimated in [58] as investment per MW of system peak growth, and given in Table8. For conversion from USD to EUR of the values in Tables
8 and9, the conversion rate given by the European Central Bank [67] was used. For estimating the yearly investment costs, straight-line amortization has been applied, assuming an expected life of 40 years for power equipment, and 10 years for the smart grid.
Table 8: Average investment cost of grid reinforcement
Cost component Total cost [e/MW] Amortized yearly cost[e/MW]
Transformers and substations 110 657 2 766
Lines and feeders 687 267 17 182
Total 797 924 19 948
The investment costs of residential AMI have been estimated in [59]. Taking into account the economy of scale, the analysis in [59] has provided a high and a low estimate for the costs. Due to the relatively low number of customers in this simulation, and in order to avoid overestimating the economical benefits of DSM, the higher estimate is used here. The results are provided in Table9.
Table 9: Average Investment Cost of Demand Side Management Cost component Total cost [e/customer] Amortized yearly cost[e/customer]
AMI 163.21 16.32
Installation 11.66 1.17
5 Results
In this chapter, the numerical results of the simulations are provided. For Case A and B, the simulation results of both methods are presented in Sections 5.2 and 5.3 respectively. Since the results of the optimization approach are superior, the rest of the analyses is focused on the opti-mization approach only. The combined effect of locally balancing the nine LV networks is analyzed in Section5.6by looking at the loading of the MV feeder. The ability of the optimization model on reducing reversed power flow is evaluated in Section5.5. In sections 5.7 and5.8 a sensitivity analysis and an economical analysis are provided respectively.
5.1 Base Case
As described in the previous section, the Base Case represents the status quo of the system. However, the active load of each LV network only comprises of the residential load and the power consumption of the heat pumps. It is assumed that in the absence of the model, the indoor temperature is kept at ✓ref. The total number of heat pumps in each network is given in Table7.
Figures3a, 3b, and3c show the net load of the InterFlex buses during Day 1, Day 2, and Day 3 respectively. For all three days and all the buses, there is a distinct afternoon peak. It can also be seen that except for Bus 5 during Day 3, the net loads of the buses have a positive value, meaning that they do not cause reverse power flows. In order to evaluate the effects of local balancing on reverse power flows, Bus 5 is analyzed in more detail in Section5.5.
(a) Day 1 (b) Day 2
(c) Day 3
Figure 3: Net Load of InterFlex Buses, Base Case
5.2 Case A
5.2.1 Numerical Approach
The peak reductions caused by the proposed numerical model are provided in Table10. The peak reductions for each day are presented both as absolute reductions in [kW], and as a percentages of original peak. As see in the table, the peaks are reduced by up to 9.48%. The model is most effective during Day 1 which is a weekday in spring. It can also be seen that the peak reduction is smallest on Day 3 which is a weekday in summer. By looking at the results for Bus 8, which does not have any EES capacity, it is concluded that the amount of flexibility provided by EES is significantly higher than the flexibility provided by heat pumps. Since the percentage of peak reduction is highest for Bus 6, this bus is selected to visually demonstrate the affects of the model. The generation, load, and net load of Bus 6 are shown in Figure4.
Table 10: Peak Reduction in Case A, Numerical Approach
Day 1 Day 2 Day 3
Bus number [kW] [%] [kW] [%] [kW] [%] 1 12.15 3.54 10.76 3.13 9.83 2.91 2 11.29 3.97 10.13 3.57 7.33 2.84 3 11.24 3.95 10.08 3.54 7.76 2.83 4 10.49 4.19 9.71 3.89 8.54 3.60 5 9.25 5.82 8.02 5.08 6.98 5.36 6 13.96 9.48 11.36 7.70 10.71 8.51 7 14.73 2.89 12.37 2.43 11.94 2.46 8 0.16 0.04 0.18 0.05 0.12 0.03 9 11.58 3.45 10.80 3.24 8.66 2.76
(a) Day 1 (b) Day 2
(c) Day 3
Figure5 shows the time series of the differences caused in the net load of the InterFlex buses, given by
Pnet= PBaseCasenet PCaseAnet (24)
As it can be seen in the figures, except for Bus 1 during Day 3, the EES are only charged and discharged once during a 24h period.
(a) Day 1 (b) Day 2
(c) Day 3
Figure 5: Differences in Net Load, Case A, Numerical Approach 5.2.2 Optimization Approach
Table 11: Peak Reduction in Case A, Optimization Approach
Day 1 Day 2 Day 3
Bus number [kW] [%] [kW] [%] [kW] [%] 1 13.12 3.83 12.25 3.57 11.89 3.48 2 12.74 4.48 12.42 4.38 8.31 3.20 3 12.22 4.29 11.58 4.06 8.97 3.25 4 11.46 4.57 11.22 4.50 9.01 3.78 5 9.75 6.14 8.80 5.57 7.11 5.42 6 19.64 13.34 18.50 12.53 14.59 11.50 7 16.19 3.18 14.58 2.86 12.54 2.57 8 1.21 0.33 1.64 0.45 0.50 0.15 9 12.55 3.73 12.28 3.68 9.05 2.86
(a) Day 1 (b) Day 2
(c) Day 3
(a) Day 1 (b) Day 2
(c) Day 3
Figure 7: Differences in Net Load, Case A, Optimization Approach
5.3 Case B
In this section, both the models presented in Section3.3are applied to the InterFlex buses, taking into account all the EES and heat pumps in each LV network. The number of heat pumps and EES capacities simulated for Case B are given in Table7.
5.3.1 Numerical Approach
Table 12: Peak Reduction in Case B, Numerical Approach
Day 1 Day 2 Day 3
Bus number [kW] [%] [kW] [%] [kW] [%] 1 26.54 7.75 17.98 5.24 17.65 5.22 2 35.35 12.43 31.87 11.22 25.90 10.03 3 26.78 9.40 23.08 8.10 19.29 7.05 4 31.98 12.76 30.76 12.34 26.10 11.01 5 19.14 12.06 16.88 10.69 14.95 11.48 6 14.93 10.14 11.88 8.05 10.81 8.60 7 71.21 13.97 56.83 11.16 54.15 11.18 8 52.91 14.52 45.61 12.50 40.20 11.72 9 37.67 11.20 34.48 10.34 28.24 9.00
(a) Day 1 (b) Day 2
(c) Day 3
Figure 8: Load, Generation and Net Load of Bus 8, Case B, Numerical Approach
(a) Day 1 (b) Day 2
(c) Day 3
Figure 9: Differences in Net Load, Case B, Numerical Approach 5.3.2 Optimization Approach
Table13shows the peak reductions caused by the optimization algorithm. As it can be seen, the peak loads are reduced by up to 21.39%, which is significantly higher than the results obtained with the numerical approach. In this case, the peak reductions are highest during Day 2, i.e. the weekday in winter. The reason for this is the high number of heat pumps in the area, and the optimal activation of heat pumps as opposed to the numerical approach where a simple algorithm is used for heat pump scheduling. Even in this case, the lowest effect is obtained during Day 3. Figure10shows the load, generation, and net load of Bus 8.
Table 13: Peak Reduction in Case B, Optimization Approach
Day 1 Day 2 Day 3
(a) Day 1 (b) Day 2
(c) Day 3
(a) Day 1 (b) Day 2
(c) Day 3
Figure 11: Differences in Net Load, Case B, Optimization Approach
5.4 Comparison of the Proposed Methods
In both simulation cases, optimization proved to be more effective. The required computation time is also lower for optimization compared to the numerical approach. Due to the simplicity of the model, to include additional devices in the optimization only requires a model that estimates the power consumption of the device as a function of the optimization parameters. Including addi-tional devices in the numerical approach however requires a custom-made scheduling algorithm for that device. One advantage of the numerical method is that the EES are mostly only charged and discharged once during a 24h period. Compared to the optimization, which many times discharges and charges the EES twice throughout a 24h period, is an advantage since it saves battery life. Due to the higher effectiveness of the optimization, the rest of the results in this thesis are only presented for the optimization method.
5.5 Reverse Power Flow
Table 14: Reverse Power flow in Bus 5, Day 3 Peak [kW] Total power [kW] Duration
[minutes] Decrease inpeak [%]
Decrease in total reverse power[%] Base Case 32.17 336.04 255 Case A 28.09 296.66 255 12.69 11.72 Case B 20.00 163.54 225 37.70 51.33
(a) Entire Day (b) Hours with Reverse Power Flow
Figure 12: Net Load in Bus 5, Day 3
5.6 Loading of the Medium-Voltage Feeder
In this section, the effect of the optimization model on the MV feeder connecting the nine LV networks is evaluated. The peak reduction and the load factor of the MV feeder are provided in Table15. Note that the amount of peak reduction does not necessarily equal the sum of reductions in the LV networks because the LV networks do not reach their peak load simultaneously. As it can be seen, the peak load is reduced by up to 4.06% and 18.01% in Case A and B respectively. It can also be seen that the model improves the load factor of the feeder. Figure13shows the net load of the feeder for all three simulation cases.
Table 15: Peak Reduction and Load Factor of the MV Feeder Peak Reduction
[kW] [%] Load Factor
Day 1 Base CaseCase A 108.87 4.06 0.570.60
Case B 450.82 16.83 0.68
Day 2 Base CaseCase A 103.28 3.86 0.600.62
Case B 481.75 18.01 0.72
Day 3 Base CaseCase A 43.89 1.83 0.570.62
(a) Day 1 (b) Day 2
(c) Day 3
Figure 13: Active Power Loading of the Medium-Voltage Feeder
5.7 Sensitivity Analysis
In order to evaluate the influence of each parameter on the effectiveness of the model, a sensitivity analysis is performed. The sensitivity analysis is based on the loading of the MV feeder in Case A. The parameters EES capacity, and number of heat pumps are each separately increased by 25%, 50%, and 100%. The results are demonstrated in Table16. The values represent the additional peak reduction as a percentage of the peak in Case A. As it can be seen in Table16, the installed EES capacity has more influence on the effectiveness of local balancing than the number of installed heat pumps. As expected, the results show that the flexibility provided by EES is greater than the flexibility provided by heat pumps.
Table17shows the sensitivity of the model towards the lower and upper limits of the thermal com-fort zone, the constraint for charging the EES with locally generated power as given in Equation
19, and the power consumption limits for the heat pumps as given in Equation16. It is apparent that these constraints have negligible effect on the effectiveness of local balancing.
Table 16: Sensitivity Analysis of EES and Heat Pumps
25% increase 50% increase 100% increase
EES Day 1Day 2 0.53%0.51% 1.03%0.97% 1.96%1.78%
Day 3 0.40% 0.77% 1.33%
Heat Pump Day 1Day 2 0.11%0.17% 0.23%0.34% 0.46%0.68%
Day 3 0.05% 0.10% 0.20%
Table 17: Sensitivity Analysis of Thermal Comfort Zone, Local-Charging Constraint, and Range of Heat Pump Power Consumption
Event SimulationDay
(✓min 1) ✓t (✓max+ 1) Day 1 0.00% Day 2 0.00% Day 3 0.00% (✓min 2) ✓t (✓max+ 2) Day 1 0.00% Day 2 0.00% Day 3 0.00%
External Generation Used for EES Charging Day 1Day 2 0.00%0.00%
Day 3 0.00% 0.8· PHP ref PT t=1PtHP 1.2 · PrefHP Day 1 0.00% Day 2 0.00% Day 3 0.00% 0.7· PHP ref PT t=1PtHP 1.3 · PrefHP Day 1 0.00% Day 2 0.00% Day 3 0.00%
5.8 Economical Analysis
As described in Section2.3, utilizing distributed flexibilities is expected to decrease grid expan-sion costs, but also increase the investment cost of Advanced Metering Infrastructure (AMI). In order to analyze the net impact of the proposed model on CAPEX, these two cost components are compared for the MV feeder in InterFlex test area. Common practice in power system design is having a capacity high enough for the highest yearly peak. Hence, for each simulation Case we look at the highest peak among the simulated days. The highest peak for the Base Case and Case B occur on Day 1, while the highest peak in Case A occurs on Day 2.
The results of the analysis are obtained using the estimated costs in Tables 8 and 9 and are provided in Table18. P represents the reduced need for distribution capacity compared to the Base Case, which in turn results in the reduced yearly investment cost of grid expansion, Cexp.
The increase in yearly investment cost of AMI, CAM I, corresponds to the cost of equipping the
flexibility-providing customers with the required metering and communication infrastructure, and nf lx is the number of flexibility-providing customers. It can be seen that applying the proposed
Table 18: Analysis of the Yearly CAPEX for the MV Feeder
Pmax[MW] P [MW] Cexp[e] nf lx CAM I [e]
Base Case 2.6788
Case A 2.5712 0.1075 2 144 36 630
Part II
Regulatory Framework
The structure and the mechanisms in the current power sector stem from the assumption that power is generated in large controllable generators connected to the HV network and is consumed at LV and MV levels by customers with uncontrollable load [3,14]. This has resulted in a system where moving from the HV level towards the LV level the roles of the agents involved become increasingly passive. Due to the increased DG capacity connected to the LV and MV networks, and technologies such as smart metering and demand response, these paradigms are undergoing a major shift [68]. To comply with these changes it is necessary to adjust the commercial and regu-latory frameworks of network operation, to enable more active roles at the LV and MV networks where the challenges are increasing [4,11,55,57,69].
This part of the thesis addresses the commercial and regulatory aspects of utilizing distributed flexibility resources. Section6 provides a list of potential flexibility services. The type of DSM program used by Avacon Netz for the purpose of InterFlex is DLC, i.e. only bilateral contracts between the DSO and the end customers and no market mechanism are required. The German DSO regulations for utilizing distributed flexibilities are reviewed in Section 7. The alternative approach, i.e. the market-based mechanisms for trading flexibility services, are reviewed in Section
8, and a flexibility market structure is proposed in Section8.2.
6 Flexibility Services
To analyze the potential of distributed flexibilities and provide a basis for creating the platforms and regulatory framework needed to exploit these resources, a number of literature have provided a list of potential flexibility services [3, 4, 14, 70, 71]. The classifications of the services are not consistent among these sources. The classification provided below is the combination of services mentioned in [3,4,14]. Based on the main application purpose, flexibility services can be classified into system flexibility services and commercial flexibility services.
6.1 System Flexibility Services
Flexibility can potentially be provided to system operators (TSOs and DSOs), to improve system performance. These services include:
• Frequency control, i.e. primary, secondary, and tertiary reserves are, in present power systems, mainly provided by synchronous generators. However, the potential of distributed flexibility in providing frequency control has been demonstrated in [72–76].
• Congestion Management (CM) or active power control services can be activated by both DSOs and TSOs to ensure safe operation of network equipment. The application of dis-tributed flexibility for congestion management has been evaluated in [7, 31, 34,35,77] • Power Quality Control includes voltage control and loss reduction [14] and is managed
through reactive power provision. Historically, generation units have been connected to the high voltage grid, and power systems are designed based on the principle that moving from generation towards load, the voltage level decreases. But high penetration of RES connected to the low voltage (LV) grid goes against this principle and may cause reverse power flows and voltage instability in the distribution grid [5]. As demonstrated in [78–80], flexibility services can be developed to counteract this phenomenon.
• Local Balancing as a supplement to the conventional load-generation balance that takes place at HV level. Although not explicitly classified as a flexibility service in the available literature, utilizing local flexibilities to increase the local load-generation balance is the basis for some new models of power supply such as local energy systems and here proposed as a system flexibility service. Note that the local balancing proposed here does not involve frequency control, and is performed prior to or in conjunction with day-ahead-market closure. The main difference between Resource Optimization and Congestion management is the planning time horizon. While congestion management is performed using day-ahead or intra-day planning, Resource optimization requires a long term planning horizon of 1-5 years [83].
The above list is not exhaustive. Depending on the responsibilities of the system operator, the list might vary from one country to another. Also, new system operator responsibilities give rise to new system flexibility services. System flexibility services defined in the literature overlap and their effects on system performance are not mutually exclusive. For example, local balancing leads to improved resource utilization and/or congestion alleviation. But the main purpose of each service is unparalleled, leading to unique operation algorithms and economic values.
6.2 Commercial Flexibility Services
Flexibility services can also be used for portfolio optimization of
• Single customers, to maximize their RES feed-in and minimize their electricity procurement from electricity providers [3], or
• Commercial actors, i.e. wholesale suppliers and retailers, to gain competitive advantage [3,14].
7 Regulatory Framework for Direct Load Control
The regulatory framework for load management through DSOs is provided in paragraph §14a of the German Energy Industry Act (EnWG). Based on §14a, the DSO is obligated to offer decreased grid tariffs to electricity suppliers and low-voltage customers who grant the DSO the control over electric vehicles and/or flexible appliances with separate metering points. The specific technical framework for the exploitation of interruptible loads, e.g. triggering events and maximum allowable interference, have not yet been described. Neither has the contractual framework, e.g. the extent of grid fee reduction, been outlined. Forming the explicit framework for flexible-load utilization through the DSO is an ongoing project at the German Federal Government. Avacon Netz GmbH, among other system operators and market players, is included in the discussion and the study at hand will be used to demonstrate a potential set up for flexibility utilization.
The concept of local balancing is based on prioritizing local flexibility and generation resources over other sources of power. On the one hand, this violates the anti-discrimination principles of grid operation. On the other hand, the proposed algorithms seek to maximize local infeed. Since nearly all the DG in LV networks are based on renewable resources, local balancing is in line with the generation merit order and the DSOs’ obligation to maximize renewable generation. However, clear directives are required to eliminate vagueness.
Also of interest is the German Act for Operation of Metering Points (MsbG) that necessitates the smart-meter roll out in Germany by 2032. Based on MsbG, the DSOs are required to install AMIs for large customers, customers with interruptible loads or prosumers. The rest of the customers only need to be equipped with smart meters, without the data management and communication properties.
