cired distributed automatic voltage control (davc)

CIRED

20th International Conference on Electricity Distribution

Prague, 8-11 June 2009 Paper 0976

DISTRIBUTED AUTOMATIC VOLTAGE CONTROL (DAVC) Murray THOMSON, Ian RICHARDSON, John P. BARTON CREST, Loughborough University – UK [email protected]

ABSTRACT The use of reactive power control as a means of controlling voltages in low-voltage (230 V / 400 V) distribution networks is considered. The proposal is to use spare capacity in power electronic interfaces such as inverters in domestic photovoltaic (PV) and micro combined heat and power (CHP) units to generate or consume reactive power according to locally measured voltage. This Distributed Automatic Voltage Control (DAVC) is proposed as a means of allowing greater penetration of low-carbon technologies into existing distribution systems. A detailed model of a real distribution system in the UK has been used to determine that DAVC can be effective in controlling voltages despite the low X-upon-R ratio typical of low-voltage distribution networks. A slight increase in network losses is predicted but this could be justified by the allowance of a greater penetration of PV and a reduced need to curtail its generation due to network voltage rise. A significant reduction in source power factor is also noted.

INTRODUCTION The traditional approach to ensuring that customer voltages in low-voltage (LV) distribution networks are kept sufficiently close to the nominal (230 V / 400 V) is largely passive: in the UK, voltages are actively controlled by onload tap changers at primary substations (eg. 33 kV to 11 kV) but thereafter there is rarely any further active voltage control. In most cases, the final distribution transformers (eg 11 kV to 400 V) have only off-load tap adjustment and the maintenance of customer voltages relies mainly on the network designer taking care with cable lengths and cross-sectional areas. In the interests of minimising cable costs and deferring network reinforcement, it is natural that the allowable range of voltages is often fully utilised. In other countries, voltage regulators (autotransformers with on-load tap changing) are more widely used within primary networks (> 1 kV) but final LV networks are again largely passive. Although it is perfectly possible to equip distribution transformers with on-load tap changers and/or to deploy voltage regulators in LV networks, capital and maintenance costs remain a concern and thus the largely passive traditional approach prevails. With this starting point, it is now anticipated that LV networks may soon be required to accommodate significant penetrations of low-carbon generation technologies,

CIRED2009 Session 4

Paper No 0976

David G. INFIELD University of Strathclyde – UK [email protected]

including photovoltaics (PV) and micro combined heat and power (micro-CHP). These micro generators will naturally alter the flow of active power through the networks and tend to cause voltages rise. The risk of voltages exceeding statutory limits has been identified as the main constraint to the allowable penetration of micro generation in existing LV networks [1, 2]. This paper considers the possibility of using reactive power control as a means to improve voltage profiles in LV networks. The basic concept is widely used in the context of high-voltage networks, and relies on the fact that the voltage drop (or rise) in a section of cable (or overhead line) is roughly proportional to PR + QX, where P and Q are the active and reactive powers flowing through the line, and R and X are the resistance and reactance of the line itself [3]. In high-voltage networks, above 10 kV for example, X >> R and thus network voltages are largely determined by the QX component, and therefore by reactive power flows. In LV networks (230 V / 400 V) however, the resistance is much more significant and so the PR component usually dominates, and thus reactive power control is much less effective as a means of voltage control. Moreover, the relatively high resistances found in LV networks mean that the real energy losses associated with reactive power flows are significant: if reactive power flows are to be increased in the interests of voltage control, then it is important also to consider these increased losses. These factors, combined with the practical challenges of deploying and maintaining switched capacitors or similar in LV networks explain why reactive power control is not normally considered as a means of voltage control in LV networks. However, if low-carbon technologies are to be widely deployed, and they are likely to have powerelectronic inverter interfaces with spare capacity, then the cost-benefits are somewhat altered. The proposal considered in this paper is that the powerelectronic inverter interfaces, associated with photovoltaic and micro-CHP units, be configured to generate or consume reactive power according to the network voltage at their point of connection: when an inverter detects a low voltage at its terminals it generates reactive power, and when high it consumes. The individual rating of these inverters is only in the order of 1 kVA to 2 kVA, and the available reactive power capacity will be significantly reduced whenever being used for the originally intended purpose of supplying active power to the grid. Nonetheless, the aggregated reactive power contribution from large numbers of such

CIRED



inverters could have significant effect on network voltages and thus form a mechanism for Distributed Automatic Voltage Control (DAVC).

NETWORK AND DEMAND MODELLING In order to investigate and quantify the potential effectiveness of DAVC, a detailed model of a real distribution network in the UK was used. This test network includes one complete 11 kV feeder and all associated LV networks serving some 1262 properties in central Leicester, UK. Details of the power flow and demand modelling were previously presented [4]. In summary, the model is written in Matlab and performs a minute-by-minute time-step simulation of the network power flow, using demand profiles such as that shown in the top graph of Figure 1. In order to reflect the diversity of demand found in real networks, the modelled demand profiles for the 1262 properties are all different. Simulations are run for one winter day and one summer day, and the aggregated demand profiles are shown in black in Figure 2.

MICRO-GENERATION MODELLING In order to see a significant voltage rise caused by micro generation, and later to see the effectiveness of DAVC, a very high penetration of micro generation was added to the model: • All of the 1262 properties were equipped with a microCHP unit rated at 3 kWth and 1 kWe as detailed previously [7]. These generators are each connected via an inverter rated at 1 kVA. • 629 (almost half) of the properties were equipped with a 2160 Wpeak PV array, as detailed previously [8]. Each array has a 1.8 kVA inverter. Example generation profiles for an individual property are shown in Figure 1, while the aggregated profiles for all 1262 properties are shown in Figure 2.

6

Summer day

2000

Power (kW)

Winter day 5

Demand (kW)

Demand (26877 kWh) CHP + PV (16043 kWh) CHP only

2500

4

1500

1000 3

500

2

0

1

0

2

4

6

8

1

10 12 14 Time (h)

16

18

20

22

24

22

24

Demand (43135 kWh) CHP + PV (29618 kWh) CHP only

2500 0

2000

1 0 0

2

4

6

8

10

12 14 Time (h)

16

18

20

22

24

Figure 1 – Profiles at an individual example property

An unbalanced power flow (load flow) calculation is used to determine minute-by-minute voltages throughout the network. This data is then used to provide ten-minute averages for consideration under European Standard EN 50160 [5]. The distributions of the ten-minute average voltages delivered to all 1262 properties are shown in black in Figure 3. Observe that in this base-case condition there is already a considerable spread of voltages: EN 50160 requires that voltages are normally within ± 10% of the nominal 230 V, which equates to a range of 207 V to 253 V. In the UK, the Electricity Safety, Quality and Continuity Regulations (ESQCR) [6] set a range of 216 V to 253 V.

CIRED2009 Session 4

Paper No 0976

Power (kW)

PV (kW) CHP (kW)

0

1500

1000

Winter day

500

0 0

2

4

6

8

10 12 14 Time (h)

16

18

20

Figure 2 – Aggregated profiles for all 1262 properties

As expected, this very high penetration of micro generation causes significant voltage rises, as shown in red in Figure 3. In both summer and winter cases, the predicted voltages are significantly in excess of the 253 V upper limit set out in both EN 50160 and the ESQCR, and would not be acceptable in practice. The remainder of this paper illustrates the use of DAVC to restore voltages to an acceptable level.

CIRED



error, the voltage error being the difference between that calculated by the power flow and a 230 V reference.

DAVC MODELLING The system being modelled includes some 1891 inverters (1262 + 629) and the DAVC concept requires that the reactive power generated or consumed by each inverter be controlled according to its terminal voltage. In other words, there are 1891 closed-loop control systems operating in parallel and interconnected by the impedances of the distribution network. In order to model this, the DAVC modelling was integrated with the power flow (load flow) within the time-step simulation framework in Matlab.

RESULTS The main results of this study are indicated by the green curves in Figure 3, which indicate that DAVC could be very effective in controlling LV network voltages. In both summer and winter cases, the predicted voltages are brought back into the acceptable range (below 253 V). It is important to emphasise that this is being achieved (the red line is being moved to the green) purely by means of reactive power control. The active power flows throughout the network are affected only marginally and, most importantly, the very large penetration of micro generation is being accommodated without need of any active power constraint.

The profile of maximum available reactive power (generation or consumption) for each inverter was precalculated, taking account of their individual active power profiles. Inverter losses associated with reactive power generation or consumption were not modelled.

DAVC CONTROL ALGORITHM

The main limitation on the effectiveness of DAVC is the available inverter capacity. In the model, this shows up particularly at midday in the winter case when much of the capacity is being used for the primary function of delivering active power. The onset of this limitation depends greatly on the assumed penetrations of PV and CHP and the timing of CHP operation.

Various algorithms for the control of the reactive power output of the individual inverters were implemented and tested within the model. Some were ineffective and others unstable, but good results were achieved with a simple integral controller that increases the reactive power consumed at each property by 50 var per minute per volt

Base case Mean: 246.2 Micro Gen. Mean: 249.2 DAVC Mean: 244.0

Probability

0.2

Summer day

0.15

0.1

0.05

0 210

220

225

230 235 Voltage (V)

Base case Mean: 239.7 Micro Gen. Mean: 245.8 DAVC Mean: 242.7

0.2

Probability

215

240

245

250

255

240

245

250

255

Winter day

0.15

0.1

0.05

0 210

215

220

225

230 235 Voltage (V)

Figure 3 – Distributions of ten-minute average voltages delivered to all 1262 properties throughout the network

CIRED2009 Session 4

Paper No 0976

CIRED



Network losses A common concern with the concept of using reactive power control as a means of voltage control in LV networks is that network losses (of real energy in the form of active power) will be increased. The model output, as summarised in Table 1, shows that the increase is actually quite small, compared to the low-carbon generation that it is facilitating. Table 1 – Losses

Demand (kWh)

Generation Network losses (kWh) (kWh) Base case 0 813 Summer 26877 Micro gen. 532 16043 DAVC 1088 Base case 0 1604 Winter 43135 Micro gen. 647 29618 DAVC 987

Source power factor A more significant concern may be the reduction of power factor as seen at the source of the 11 kV feeder at the primary substation. Table 2 summarises the results from the model. Notice first that, in the two base cases the power factors are typical of those found in practice today – in the absence of any significant penetration of micro generation. Next, the two cases with micro generation show a substantial reduction in energy demand but almost no change in reactive power, and thus a reduction in the source power factor. The significant results however are in DAVC cases where, a very large increase in reactive power demand is indicated, bringing a very significant reduction in power factor. Table 2 – Feeder source data

Energy Reactive Average (kWh) (kvarh) power factor Base case 27690 6959 0.97 Summer Micro gen. 11366 6578 0.87 DAVC 11923 29874 0.37 Base case 44738 12735 0.96 Winter Micro gen. 14164 11470 0.78 DAVC 14504 25892 0.49

CONCLUSIONS Voltages in existing LV networks already exploit the full allowable voltage range (black lines in Figure 3) and the addition of high penetrations of micro generation could raise voltages beyond the upper limit (red lines in Figure 3). Implementation of Distributed Automatic Voltage Control (DAVC), which uses the reactive power capabilities of the inverters associated with the micro generators, could be an effective means of bringing the voltage back down to acceptable levels (green lines in Figure 3). Thus, the DAVC facilitates a very high penetration of micro generation and does so without need of any active power constraint.

CIRED2009 Session 4

Paper No 0976

Implementation of DAVC within the software of the inverters would have low cost and the increased network losses do not appear to be an obstacle. The reduction of source power factor requires further consideration. This study has been concerned with micro-generation. However, it is widely anticipated that demand side management, particularly demand response, will become increasingly common as a means of easing the integration of non-schedulable generation, not only of the sort considered in this paper, but also large scale renewable energy generation. In this context, it is reasonable to consider the implementation of DAVC in loads with suitable power electronic converter interfaces.

Acknowledgments This work was supported by the Engineering and Physical Sciences Research Council, UK (Project GR/T28836/01). The authors would like to thank Melody Stokes and Mark Rylatt at IESD, De Montfort University, for their work on the original demand and network modelling. We are grateful to East Midlands Electricity (now E.ON Central Networks) for supplying the network data.

REFERENCES [1] S. Ingram, S. Probert and K. Jackson, 2003, "The Impact of Small Scale Embedded Generation on the Operating Parameters of Distribution Networks", P B Power, UK DTI Report: K/EL/00303/04101. [2] Mott MacDonald, 2004, "System Integration of Additional Micro-Generation (SIAM)", UK DTI Report: 04/1664. [3] L. L. Freris and D. G. Infield, 2008, Renewable Energy in Power Systems, John Wiley & Sons. [4] M. Thomson and D. G. Infield, 2007, "Network PowerFlow Analysis for a High Penetration of Distributed Generation", IEEE Transactions on Power Systems, vol. 22, pp. 1157-1162. [5] European Standard EN 50160, 1999, "Voltage characteristics of electricity supplied by public distribution systems", CENELEC. [6] UK Government, 2002, "Statutory Instrument 2002 No. 2665: The Electricity Safety, Quality and Continuity Regulations 2002". [7] M. Thomson and D. G. Infield, 2008, "Modelling the impact of micro-combined heat and power generators on electricity distribution networks", Proc. IMechE, Part A, Journal of Power and Energy, vol. 222, pp. 697-706. [8] M. Thomson and D. G. Infield, 2007, "Impact of widespread photovoltaics generation on distribution systems", IET Renewable Power Generation, vol. 1, pp. 33-40.