Smart Curtailment for Congestion Management in LV Distribution Network

Transcription

1 Smart Curtailment for Congestion Management in LV Distribution Network A.N.M.M. Haque*, M. T. Rahman, P.H. Nguyen Department of Electrical Engineering Eindhoven University of Technology 5600 MB Eindhoven, The Netherlands *: F.W. Bliek DNV GL Energy 9704 CA Groningen, The Netherlands Abstract The rapid proliferation of distributed energy resources (DERs) leads to capacity challenges, i.e. network congestions, in the low-voltage (LV) distribution networks. A number of strategies are being widely studied to tackle the challenges with direct switching actions such as load shedding or power curtailment. On the other hand, various market-based demand response (DR) programs have been developed to influence the large number of DERs to use their flexibility to deal with network congestions. However, most of the market-based solutions rely on the flexibilities of the DERs, thus cannot solve the congestion when flexibility is not available in the network. To complement the market-based solutions, a smart active power curtailment based mechanism is necessary for managing the congestions in the distribution network. In this paper, we propose a novel method for congestion management by active power curtailment based on a Mixed-Integer Programming technique. In addition, two greedy selection methods together with fair power curtailment and security constrained OPF methods have been developed for the sake of comparison. The overall performance of the proposed approach and the comparison with other methods have been verified by a simulation with a typical LV network of the Netherlands. Index Terms Congestion management, graceful degradation, thermal overloading, curtailment cost. I. INTRODUCTION In recent years, the electrical power distribution network has been undergoing a radical shift towards an active distribution network (ADN). This paradigm shift to ADN is mostly influenced by the emerging generation technologies based on the renewable energy sources (RES) as well as the electrification of the transport and heating sectors. While enabling opportunities for large-scale penetration of DERs, this transition also increases the uncertainty and complexity in the distribution network operations [], [2]. For instance, different distribution system operators (DSOs) in Italy, Spain, Ireland and Germany with large share of installed RES-based distributed generation (DG) and long feeder lengths experience voltage limit violations in their networks [2]. Meanwhile, network congestions can be a further critical issue in more densely clustered networks like in the Netherlands [], [3]. Reinforcing these network assets necessitates huge amount of investment although the peak loads will generally occur only for few hours in a year [3]. Congestion management techniques in the distribution network can be roughly subdivided into two categories namely, direct and indirect control approaches [4]. The direct approach aims to mitigate the congestions by curtailment of load and local generation while the indirect approach influences the individual prosumers with price and/or incentive based Demand Response (DR) mechanisms. Among the direct approaches, a load shedding scheme based on system eigenvalues and bus modal participation factor is presented in [5]. A load curtailment scheme based on power flow tracing is proposed in [6], where the source of overloading and amount of load to be shed are determined using power tracing simulations on a real time basis. An undervoltage load curtailment method based on Particle Swarm-based Simulated Annealing (PSO-B-SA) optimization is applied in [7] that aims at ensuring the voltage stability in the distribution network. A GA based load curtailment scheme is presented in [8] that aims to minimize the amount of curtailed load and system losses. An intelligent centralized coordinated algorithm based load shedding approach is presented in [9] to relieve the network congestions while affecting the least number of consumers. Recently, Universal Smart Energy Framework (USEF) has been introduced as a conceptual approach to invoke the flexibility from end-users for various services including congestion management in the distribution network [0]. USEF aims to inherently combine the indirect and direct approaches of congestion management to maximize infrastructure utilization for the DSO. This enables the DSO to procure flexibility from a local capacity market to resolve the congestion, while a process of graceful degradation is activated to limit network access in terms of connection capacities. This assists the DSO to maintain the network reliability even when sufficient flexibility is not available in the network. An approach of realizing graceful degradation is proposed in [] through advanced power curtailment mechanisms. To obtain a synergy between direct and indirect control approaches for congestion management, this paper proposes a smart active power curtailment mechanism based on the USEF

2 process. The following sections discuss the formulation of the study in detail. First, an optimization problem is formulated to select suitable locations for active power curtailment by a Mixed-Integer Programming (MIP) method. The proposed method is evaluated against four selection rules. Finally, a case-study is performed and numerical results are explained. II. PROBLEM FORMULATION A typical LV network is designed and operated radially without any redundancy. Thus, any incident occurring at the MV/LV transformer, e.g. congestion when transformer loading S total exceeds the rated capacity S max, will affect all the end-users downstream of the transformer. To cope with such increasing occurrences, USEF describes a process so that the DSO can limit the network access for certain number of connected end-users to alleviate the load of the transformer in order to relieve the congestion. This can be done by curtailing active power consumptions at the connection points by the total curtailed power that can be expressed as follows: Scurtailed Stotal Smax () The problem for the DSO is to calculate the optimal amount of curtailment at each of the selected connection points considering the individual priorities. One way to address the problem is to define the priorities in terms of firm and non-firm capacity contracts []. A. Firm and Non-Firm Capacity In this paper, a flexible capacity contract is introduced to determine the curtailable power of the end-users with different levels of reliability for various types of connections. This can be realized by means of firm and non-firm capacities, defined by the following: Firm capacity is defined as the amount of power that is needed to be maintained at all times, as conventional contract between the end-users and DSO. Non-firm capacity is defined as the amount of power that can be curtailed during network emergency situations. This can be organized as an extra contract next to the conventional (firm capacity) contract due to available flexibility that the end-user can offer. Therefore, the total capacity at any connection point i can be expressed as the summation of the contracted firm and capacities as follows: firm Pi = Pi + Pi (2) For instance, if an end user has a rated connection capacity of 3.8kW based on a 6A single-phase connection, a contract with a firm capacity of 80% of connection capacity will enable the DSO to curtail maximum 20% of the connection capacity (760W) during network congestions. B. Minimizing the Curtailment Cost From the DSO s perspective, a curtailment cost is implemented to account for the penalty of the curtailed power. This curtailment cost can be related to the contracted firm capacity of the connection point, as an end-user with higher firm capacity opts for a greater level of required reliability compared to a consumer with a lower firm-capacity. Once congestion is expected in the network, a curtailment plan will be developed based on the curtailment costs and the capacity contracts. The connection points can be selected for curtailment either by calculating the optimal amount of curtailment to minimize the total curtailment cost or based on predefined selection rules as explained in the later subsections. The optimization problem is formulated considering the radial topology of the LV network as illustrated in Fig.. Each of the load buses is considered to represent the aggregated load of certain number of households. The MV/LV transformer is considered to be comprised of NF number of LV feeders. The decision variable, u i selects the feeders to curtail the active power and is binary in nature. The optimization problem can be mathematically presented as: subject to, NF ui. Ci + Pij, curtail Cij i NF, j i = j = (3) min (. ) { if feeder i is selected u i = 0 otherwise (4) Pij, curtail Scurtailed (5) j= 0 P u. P (6) ij, curtail i ij, where, : number of buses in i-th feeder; P ij, curtail : curtailed power at j-th bus of i-th feeder; P ij, : non-firm capacity at j-th bus of i-th feeder; C i : curtailment cost of i-th feeder; C ij : curtailment cost of j-th bus in i-th feeder. Eq. (4) (6) represent the constraints of the optimization problem. Constraint in eq. (5) is imposed to limit the curtailment to keep the loading of the transformer close to the nominal rated value, while eq. (6) ensures that only the capacity of the selected feeder is considered for the curtailment in corresponding buses. The optimization problem described in (3) (6) consists of both integral and binary variables and can be solved by a mixed integer programming (MIP) method. C. Greedy Selection Rules Beside the MIP optimization formulation, some other greedy selection rules have been introduced in this paper for the sake of comparison. These selection rules are discussed as follow: ) Selection Based on Non-Firm Capacity in the Feeder The aim of the method is to incorporate the minimum number of feeders to resolve the congestion. Whenever a feeder is selected, all the buses in the selected feeder will be considered for curtailment. This process is deployed by the following steps:

3 Utility grid MV Stotal LV 2 NB- NB P 2 P2 NB2- P2 P22 P2NB2 2 - NB2 Pi Pi2 Pi Fig.. Radial topology of the LV distribution network. Step : Selection of the feeders is performed by ranking available non-firm capacities in all the feeders. It is assumed that the feeders and buses having higher amount of non-firm capacity cause lower curtailment cost if selected for curtailment. Step 2: Regarding a selected feeder, if the available capacity of the selected feeder is higher than the required amount curtailment, an optimization problem is formulated, as shown in eq. (7), to minimize the total curtailment cost within the feeder. Linear programming is used to determine the optimal values of curtailment from the individual buses of the selected feeder. Actually, the optimization problem is a simplified form of (3) and can be reformulated as: subject to, min Pij, curtail. Cij j j = PNB2 (7) Pij, curtail Scurtailed (8) j= 0 Pij, curtail Pij, (9) As shown in eq. (8) and (9) the constraints limit the amount of curtailment in the buses within the available non-firm capacity while the total curtailment in the network must be enough to solve the congestion. Step 3: If the total available non-firm capacity of the selected feeder is curtailed and the congestion is yet to be relieved, the next feeder with the highest non-firm capacity is selected and the process continues till the congestion is fully resolved. 2) Selection Based on Power Flow in the Feeder The feeders can be selected for curtailment by ranking them based on the flows revealed by an initial load-flow calculation. The feeder having the maximum loads contributes more to the congestion and curtailing active power in the feeder can resolve the congestion efficiently. Once the feeders are selected the problem can be solved using the same methodology as discussed in step 2 and step 3 for the selection of feeders based on the available non-firm capacity following eq. (7) - (9). 3) Fair Power Curtailment The objective of this method is to incorporate all the buses in the network in a fair curtailment process. The fair approach of curtailment is deployed as follows: Step : A curtailment co-efficient is defined that expresses the share of the non-firm capacity of a bus in the total capacity available in the network. The coefficient is calculated by dividing non-firm capacity of the bus to the total non-firm capacities available for the whole network. P ij, being the non-firm capacity of j-th bus in the i-th feeder, the curtailment co-efficient, w ij is given by, Pij, wij = NF (0) P i= j= ij, Step 2: The total required curtailment is distributed among all the buses of the network based on the curtailment coefficient. The required curtailment for each of the buses is given by, P = w S () ij, curtail ij. curtailed 4) Security Constrained Optimal Power Flow (OPF) A security constrained OPF problem can be formulated to calculate the required curtailment for each of the buses with an objective of supplying the maximum amount of loads while ensuring the security constraints of the network. In other words, the objective of the OPF problem is to maximize the social benefits considering the rated thermal capacities of the transformer and the cables []. If P Dij and P Sij are the demanded and injected power at j-th bus of i-th feeder respectively, the objective of the OPF can thus be formulated as: NF NB ( i CDij. PDij CSij. PSij i= j= j = (2) min ) where, CDij and CSij are demand and supply costs for the j-th bus of i-th feeder respectively. The OPF problem is subjected to the constraints illustrated by eq. (3) (6). Eq. (3) and (4) denote the power flow equations and the power flow limits of the components respectively, while eq. (5) and (6) denote the limits of the supplied and demanded power. g( δ, V, Q, P, P ) = 0 (3) G S D Pijj ' Pijj ',max j, j' i (4) 0 PSij PSij max 0 PDij PDij max (5) (6) III. CASE STUDY A. Description of the Test Network A case study is performed for evaluating the proposed methodology in a Dutch LV residential network as illustrated in Fig. 2 considering full penetration of solar PV, heat pump (HP), EV along with uncontrolled loads. A 250kVA MV/LV transformer feeds the radial LV network from the MV bus. Each of the load buses in the network represents the aggregated load of 0 households. The uncontrolled base loads can be represented as aggregated loads of residential consumers and modeled using normalized profiles. Actual consumption from such normalized profiles of 400 households are calculated by

4 multiplying them with 3400kWh that represents the average annual energy demand per household in the Netherlands [2]. Behaviors of the individual flexible loads like heat pump and EV and local generation technologies like solar PVs are modeled according to the functionalities used in [3]. Information related to the case study are summarized in Table I. Upon observing the overload of the transformer from loadflow calculations, the curtailment methods are run separately for the time instants when congestion is occurring. The curtailment methods and selection rules can be organized for the case study as follows: Case : Selection based on non-firm capacity in the feeder Case 2: Selection based on power flow in the feeder Case 3: Fair curtailment Case 4: Security constrained OPF Case 5: Selection based on MIP optimization As illustrated in Table II, four different cost categories are considered as curtailment costs of the buses and three categories are considered for each of the feeders of the network shown in Fig. 2. The cost categories are formulated based on the available non-firm capacities of buses and feeders. A higher amount of non-firm capacity corresponds to a lower priority of load which is reflected by a lower curtailment cost. B. Simulation Setup The simulation is performed for a typical winter week in February which is usually the coldest time of the year. The network model is constructed in MATLAB environment using the Power System Analysis Toolbox (PSAT) []. PSAT is also used for performing the load flow calculations to observe the loading conditions in the network. The MIP optimization problem is formulated in MATLAB environment using YALMIP, which is an open-source MATLAB toolbox that supports with rapid development of complex optimization algorithms [4]. GUROBI optimization solver is used to solve the problem defined by YALMIP. The OPF problem is solved with the security constrained OPF routine of the PSAT that uses an interior point method (IPM) optimization solver. The linear programming (LP) optimization problems of the remaining cases are solved by the LP function of the MATLAB optimization toolbox. TABLE I. PROPERTIES FOR SIMULATION Property Values Transformer rating 0kV/0.4kV, 250kVA Transformer R0, Z Ω, Ω Base power for load-flow 250kVA Installed PV capacity per house kwp Charging rate of EVs 3kW/hour Temperature range for HPs 8 C -22 C Capacity of the HPs 3.5kW TABLE II. COST CATEGORIES FOR CASE STUDY Category C ij (/kw) C i ( ) A 5 8 B 5 8 C D 30 Cable types A: 50mm 2 Al B: 95mm 2 Al C: 50mm 2 Al D: 6mm 2 Al E: 0mm 2 Cu F: 6mm 2 Cu External network 0kV m, B m, C 2m, E Feeder 28m, C kV 35m, A m, A 40m, B 204m, B 42m,B 0m,E 3 24m, C 5m, F 40m, B Feeder m, B 20m, D m, A 320m,B m, D 80m, A 69m, A Feeder m, E 9 8m, D 2 Fig bus LV network model for simulation. IV. NUMERICAL RESULTS The case is run for one week with a time step of 5 minutes. First, the loading of the transformer is observed without any congestion management measures. Next, the cases are implemented to curtail the loads at the buses. The results are validated with another load-flow calculation. The following sections discuss the key-findings more in detail. A. Transformer Loading The loading of the transformer before and after the curtailment are shown in Fig. 3. After running the initial loadflow, total 9 time steps are observed when the transformer is overloaded by certain margins. The curtailment mechanisms are therefore implemented for all the 9 steps. The constraints in eq. (5) and (8) allow the methods to keep the loading of the transformer close to the nominal capacity. B. Curtailment Cost A quantitative comparison among the calculated curtailment costs is presented in Table III. The results are tabulated for all the cases in terms of the total curtailment cost, total number of buses involved in curtailment and number of time steps when each feeder is used for curtailment. It is evident that the curtailment plan based on proposed MIP optimization outperforms the other selection rules. Case 2 and Case 3 represent relatively higher curtailment costs. This is consistent with number of buses involved for curtailment for these two cases. For Case 2, feeder is chosen for curtailment due to the maximum power flow among all three of the feeders. The total number of buses in feeder is also higher than the other two. For Case 3, all of the buses in the network are considered for curtailment which is responsible for the highest curtailment cost among all the cases. TABLE III. RESULTS OF THE CASE STUDY Number Total time steps of curtailment Curtailment Case of buses Cost ( ) Feeder Feeder 2 Feeder 3 involved Case Case Case Case Case

5 .2 Power (p.u.) No. of days Rating Initial load Case Case 2 Case 3 Case 4 Case 5 Fig. 3. Transformer loading. The OPF routine minimizes the loss over the distribution cables and curtails the non-firm capacities completely at the buses located at the end of all the feeders while supplying the full capacities to buses at the beginning of the feeders. This results in a higher curtailment cost compared to Case 5. The MIP optimization algorithm uses the optimum combination of the buses and feeders based on the individual curtailment costs. The inclusion of the binary decision variable lets the algorithm to check for the buses individually for the corresponding feeders. C. Effects on Network Loss Total network loss for the simulated week for all the cases are calculated and compared with the losses for initial loading. As shown in Fig. 4, network losses are reduced for all five curtailment cases. The notable difference in network losses for Case 4 compared to other cases is due to the variation of curtailment process for the OPF. The OPF routine ensures that the maximum amount of load is supplied satisfying the imposed constraints. To do so, it curtails the non-firm capacities completely at the buses furthest from the transformer. This facilitates a reduced flow of power through the cables thus resulting in a lower active power loss. For all the other cases, the non-firm capacities of the buses are partially curtailed and the resulting losses become higher than Case 4. V. CONCLUSIONS The focus of this paper has been to identify the possibilities of a smart active power curtailment mechanism to complement the market-based approaches to tackle the network congestions. From the DSO s perspective, a Mixed- Integer Programming based curtailment algorithm is proposed to select suitable buses for curtailment in a radial LV network. The proposed method is evaluated against a number of different selection rules simulated in a case study of a Dutch LV network. The simulation results indicate that an MIP based optimization method can effectively improve the selection process of the buses for active power curtailment. Fig. 4. Network loss reduction compared to the initial loading. The future research in this topic will be directed to a more detailed analysis with a bigger network in order to validate the performance of the tool for large-scale network simulations. REFERENCES [] A. N. M. M. Haque, P. H. Nguyen, and W. L. Kling, Congestion Management with the Introduction of Graceful Degradation, in PowerTech, 205 IEEE Eindhoven, 205, pp. 6. [2] A. N. M. M. Haque, P. H. Nguyen, W. L. Kling, and F. W. Bliek, Congestion Management in Smart Distribution Network, in Power Engineering Conference (UPEC), th International Universities, 204, pp. 6. [3] R. A. Verzijlbergh, L. J. De Vries, and Z. Lukszo, Renewable Energy Sources and Responsive Demand. Do We Need Congestion Management in the Distribution Grid?, IEEE Trans. Power Syst., vol. 29, no. 5, pp , 204. [4] S. Huang, Q. Wu, Z. Liu, and A. H. Nielsen, Review of congestion management methods for distribution networks with high penetration of distributed energy resources, in IEEE PES Innovative Smart Grid Technologies, Europe, 204, pp. 6. [5] A. Arief, M. B. Nappu, Z. Y. Dong, and M. Arief, Load curtailment strategy in distribution network with dispersed generations, in Universities Power Engineering Conference (AUPEC), 20, pp. 6. [6] M. R. M. Dahalan, H. Mokhlis, A. H. A. Bakar, and N. M. Sapari, Overload Alleviation Scheme Based On Real Time Power Flow Tracing In Distribution Network, in Power Engineering and Optimization Conference (PEOCO), 204 IEEE 8th International, 204, no. March, pp [7] N. Sadati, T. Amraee, and A. M. Ranjbar, A global Particle Swarm- Based-Simulated Annealing Optimization technique for under-voltage load shedding problem, Appl. Soft Comput. J., vol. 9, pp , [8] A. R. Malekpour, A. R. Seifi, M. R. Hesamzadeh, and N. Hosseinzadeh, An optimal load shedding approach for distribution networks with DGs considering capacity deficiency modelling of bulked power supply, in 2008 Australasian Universities Power Engineering Conference, 2008, pp. 7. [9] D. Panasetsky, N. Tomin, D. Yang, and V. Kurbatsky, A New Intelligent Algorithm for Load Shedding Against Overload in Active Distribution Networks, in Power System Technology (POWERCON), 204 International Conference on, 204, pp [0] An introduction to the Universal Smart Energy Framework, USEF Summary, 204. [Online]. Available: [] F. Milano, An open source power system analysis toolbox, IEEE Trans. Power Syst., vol. 20, no. 3, pp , [2] Energie Data Services Nederland. [Online]. Available: [3] J. A. W. Greunsven, E. Veldman, P. H. Nguyen, J. G. Slootweg, and I. G. Kamphuis, Capacity management within a multi-agent market-based active distribution network, in IEEE PES Innovative Smart Grid Technologies Conference Europe, 202, pp. 8. [4] J. Lofberg, YALMIP: a toolbox for modeling and optimization in MATLAB, 2004 IEEE Int. Conf. Robot. Autom. (IEEE Cat. No.04CH37508), pp , 2004.