Sei sulla pagina 1di 7

1

Incorporating Control System Reliability in Active Management of Distribution Systems with Distributed Generation
Kang Ma, Student Member, IEEE, and Joseph Mutale, Senior Member, IEEE
emerge associated with the unprecedented changes in operational arrangements. Traditionally power flows in one direction in the distribution network, from bulk supply point to end user; with increasing penetration of distributed generation (DG), power flows become multi-directional causing new problems, such as overvoltages, which impose limitations to DG penetration level [1]-[3] [10]. Besides the power flow and voltage rise challenges, there are also new challenges in the power quality management and the fault level management [11]. Various approaches have been proposed to address the problems which can be put into two main categories. The first category includes all methods that advocate changing the basic operation philosophy of distribution system from passive management to active management (AM) and the second includes traditional network reinforcement approaches in which sufficient system margin is provided in advance within a passive network operation philosophy also referred to as fit-andforget approach. The impact of AM on the level of DG penetration is summarized in [4] and it is discovered through various sensitivity analyses that the DG penetration level increases with the increase of network flexibility [4]. The structure and control strategy of AM is described in detail in [4] including quantification of the increase in DG penetration due to different AM scenarios. The technical and economic impacts of AM are investigated in [10] from various perspectives including the voltage profile, line losses, power generation and net benefit. The transition from traditional passive network to AM is not a one-off construction but rather an accumulative process with gradual increase of network flexibility [11]. The AM system consists of four essential parts with different functions: measurement and data analysis part (the Supervisory Control And Data Acquisition or SCADA system), decision making part which makes real-time decisions based on the feedback from measurement and data analysis part, communication part which transmits the information among different parts, and local control devices which execute the decision from the decision making part by adjusting the set points of tap changing transformer and reactive compensation, changing the output of DG, and performing other corrective actions if necessary [11]. However, in all these previous works, AM is the preferred choice with the underlying assumption that its control system is 100% reliable. As no control system is 100% reliable such

Abstract-- Several factors affect the viability of Active Management (AM) schemes in distribution systems with distributed generation. One of the critical factors is the control system reliability. While a lot of work has been done on the technical and economic aspects of active distribution network management, almost no attention has been paid to the impact of the reliability of control and communication systems on the expected benefits of active management strategies. This paper investigates the impact of control system reliability on the benefits of AM. The reliability of the control system is assessed through a state enumeration approach, with different load levels. The results for the specific test case suggest that the load level and AM control system reliability should be jointly taken into account for a comprehensive analysis on the overall cost benefit of AM compared to Traditional Reinforcement Scheme (TRS), and that under each load level the economic benefit of AM improves with the improvement of AM control system reliability. Index TermsPower systems, Reliability, Wind power generation, Static Var Compensators

he global drive to mitigate the now widely recognized negative effects of CO2 on the world climate has lead to CO2 reduction targets being set under the Kyoto protocol. To fulfill its obligations under the EU Kyoto targets, the UK government has set very ambitious targets of its own which will require installation of 8GW-10GW of renewable generation and 10GW of CHP schemes by 2010 [1]. If these targets are to be achieved, revolutionary changes in planning and operation of distribution networks will be required [2]-[6]. There are quite a few published works focusing on the economic aspects of different types of Distributed Generation (DG). Reference [5] quantifies the impact of DG on environments as part of the overall economic assessment; reference [7] creates the business model of DG. Results from these works have shown that DG is potentially beneficial in the economic sense, as well as being a promising means to address environmental concerns. However, new problems
This paper was supported by Supergen Flexnet, which is a consortium under Engineering and Physical Science Research Council, UK. K. Ma is a PhD student with Electrical & Electronic Engineering Department, University of Manchester, Manchester, M14 6GB, UK (e-mail: Kang.Ma@postgrad.manchester.ac.uk). J. Mutale is a senior lecturer with Electrical & Electronic Engineering Department, University of Manchester, Manchester, M14 6GB, UK (e-mail: J.mutale@ieee.org).

I. INTRODUCTION

an assumption may lead to overly optimistic decisions about the benefit of AM. Two questions should be thoroughly investigated before the final system development strategy is decided: 1) whether AM is always preferable to network reinforcement schemes in terms of costs and benefits; 2) under what circumstances AM is preferable. This paper explores one of the crucial factors that have a significant impact on the viability of AM, namely control system reliability. A methodology is proposed to compare the AM and reinforcement schemes on a characteristic test case from the perspective of economic benefits incorporating the risk of control system failure. The paper will suggest best reinforcement scenarios for each system with the same network topology but different parameters. Sensitivity analysis will be performed to quantify the impact that determines the viability of AM. II. BASIC TECHNIQUES OF RELIABILITY ASSESSMENT A crucial factor that determines the viability of AM is the reliability of the control system. This paper applies well known reliability assessment methodology and techniques to evaluate the reliability of the control system. Based on this approach risk events are defined for two cases (i) when wind generation is high and demand is low leading to overvoltage problem and (ii) when the wind stops blowing and demand is at peak giving rise to undervoltage problem. When the voltage violations occur, corrective actions are taken to eliminate the violations without DG and load curtailment or to minimize the consequences of such curtailment if unavoidable. The possible failure of the control system poses another threat to system reliability. The whole process of reliability assessment and basic assumptions and techniques have been developed for many years and many papers have been published in this field, where Roy Billinton and Ronald Allan have done a lot of pioneering research [8]. The well known techniques developed by these authors are applied under a new set of conditions in the paper. The failure of components is assumed to follow exponential distribution. By studying the physical structure and behavior of components, state space model is created and transition rates assigned. One of the simplest models is the two-state model consisting of only up and down states. Different components are assumed to be independent to simplify the analysis. State enumeration and Monte Carlo Simulation (MCS) are two basic techniques. Both techniques have their pros and cons. State enumeration is straightforward but the full enumeration is usually not possible because of the prohibitive number of system states, or combinatorial explosion. Hence only some of the system states satisfying a predefined criterion are considered. MCS are capable of dealing with complicated systems and considers all possible events, but it imposes a heavy computational burden. State enumeration method is applied in this work. The corrective control actions are not predefined and should be generated automatically in real time. Hence an

optimization with the objective to minimize the user defined objective function subject to power balance and other system constraints is applied. The objective function expresses the cost of different actions and distinguishes between them by assigning different weighting factors. Depending on the power flow model, the optimization can be either based on DC load flow or AC load flow. The optimization using DC load flow is simpler and more efficient than that using AC load flow. However, the former is not capable of dealing with voltage variations. The AC load flow is more accurate and yields more comprehensive results. Hence an optimization based on AC load flow is applied in this work. III. TECHNICAL AND COST BENEFIT ANALYSIS OF SYSTEM REINFORCEMENT SCHEMES The methodology to quantify both the technical and commercial impacts of AM and traditional reinforcement schemes on distribution networks is presented below. A. Technical analysis of Active Management Scheme Active management in essence runs under corrective control mode, based on FACTS devices, on-load tap changers, DG output control (DG curtailment control) and the AM control system. When potential system violation occurs (such as bus over voltage or line overflow), corrective control actions are generated by the control system, transmitted through communication system (control system and communication system is collectively called control system in this paper) and executed by local control devices to eliminate violations. Different types of corrective actions are given different priority based on cost and time to respond. The actions from highest to lowest priority are: SVC (Static Var Compensator) adjustment, OLTC (On-Load Tap Changer) adjustment, DG curtailment and load curtailment. SVC adjustment is the fastest to respond when violation occurs, normally with a response time far below one second. DG curtailment ranks the second in speed. However, DG curtailment is uneconomic due to the potentially large opportunity cost of energy not generated, thus its given a lower priority than OLTC adjustment. OLTC takes minutes to respond but costs little to operate. The paper assumes that the system can withstand violations for the short period required by OLTC response. The least favorable action is to shed load, which will incur a large social cost.

Fig. 1. The whole structure of the control system and executive devices.

The whole structure of the control system including control devices is shown in Fig. 1 where the control system is the brain whose function is to generate real-time control actions (namely, to determine the set points of control devices) according to the real-time voltage feedback information. Such structure is consistent with that in [2]. In this structure, the failure of any local control device (SVC, DG output control and OLTC) does not affect the operation status of the control system, which means, by way of analogy, a disabled hand does not affect the functionality of the brain. Moreover the failure of any control device does not affect other devices or, in other words, the paralysis of one limb does not affect other limbs. However, all devices cease to operate when the control system fails. Failures of other components such as power lines and transformers are not considered so that the impact of control system unreliability can be quantified separately without distortion by other components. The state enumeration method is applied to evaluate the expected Generation Curtailment and Unserved Energy (from load shedding). Ideally, all possible states, including all overlapping failures should be enumerated to achieve an accurate result. However, since the number of system states is very large and the probability of overlapping failures is negligible, only first order failures are considered: the single OLTC, SVC, DG output control, and control system failure is considered. All states considered are mutually exclusive so that the probabilities can be simply summated. When all control devices and the control system are in the up state, the optimization model based on AC load flow is expressed in Model I:
Cur Cur min (Tk 1) 2 + 1 PGi + 2 PLi k i i

PLi , QLi PGi , QGi


Cur PGi

Active and reactive load at bus i Active and reactive generation at bus i Active generation curtailment at bus i

Cur Cur PLi , QLi Active and reactive load curtailment at bus i

Pi inj , Qiinj Active and reactive power injection at bus i


Tk X SVC
Tap setting of on load tap changer k Equivalent reactance of SVC Load flows at branch ij Voltage magnitude and angle at bus i

Sij
Vi , i

pf min , pf max Min and max power factor allowed


A large implies a low priority that the particular term is assigned. To clearly distinguish the priority level,

and

should not be too close to each other. SVC adjustment is not represented in the objective function because it is always preferable within its limit. For example, for a typical solution with 1 =100 and 2 =10000, the change of OLTC tap (represented by the deviation of T from 1, the nominal tap) is not amplified at all, whereas DG curtailment is amplified by 100 times and load curtailment by 10000 times. Thus, a small increase in DG curtailment will result in considerable increase of the objective function value which is to be minimized. The impact is even greater when increasing load curtailment. In that way, the algorithm favors SVC adjustment over OLTC adjustment, DG curtailment and load curtailment in decreasing order. When OLTC fails and all other components are in normal state, the OLTC is simply bypassed. The model is similar to Model I with the only difference being that all terms including the tap setting variable T are omitted. When the SVC is down and all other components remain up, the SVC branch is an open circuit. Similarly, terms with SVC equivalent admittance are not considered. When the control system fails, corrective control is suspended: OLTC, SVC and DG partial curtailment are not available. OLTC and SVC are bypassed in the model and DG is completely disconnected in the case of any system violation. In such a case, loads are subject to curtailment when demand is high and DG is disconnected. DG power factor is assumed to be fixed at 0.95 lagging. The load shedding optimization is called only in the case that the system violation cannot be eliminated by DG curtailment, since such action will incur significant cost. Half hourly DG curtailment and load curtailment over a year are calculated for each system state, from which the set of energy indices including DG Energy not Generated (ENG), DG Energy Generated (EG) and Energy not Served (ENS) are derived. ENG indicates the amount of energy curtailed due to system constraints and EG is the complement of ENG. Energy not Served represents the energy curtailed at load points.

s.t PGi PLi P


max Sij < Sij

Cur Gi

+P

Cur Li

= Pi (V , , T , X SVC )
inj

(1) (2) (3) (4) (5) (6) (7)

QGi QLi + Q

Cur Li

= Q (V , , T , X SVC )
inj i

Vi min < Vi < Vi max 0<P


Cur Gi

< PGi

Cur 0 < PLi < PLi

pf min <

Cur PGi PGi < pf max Cur ( PGi PGi ) + j QGi )

Cur PLi PLi P = Li Cur QLi QLi QLi


min max X SVC < X SVC < X SVC

(8) (9) (10)

Tkmin < Tk < Tkmax


where

1 , 2 > 1

Coefficient corresponding to priority level

4
Cur ENG = t PDG (ti ) i =1
Cur EG = t ( P (ti ) PDG (ti )) Wind i =1 i =n

i =n

(11) (12) (13)

NB = AR Capital DG Operation DG Cost loadShedding ACCscenario AOCscenario

(18)

The goal is to maximize NB. Since ACCscenario and

AOCscenario are assumed to be independent from DG


capacity, energy generated and energy not served, the problem can be broken down into two sub problems. First sub problem is to find the optimal DG capacity and second is to select the most cost efficient system reinforcement scenario. C. Traditional Reinforcement Scheme (TRS) Active management runs under corrective mode, which means that it tries to eliminate violations after they occur, whereas TRS aims at providing sufficient system margin in advance; that is at the planning stage. In a typical distribution network with DG, buses to which DG is connected (DG bus) are prone to overvoltage problems. Other buses that are not along the same radial branch as DG buses but fed from the same grid supply point may suffer from unacceptable voltage drop. In theory, a possible solution is to increase the impedance of lines connecting DG bus, at the same time to decrease the impedance of lines connecting buses where the problem of voltage drop is more likely to happen. In practice, that means in some places to invest in parallel lines and in other places to replace existing lines or to install series reactors. Under traditional passive mode of operation, no corrective actions are available: DG cannot be partially curtailed. SVC or OLTC is not available. Although the concept is far from that of active management, the optimization model, given below in Model II, is quite similar: Model II:
Cur min PLi i

ENS = t PLCur (ti )


i =1

i =n

The overall result is the aggregation of the result from each state weighted by the probability of that state. B. Cost Benefit Analysis of Active Management Scheme The cost benefit analysis, based on the overall result, requires indices such as ENG, EG and ENS. Annual revenue of DG (AR) is calculated from (14): (14) AR = (DG + ROC ) EG In (14)

DG

is the energy price in (/MWh) which is

assumed to be constant and

ROC

is the income of DG from

the sales of Renewable Obligation Certificates (ROCs). One ROC is issued to an accredited renewable generator by OFGEM (Office of Gas and Electricity Markets) for each MWh of energy generated. DG makes a profit by selling ROCs to electricity suppliers who need ROCs to meet their renewable obligation or face a penalty charge. In this paper, ROC is not considered. Therefore, ROC = 0 . The economic assessment is based on annualized capital costs of DG, additional lines and investment in active management system. In reality, the capital cost of DG depends on a lot of factors including the DG capacity, the location and policy etc. It is rather difficult to summarize a universal formula to calculate DG capital cost. In this paper, the DG capital cost has been linearized and is assumed to be proportional to DG capacity as given in (15). CapitalDG = kDGCapital Capacity DG (15) In (15)

Cur s.t PGi PLi + PLi = Pi inj (V , )

(19) (20) (21) (22) (23) (24)

Cur QGi QLi + QLi = Qiinj (V , )


max Sij < Sij

kDGCapital (/MW) is the DG unit capital cost The

Annual DG operation cost is modeled in a similar way (16): Operation DG = kDGOperation Capacity DG (16) The cost of possible load shedding is assumed to be directly proportional to the energy not served. Cost loadShedding = kloadShedding ENS (17) It should be noted that under the assumption that network components (lines, buses and generation) failure is not considered, only AM control system failure may contribute to ENS. Thus, ENS is zero under traditional reinforcement scheme and so is the cost of load shedding. Besides, every scenario (Active management and traditional reinforcement plan) incurs an annualized capital cost ( ACCscenario ) and operation cost ( AOCscenario ). The net benefit ( NB ) of every scenario is calculated as follows:

Vi min < Vi < Vi max


Cur 0 < PLi < PLi Cur PLi PLi P = Li Cur QLi QLi QLi

Several network reinforcement plans are proposed. The optimization runs for every half hour of the year and for every plan. For every plan, the optimization should provide a feasible solution for every half hourly load and generation. If there is no feasible solution for all half hourly load and generation, such a plan is considered to be unfeasible and should be abandoned. IV. CASE STUDY AND DISCUSSION The test case is based on the work reported in [4]. It is a

typical 6 bus distribution network with wind generation. The network diagram is presented in Fig. 3:

vary within 0.9 to 1 (absorbing vars). The income from ROCs is not considered in this paper ( ROC = 0 ). Under active management, wind generation is curtailed in the periods when wind generation reaches its peak output, load is relatively low and the system state allows the corrective control action to be performed. Annual wind energy output and curtailment is calculated for each load case. The results obtained are consistent with the expectation that DG curtailment decreases when load level increases. The DG energy curtailment for the three load cases are shown in Fig. 4, Fig. 5 and Fig. 6, respectively:

The diagram and network parameters are the same as in [4] and the annual wind generation and load data is obtained from [9]. The 33kV distribution system is fed from a 132kV bulk supply point at bus 1. Bus 1 and 2 are connected through an OLTC transformer. Loads are connected to bus 2, 3, 4 and 5. Wind generation with SVC is connected to bus 6. Different types of loads, including residential, commercial and the industrial, are allocated to different buses. Half hourly load and wind generation data is obtained for a period of one year. At a specific time, the corresponding load and wind generation forms the snap shot of the system at that moment. Under active management, three kinds of corrective actions are considered and are listed in decreasing order of priority: I) Area based voltage control by OLTC between bus 1 and 2; II) Area based voltage control by SVC at bus 6; III) Partial DG curtailment at bus 6. If system violation cannot be eliminated after corrective actions are exhausted, load curtailment is performed as the last resort. Five first-order states of the AM control system are enumerated where only single failure at a time is considered: I) The control system and all control devices (SVC, OLTC and DG output control) are in the normal state; II) The control system fails but executive devices are up; III) OLTC fails. Other devices are up; IV) DG output control fails. Other devices are up; V) SVC fails. Other devices are up. The failure of network components (transformers, lines and generation) is ignored so as not to mask the failure of AM control system. The impact of the reliability of AM control system on the relative merits of AM and TRS are studied. A. Case results on different load levels Three typical load cases are defined, namely: the low load, the medium load and the high load. The high load case is defined as taking up system margin completely, which, in other words, means that any incremental load will cause system violations for the base case. The total load of the low load case and the medium load case are 49.62% and 84.12% of that of the high load case, respectively. All other parameters, including the network parameters, wind generation and probabilities of control system states, are the same for all three cases. The power factor of wind generation is controlled by the SVC at bus 6. It is allowed to

Fig. 3. Network diagram.

Fig. 4. DG energy curtailment for low load case.

Fig. 5. DG energy curtailment for medium load case.

Profit of wind generation under AM is calculated for each capacity and load case. Fig. 7 shows the results:

Fig. 6. DG energy curtailment for high load case.

B. The impact of AM control system reliability The impact of control system reliability on the net benefit of AM is quantified through sensitivity analysis. For simplification the sensitivity analysis considers only state I and state II. The net benefits for the three load cases with different combinations of state probability are shown in Fig. 8-10 (Note that in the legend the Prob means the probability of state II, the control system outage state. The larger Prob means the poorer reliability of the control system):

The optimal DG capacity (where DG profit reaches maximum) under AM is 10MW, 15MW and 20MW for low, medium and high load case, respectively. Active Management (AM) and Traditional Reinforcement Scheme (TRS) are compared on the basis that DG actual energy generated is the same under the different scenarios. Since no corrective action including DG curtailment is available under TRS equivalent DG capacity is determined solely by wind profile independent of the reinforcement scheme. Based on the energy delivered with optimal DG capacity under AM, the equivalent DG capacity under TRS is 9.249MW, 14.184MW and 18.421MW, corresponding to the DG capacity of 10MW, 15MW and 20MW under AM, respectively. Different TRS under different load levels are developed to accommodate the above equivalent DG capacity. For the low load case a series reactor is installed in lines 4 and 5, in order to relieve the overvoltage problem at bus 6. For medium and high load cases a series reactor is installed in lines 4 and 5.For these two load cases identical lines parallel to lines 2 and 3are also installed to solve the voltage drop problem at bus 3 and bus 4. The net benefit of AM and TRS is calculated and the results are summarized in Table I: Load Scale
TABLE I: NET BENEFIT OF AM AND TRS

Fig. 7. Profit of wind generation under AM.

Fig. 8. Net benefit of AM with different reliability profile for low load case.

Low load Medium load High load

DG Optimal Capacity (MW) 10 15 20

Net Benefit under AM () 68,537.2 103,446.0 111,918.0

Net Benefit under TRS () 103,590.9 82,063.1 132,263.3

Fig. 9. Net benefit of AM with different reliability profile for medium load case.

It is interesting to note that in this case study, TRS yields greater benefit than AM in both low load and high load case. However, for medium load case, the net benefit under AM exceeds that under TRS by about 26.1%. These results are rational since in the low load case TRS does not require the installation of new lines, thus prohibitive cost is avoided. In high load case the capital and operation cost of DG under TRS is much lower than that under AM due to reduced equivalent DG capacity under TRS which is 7% lower than DG capacity under AM.

Fig. 10. Net benefit of AM with different reliability profile for high load case.

With reference to Fig. 8-10 it is consistent with qualitative analysis that the net benefit decreases or even becomes

negative as the probability of control system failure increases provided the installed DG capacity remains unchanged. In the medium and high load cases, the net benefit is more sensitive to the reliability of control system than that in the low load case. For a given DG capacity under AM, the threshold of probability where net benefit under AM is exactly the same as that under TRS can be found by reducing the reliability level in small steps from the 100% reliable point. In this way, the threshold probability of state II is found to be 0.948 for DG capacity of 15 MW in medium load case. V. CONCLUSION This paper has investigated the impact of control system reliability on the benefits of active network management (AM) compared to traditional network reinforcement schemes. The viability of AM over network reinforcement has been evaluated considering a crucial factor, the reliability of AM control system. The case study confirms the general assumption that the viability of AM improves with improvement in control system reliability for all three load levels. However, the load level and AM control system reliability should be jointly taken into account for a comprehensive analysis on the overall cost benefit of AM compared to TRS. Load level has a complex effect on AM viability and is system specific. In this case study however it was found that extremely low or high load levels, even though with reliable AM system, discourage the application of AM compared to the medium load level with all other conditions remaining the same.

[11] R. A. F. Currie and G. W. Ault, Fundamental research challenges for active management of distribution networks with high levels of renewable generation, presented at the 39th International Universities Power Engineering Conference, 2004.

VIII. BIOGRAPHIES
Kang Ma (IEEE Student Member) was born in Beijing, China. He graduated from the Tsinghua University with a Bachelor of Engineering Degree in 2007. From January 2008 onwards he has been pursuing a PhD degree at the University of Manchester in the Electric Energy and Power Systems Group. His research topic is Risk Assessment of Power Systems under Corrective Control Paradigm.

Joseph Mutale (IEEE Senior Member) is a senior lecturer at the University of Manchester in the Electric Energy and Power Systems Group. His research interests include the area of pricing of transmission and distribution networks and integration of distributed energy resources into power systems.

VI. REFERENCES
http://www.sedg.ac.uk/Publication/Active%20Management%20of%20D istribtion%20Networks.pdf , accessed Nov 2008 [2] A. Shafiu and T. Bopp, Active management and protection of distribution networks with distributed generation, presented at Power Engineering Society General Meeting, 2004. [3] H. B. Puttee and P. R. MacGregor, Distributed generation: Semantic hype or the dawn of a new era?, Power and Energy Magazine, IEEE 1(1): 22-29, 2003. [4] J. Mutale, Benefits of Active Management of Distribution Networks with Distributed Generation, presented at Power Systems Conference and Exposition, 2006. [5] G.A.R. Rodriguez and E. O'Neill-Carrillo (2005), Economic assessment of distributed generation using life cycle costs and environmental externalities in Proceedings of the 37th Annual North American Power Symposium, 2005. [6] J. Mutale and G. Strbac, Methodology for Cost Reflective Pricing of Distribution Networks with Distributed Generation presented at Power Engineering Society General Meeting, 2007. [7] J. Gordijn and H. Akkermans, "Business models for distributed generation in a liberalized market environment," Electric Power Systems Research 77(9): 1178-1188, 2007. [8] R. Billinton, and R. N. Allan, Reliability Evaluation of Power Systems, Plenum Publishing Corporation, 1994. [9] http://www.sedg.ac.uk, accessed Nov 2008 [10] J. Zhang and H. Cheng, Technical and economic impacts of active management on distribution network, International Journal of Electrical Power & Energy Systems 31(2-3): 130-138, 2009. [1]

Potrebbero piacerti anche