ERDF EXPERIENCE IN REDUCING NETWORK LOSSES

ERDF EXPERIENCE IN REDUCING NETWORK LOSSES Michel ODDI Frédéric GORGETTE Guillaume ROUPIOZ EDF R&D France ERDF France EDF R&D - France michel.oddi@edf.fr frederic.georgette@erdfdistribution.fr guillaume.roupioz@edf.fr ABSTRACT Optimising the normal opened points of its MV feeders allows ERDF, the main French distribution network operator, to reduce its losses to approximately 2%, while adopting a summer/winter scheme for its (V) substations should allow another rapid reduction of 2% but at the cost of a degradation in supply quality whose acceptability remains to be discussed. The adoption of new MV/LV transformers will also allow new and significant savings whose effects will appear progressively. The experiments and complimentary studies to come will allow ERDF to judge if it is opportune to modify the voltage system to operate the MV network at the lowest possible voltage level with the prospect of gains in the order of 1%. The connection of decentralised generation is shown to be unfavourable overall in terms of losses, except in the case of a connection very close to a (V) substation. INTRODUCTION Faced with a new situation of rare energy, the necessity of reducing greenhouse gasses and the massive introduction of decentralised generation, ERDF, a subsidiary of the EDF group and the main French distribution network operator, launched, in 2008, a vast study to identify possible paths to improve the losses in its network and evaluate their potential. It is evident that there are only two types of measure possible to reduce them, either by modifying the network operation or utilising better energy efficiency network components. The first one will necessarily have an impact on the quality of supply that we can legitimately assume will be negative but has the advantage of giving immediate results, whereas the second one can only have long term effects. The anticipated equipment replacement cost is such that it can only be done progressively. LOSSES IN THE ERDF NETWORK Like the great majority of network operators, ERDF distinguishes between technical and non-technical losses. The technical losses are estimated, from manufacturing data, iron losses in the transformers and conductor impedances, and from statistical calculations in order to take load curves into account. Thus, in 2006, the technical losses in the ERDF network represented 12 TWh for an injection of 354 TWh, i.e. a loss of 3.43%. They are distributed as follows: (V) 17% 50% zero-load and 50% under load substations MV feeders 28% MV/LV substations 36% 69% zero-load and 31% under load LV network 12% Meters 2% Table 1: losses in the ERDF network An examination of this table allows us to prioritise the stakes, the MV/LV transformers, particularly regarding their iron losses, the MV feeders and the (V) transformers. LOW LOSSES MV/LV TRANSFORMERS The reference standard for these equipments is the European standard EN 50464-1 which classifies the losses in levels: five iron loss levels classified from A 0 to E 0 and four load loss levels classified from A k to D k. As an example, for a 400kVA transformer, which is the most common model in Europe, the levels are as follows: Iron losses Load losses Level Gains over E 0 Level Gains over D k A 0: : 430W 54% A k 3250W 46% B 0 : 520W 44% B k : 3850W 36% C 0 : 610W 34% C k : 4600W 23% D 0 : 750W 19% D k : 6000W - E 0 : 930W - - - Table 2: loss levels in MV/LV transformers The yield of a transformer, which is the ratio between the power it absorbs and the one it puts out, naturally depends on the load. It is very good, typically between 0.975 and 0.99 for classic MV/LV transformers. However, there are so many that the losses in all of this equipment are not negligible as the figures given above show. The situation at ERDF For ease of comprehension, the performance of the equipment is reduced to the levels of the standard in force. The energy performance required by ERDF specifications has, up to now, been located in the lower part of the ranges C o D k, D o D k, E o C k, and E o B k from low power (50kVA) up to the highest (1000kVA). Aware of the weakness of the

energy efficiency of its equipment, ERDF has modified its specifications with a view to only purchasing low losses equipment from now on. The new specification, applicable to cabin transformers from 160 to 1000 kva purchased from 2009, imposes the following characteristics: Iron losses: minimum level C o; Load losses: level C k. The equipment offered by the suppliers will be accepted or not depending on the technical-economic criterion of their complete cost, which is equal to: C complete = P purchase + P o + P k where P purchase is the purchase price of the equipment, P o the updated and accumulated purchase cost of the iron losses and P k the same cost relative to load losses. Order of magnitude of expected gains The most common cabin transformers at ERDF are of 400kVA. The old ones were of type E o C k, the lowest performance new ones will be of type C o C k, and the highest performance new ones will be of type A o C k. As the levels of load losses are identical for the three transformer types, changing models has no impact on them. The C o C k equipment allows a saving of 34% of the iron losses and the A o C k, 55%. Over time, the gain should be from 8 to 12% of total losses, but only at a slow rate. If we consider that ERDF has 700,000 transformers and that only 13,000 transformers are purchased each year, we can quickly estimate the potential improvement in iron losses on the MV/LV transformers assuming identical reductions for all equipment. It is thus 0.6% to 1% per year. We can therefore hope to reduce total network losses by 0.15 to 0.25% each year, at least at the beginning. These numbers tend to increase as more equipment is replaced. It should be noted that the figures could be approximately doubled for the next four years as it is planned to replace 13,000 units per year early as they are polluted with PCBs. OPTIMISATION OF NORMAL OPEN POINTS The ERDF MV network has a tree structure, which can be looped in degraded situations. Normal Open Point Figure 1: tree networks and normal open points The position of the normal open points has long been established to offer the best possible power quality. However, a rapid calculation on a linear segment shows that the two criteria, optimisation of power quality and limitation of feeder losses, both lead to placing the normal open point at the middle of the segment. Figure 2: normal open point of a linear feeder Despite this positive assessment, ERDF has modified its criteria in 2007 to give a little more weight to losses with respect to power quality. ERDF now applies the method of products P*Leq. This new concept consists of balancing the feeders P*Leq products where P is the average power consumed from the feeder and Leq is the equivalent length 1. As the feeder load is only taken into account as an average, you can reasonably ask what influence the real load curve has. To answer this, we have dealt as an example with a French administrative department, judged to be representative, with 72 MV feeders of all types, urban, semi-urban and rural. Optimisation of losses Using a heuristic branch combinatorial method that we had to develop, we determined the position of the normal open points that lead to minimum losses without considering other criteria. In parallel, we evaluated the impact of operating schemes corresponding to the power quality characterised by the SAIDI. After having modelled the MV feeders, we entered the real load curves and studied several configuration cases, the old scheme favouring power quality, the schemes developed to balance the P*Leq products and the schemes resulting in minimum losses. For the last case, we assumed the ability to move the normal open points without constraint and envisaged several reconfigurations: fixed (no reconfiguration), seasonal (summer/winter/half season), monthly, weekly, semiweekly (week/weekend), daily and hourly. The results obtained on the studied examples are summarised in the following table. Operation scheme Losses saving in % Operation number per year SAIDI in mn Old policy Reference 0 62,3 P*Leq products 2,07 0 62,2 Losses optimisation -Fixed configuration -Hourly configuration 2,48 2,50 0 35 594 62,6 62,8 Table 3: Optimisation of MV feeder losses 2 N Lsec( n)* τ ( n) n= 1 Leq = 1 τ aerial where L sec is the length of each section, τ its probability of failure and τ aerial the reference value for overhead network. 2 We have not shown the results for other reconfigurations to reduce the table size. In fact, they are very close to those for hourly reconfiguration.

In view of the results, we see that the balancing of the P*Leq products is well founded and allows a saving of 2% of the MV feeders losses which results in a very good compromise and has the advantage of being easy to apply. We also notice, that reconfiguring the network, even seasonally, does not provide any significant benefit. Impact of distributed generation The preceding study did not take into account the possible presence of distributed generation, but a simplified reasoning shows that it is truly difficult to establish general rules. Suppose that two feeders are operated looped in the presence of a distributed generation facility: there is a naturally supplied zone by this generation bordered by the two points where the current intensity cancels out. Figure 3: naturally supplied zone by a distributed generation facility If we place the normal open point near the edge of the natural supply zone, the current circulation is hardly modified and the losses are optimised. On the other hand, if the normal open point is distant from the edge of the natural supply zone, power flow is modified and the distributed generation facility will hold back and the losses will increase. Figure 4: normal open point increasing losses The natural supply zone certainly depends on all the time the instantaneous consumption and the power of the distributed generation facility at the same instant. In addition, choosing one of its terminals as the normal open point has little chance of optimising power quality. Calculations have been carried out assuming three production types, hydraulic, wind and cogeneration, which have very different profiles. We assumed a flat profile for cogeneration and used recorded logs for the wind and hydraulic profiles. Three connection points were selected, at the head of the output, the middle and the end. These configurations are not entirely realistic (for example, wind generators are mostly connected to dedicated feeders) but they do cover the conceivable range. Connecting a distributed generation facility to the head of the feeder is the optimal solution as it reduces the (V) transformer losses without affecting the feeder losses or impacting the normal open points and thus the power quality. In the other cases, connecting to the middle or end of the feeder, and in general, the losses increase The precision of 10-3 is obviously illusory but allows showing the impact of the envisaged reconfigurations. significantly if you do not change the normal open points: in our examples, the losses may be doubled in some cases. If we modify the normal open points to optimise the losses, we notice a strong degradation of power quality: the substation SAIDI can increase by 60mn in the worst cases. The use of automatic reclosers upstream of the zones fed by the distributed generation facility may allow the optimisation of both losses and power quality; this remains to be studied. In the case of dedicated feeders to connect the distributed generation, the global level of losses obviously depends on the length of the outgoing feeders: extra load losses in these feeders and gains in the (V) transformer. OPERATION OF (V) SUBSTATIONS ERDF operates approximately 2250 (V) substations of which approximately 90% are of the twotransformer type. These substations are permanently operated with the two transformers in service according to an operating scheme known as open busbar operation. Figure 5: open busbar scheme This scheme was chosen as it allows good power quality. If a MV feeder circuit breaker fails and does not open, backup is provided by the corresponding MV input circuitbreaker and only half the substation users are deprived of electricity. However, it does not allow optimisation of transformer losses. Closed busbar operation The normal ERDF network planning rules allow that normally a single transformer can feed the entire substation as shown in the following diagram, known as closed busbar operation. Figure 6: closed busbar scheme This scheme, which is currently unusual, could be permanently adopted. In relation to the open busbar scheme, the iron losses are halved but the load losses increase. We can easily show that, if the two transformers are balanced, the overall balance of losses with closed busbars is positive compared to open busbars as the load losses in the transformers are lower than the iron losses. This scheme has the advantage of being able to be implemented without prior adoption of equipment but has

the disadvantage of degrading power quality. If HV network failures have no impact, those of the MV network certainly will. If a MV feeder circuit breaker fails and does not open on a fault, the backup by the input circuit breaker will deprive all the station customers of electrical supply. In addition, polyphase faults occurring on the MV network will cause voltage sags for all the station customers whereas only half of them will be affected under opened busbar operation. Parallel operation This operating mode, illustrated by the figure below, is often employed as it offers good power quality. In the case of the loss of a transformer, no customer is deprived of electrical supply. Figure 7: parallel scheme This operating scheme also limits load losses by perfectly balancing the load on the two transformers. However, the gains in term of losses are limited: if, with open busbars, one transformer delivers three times more power than the other one, the load losses of the station are only reduced by 20% for parallel operation and if it delivers one and a half times the power of the other one, they are only reduced by 4%. On the other hand, parallel operation requires, in particular, limiting the short-circuit power of the MV network (by rapid opening of the parallel or by short-circuit reactances) and the design of new voltage regulators and MV neutral grounding which are difficult to implement and expensive. The parallel scheme has no interest in our case. Summer/Winter scheme We saw above that the closed busbar scheme was interesting in terms of losses when load losses were low. The optimisation solution therefore consists of changing the operating scheme depending on transformer loads. In current conditions, adopting a real-time scheme is not envisageable because, on the one hand, the operation staff will be overloaded and, on the other hand, there are no simple criteria for manual implementation. The solution that currently seems to be the most suitable is a Summer/Winter scheme. Simulations carried out on the real ERDF network lead to a reduction, thanks to a Summer/Winter scheme (summer period from 15 April to 15 October, in (V) substation losses of 13% and 2.2% over the entire ERDF network. On the other hand, the supply quality should be degraded, each customer experiencing an extra short outage (less than 3mn) every 8 years and, above all, 25 extra voltage sags each year. VOLTAGE AND POWER FACTOR The impact of power factor on feeder losses has long been known, which is why network operators try to keep the phase angle between voltage and current as near to zero as possible. In this area, ERDF policy is the automatic putting into service of MV capacitors in the (V) substations. It should be noted that they also contribute to optimise the transmission network. This apparently simple process is in fact extremely complex as many parameters enter into the equation. The examples below demonstrate this. Impact of burying the network Underground network are being developed more and more both for protection from climatic events and aesthetic reasons. One of the effects is the distributed generation of reactive power all along the feeder and the resulting increase in voltage. We can estimate that the complete burying of an ERDF type entirely overhead feeder, all other factors being equal, results in a reduction of load losses for the feeder of between 10 and 15% thanks to the improvement in power factor and an increase in MV/LV transformer iron losses of between 1.5 and 2% because of the increase in voltage. The overall total will however be positive with a reduction of total losses for the feeder of between 2 and 3%. Distributed generation and in-network capacitors In one of the preceding paragraphs, we analysed the effect of distributed generation on losses uniquely in terms of the position of the normal open points. Another approach consists, for a given MV feeder, of optimising the connection position of the generation facility and its generation profile. Under these conditions, we see that, if the consumption is evenly distributed along feeder, the optimum in terms of losses is reached when the producer is connected at 2/3 of the feeder and it produces 2/3 of the apparent power that it consumes. The load losses in the feeder can then be reduced by 90%. Obviously, this proposition is not realistic, on the one hand because the production level is either unpredictable (wind for example) or subject to other requirements (cogeneration for example) or, on the other hand, because the reactive power is not always adjustable. In addition, even when it is, it is most often used to adjust the voltage along the feeder to the detriment of the power factor. The installation of capacitors in the network is just a particular case of distributed generation whose active power production is null. Their optimal position is certainly 2/3 along the outgoing feeder and they must generate 2/3 of the reactive power consumed in real time by the feeder. Theoretically, this could reduce the under-load losses in the outgoing feeder by approximately 15%.

Voltage, consumption and losses A commonly held belief is that increasing the voltage reduces losses as the current is reduced. In practice, other factors come into play, such as an increase in transformer iron losses and in customer consumption. Several measuring campaigns carried out in North America lead us to believe that consumption tends to decrease globally with the reduction in voltage. We could expect a gain of approximately 1% of MV & LV network losses by operating the network at the lowest possible voltage compatible with the regulation if the behaviour is the same as in North America. We notice that the presence of network capacitors allows smoothing of the feeder voltage profile, which is a favourable factor. Which is why ERDF plans to carry out experiments in 2009 to evaluate, on the one hand, the consumption profile for typical customers and, on the other hand, the possible change in losses depending on the voltage. to 12% but at a low annual rate of 0.15 to 0.25% at the beginning even though the anticipated replacement of PCB contaminated equipment allows doubling of these figures for the first four years. The integration of distributed generation of significant capacity does not have a favourable impact on distribution network losses except in the particular case of connection very close to a (V)HT/MT station. In all other cases, the production profile type and the connection position will lead to very disparate results but they will always be more or less unfavourable for the losses/supply quality compromise. (V)HT network voltage plan Regulations fix the (V)HV voltage range. Given that the MV voltage is independent of the transmission network voltage because of the load tap changer using, it is interesting to operate the transmission network at its maximum voltage level possible to reduce its load losses. Under these conditions, what happens to distribution network losses or, more precisely, those in the (V) transformers? With a ratio of 1.4 between the resistances of the primary and secondary windings, as we can find on ERDF transformers, an increase of 5% in the HV voltage leads to a 5% increase in the total losses of this equipment. CONCLUSION The implementation of certain operating policies can allow simple and very fast optimisation of the losses in distribution networks. In the case of ERDF, the optimising of the normal opened points on its MV feeders can reduce them by approximately 2% without affecting power quality and the adoption of a Summer/Winter scheme for the (V) substations can allow another 2% savings but at the price of a degradation in power quality whose acceptability remains to be discussed. Operating the network at the lowest possible voltage could give ERDF another reduction in losses on the order of 1% without affecting supply quality. Complementary experiments and studies are necessary, on the one hand to confirm the correlation between lowering the supply voltage and lowering of consumption and, on the other hand, to revisit the plan compromise between voltage system and power factor. It is in view of these results that we will be able to determine the feasibility and interest of this type of solution. The replacement of the MV/LV transformers with better energy efficiency equipment will allow a substantial gain over the ERDF network losses, ultimately on the order of 8