Dynamic Grid Edge Control

Transcription

1 Dynamic Grid Edge Control Visibility, Action & Analytics at the Grid Edge to Maximize Grid Modernization Benefits The existence of greater volatility at the grid edge creates a set of problems that require the use of decentralized, dynamic controllers that monitor and act at the point of the problem. This white paper introduces the concept of dynamic grid edge control and the associated grid edge devices that can be used as shock absorbers to manage voltage volatility and to improve precision of voltage control along a distribution feeder. Improving control of the grid edge allows utilities to fully realize the benefits of many grid modernization programs, including advanced VVO/CVR schemes, peak demand management, line-loss reduction, integration of renewables/pv solar, as well as reliability and power quality enhancement. Varentec, Inc

2 Dynamic Grid Edge Control A Critical Gap in Realizing Smart Grid Benefits Executive Summary The existing centralized command and control architecture currently used by utilities presents challenges in delivering many of the possible benefits from grid modernization initiatives. Greater volatility at the grid edge creates a set of problems that require the use of decentralized, dynamic controllers that monitor and act at the point of the problem. This white paper introduces the concept of dynamic grid edge control and the associated grid edge devices that can be used as shock absorbers to manage voltage volatility and to improve precision of voltage control along a distribution feeder. Improving control of the grid edge allows utilities to fully realize the benefits of many grid modernization programs, including advanced VVO/CVR schemes, peak demand management, line-loss reduction, integration of renewables/pv solar, as well as reliability and power quality enhancement. I. INTRODUCTION The electric grid has offered highly reliable, low-cost energy to power the industrial and consumer customer base for many decades. This performance is remarkable, relying on centralized large generators with a centralized command and control approach to grid operations. Recent Smart Grid initiatives have targeted further improvements in reliability, cost, asset utilization, and system performance through the addition of smart meters, sensors, software, and analytics to traditional grid assets. The intent of Smart Grid investment was that existing controllers such as capacitor banks, LTCs, and Line Voltage Regulators would be able to deliver the improved grid control needed to realize the targeted benefits. New trends such as distributed generation, energy conservation, electric vehicles, and unprecedented voltage visibility from AMI are driving changes that primary side control equipment was not designed to handle. Utility operational managers are stating that they do not currently have the tools in their toolkit to effectively manage and optimize the grid in the changing environment, and that a new layer of decentralized, distributed control capability is needed at the edge of the grid. It is important to first understand why, after decades of success, traditional centralized command and control strategies are now experiencing significant limitations. Distribution grid operation has typically been based on radial networks fed from substations, with operating margins calculated based on averaged feeder loading models that could be predicted with a high level of assurance. The primary operations objective has been to maintain the feeder voltage within a specified ANSI band of +/-5% of nominal value. Although the voltage compliance requirement is at the point of common coupling (PCC) at the customer meter, most utilities have used the primary side voltage estimate as a proxy for the customer voltage. This white paper will shed new light on secondary side grid edge control. 2

3 II. PRIMARY SIDE VOLTAGE CONTROL The primary side voltage typically decreases along the length of the feeder, driven by line impedance and line currents which are a summation of individual customer load currents. It can be seen that longer feeders with heavier loading will experience greater levels of voltage drop. As the overall feeder loading varies (based on season, day and time), the voltage at points along the feeder also varies. An additional factor is the voltage at the substation end, which can also fluctuate based on the transmission side voltage level. Figure 1 shows a typical primary side voltage profile along a feeder at different times of day. Figure 1: Conventional model of VVO implementation with expected primary side voltage profile along feeder length, including variations at different times. Utilities have traditionally ensured voltage compliance by estimating the level of voltage variations using model-based analysis, and then using Load Tap Changers (LTC), Line Voltage Regulators (LVR) and fixed and switched capacitor banks to try to maintain voltages within the desired band. Control techniques include operating the assets based on line current level (or using time of day as a proxy for line current level), or using an end-of-line sensor to measure voltage at the end of the feeder. These simple approaches have worked well in the past, but are failing to provide the enhanced level of control that the utilities now need to achieve the new objectives that are being imposed on them. There are two key factors that utilities have not fully accounted for. The first is the level of variability that is seen at an individual customer load, which is typically not reflected in the averaged models assumed in feeder level voltage analysis. The second is the impact that the voltage drop across the distribution transformer has on the secondary side voltage. Utilities use the primary side voltage estimate as equivalent to the secondary side voltage. This would be reasonable if every individual load is assumed to have the averaged characteristics at the substation. However, what is seen at individual loads are currents that fluctuate by an order of magnitude over the course of a day, with power factors that vary from unity to as low as This can result in voltage drops across the distribution transformer of 2 to 10 volts, about 1 5% of the line voltage or almost half of the allowed ANSI band. This causes the minimum voltage point along a feeder to not be at the end of the feeder, but at arbitrary locations along the feeder, depending on the transformer impedance and loading level at a particular time. This does not seem to have been anticipated, but has been validated by many utilities that have been examining their AMI data. The net result is that primary side voltage, assumed to be a proxy for the secondary side voltage, does not always represent the secondary side voltage the metric by which voltage compliance and power quality are measured. Further, as many grid optimization programs, such as CVR or peak demand reduction, have been predicated on the ability to achieve tight control of feeder voltage profiles, it is not surprising that they have not been able to deliver the projected performance or return on investment. 3

4 Figure 2a shows real measured secondary side voltage data for a typical feeder at 10 pm, showing that the secondary side voltage does not drop at the end of the feeder, but at Node 13 in the middle of the feeder. Figure 2b shows the same feeder at 11 pm, when the minimum occurs at Node 29 again, not at the end of the feeder. Further, neither voltage profile exhibits the smooth characteristics predicted by averaged models, and has low-voltage points that shift from node to node at different times of the day. The red dot shows the minimum voltage across the feeder at any given time. It is seen that the secondary voltage for a normal average feeder varies by as much as 10 volts on a 240 volt base, or by 4.2%, almost half the allowed ANSI band. The dynamics and variability makes it impossible to predict or model the low voltage points that would, for example, determine the CVR gains that could be realized for a specific feeder. The inability to predict, model or respond to secondary side voltage changes is being referred to as the grid edge control problem. This is a problem that is only now being understood, and for which traditional averaged model-based analysis and centralized command and control techniques do not work. This has been a challenge for traditional model-based and heuristic rule-based Volt/VAR Control techniques that are being used by utilities today. Figure 2. Dark blue line shows the measured secondary voltage profile along an actual feeder with 33 nodes. (a) Minimum voltage occurs at node 13 at 10 pm. (b) Minimum voltage occurs at node 29 at 11 pm Why is voltage variability at the grid edge such a concern? Utilities have clearly operated for decades with this variability and have not been greatly impacted in any way. Traditionally, the utilities have been able to manage their radial distribution grids based on conservative planning models to ensure that voltages remained within a fairly wide +/-5% ANSI band. That paradigm is now changing. Network reconfiguration under FLISR, distributed PV solar, grid optimization and energy conservation, improved asset utilization, electric vehicle charging, consumer pricing signals these are all new trends that require significant changes in the way the grid is controlled and operated, including substantially tighter control of the voltage at the point of common coupling (PCC), or the grid edge voltage. Smart grid investments have assumed that the existing control layer i.e. primary side LTCs, LVRs and capacitor banks could deliver this enhanced performance by adding sensing, communications (e.g. AMI), advanced analytics, and complex system optimization to squeeze out the desired savings as shown in Figure 3. If the grid edge had behaved as expected, this may have been possible. With a highly volatile voltage profile along the grid edge, it is not possible for centralized command and control, when operated in the traditional manner, to achieve the distributed control objective of regulating the voltage at all points along the grid edge. It is clear that a new paradigm is indicated Grid Edge Control. 4

5 Primary Distribution Network - LTC, Cap Bank, LVR, EOL sensor etc. Node 1 Node 2... V VVO SCADA Service Transformer Meter Headend Loss of control at grid edge Figure 3. Typical VVO/CVR implementation based on LTC/LVR and capacitor bank control, with set points based on feedback from AMI network. Loss of voltage control at grid edge limits CVR gains. III. GRID EDGE CONTROL Grid edge control requires that voltage at multiple points along a feeder be independently regulated so that broader grid operational objectives are achieved at lowest cost. This requires differing levels of control effort at various points which indicate a need for distributed control. An example is the case of a feeder where a grid optimization objective such as VVO/CVR is to be implemented. Conventionally, this requires that the LTC voltage tap is lowered until the lowest secondary voltage along the feeder is at the lower end of the ANSI band. At this time, if one node is 1 volt below the desired set point, the only option is to increase the entire feeder voltage level by 1 volt to ensure ANSI compliance. The corrective action is in the opposite direction of what is needed for achieving system-wide CVR goals, and results in sub-optimal CVR performance. Given that the lowest voltage continuously moves from node to node, best-in-class AMI-based CVR schemes track these ever-changing nodes and control the LTC set-point accordingly. Because an individual node voltage is stochastic and dynamic in nature, and because the level of variability is often very high, the lowest voltage along the feeder is not predictable and can vary widely by time of day and by season. One can begin to see why lack of grid edge control is a major limitation for grid optimization schemes that rely only on centralized command and control, particularly where the voltage at the PCC is the metric by which success is measured (all VVO/CVR and compliance programs). Grid edge control can be achieved through the use of individual controllers that mitigate the volatility at the problematic nodes along a feeder, allowing overall grid operational and optimization objectives to be achieved. Grid edge control can deliver the following benefits to distribution system operators: BENEFITS 2x improvement of grid optimization: CVR, VVO, peak demand reduction Improved grid integration of distributed PV solar and smoothing of dynamic loads Energy loss minimization, power quality enhancement Minimize need for peaking plants, investment deferral, and other programs Improved power quality and strengthening of weak feeders 5

6 Since the control effort needed at individual nodes varies with location and time, a centralized command and control technique cannot be used. A distributed set of controllers is needed at specific nodes. It is not possible to use slow, centralized command and control techniques to manage a distributed fleet of such controllers because of the dynamic nature of certain loads and distributed solar dynamics. Further, because these controllers are randomly deployed along the feeder, it is not practical to control them based on feeder models and peer-to-peer communications. Rather, these distributed controllers must operate in a truly decentralized manner, with dynamic and autonomous response based only on locally measured variables. In such a system it is critically important that, even as these devices operate to improve local and system level performance, their control algorithms decouple their individual response from that of other units in the vicinity. If this cannot be achieved, system stability would be compromised and oscillations could result. Slow communications to these devices from the central command and control can help coordinate their behavior with existing slower grid assets. This can allow the system to simultaneously achieve local and global optimization objectives, a unique capability. In addition, the distributed devices can provide visibility and diagnostics along the grid edge. This overall functionality represents a new paradigm in active grid control. In summary, grid edge control uses distributed and decentralized controllers that decouple the dynamics at the grid edge, allowing existing centralized command and control assets to achieve unprecedented levels of local and global grid optimization. The proposed approach fully complements all existing grid operation and optimization approaches by inserting a layer of distributed and decentralized control at the grid edge, as seen in Figure 4. Offline ENGO Manager Primary Distribution Network - LTC, Cap Bank, LVR, EOL sensor etc. V Node 1 Node 2... VVO SCADA Service Transformer Grid Edge Network - ENGO V-10 devices Meter Headend ENGO-V10 Unit ENGO-V10 Unit ENGO-V10 Unit Varentec s responsibility areas Figure 4: Typical feeder with centralized command and control with added Grid Edge Control capability IV. IMPLEMENTING GRID EDGE CONTROL Grid edge control employs a methodology that places individual regulators at specific nodes where localized control is needed. Locations are determined through a process of reviewing feeder models, AMI data, and other data available about the feeder. Secondary side control, or grid edge control, is distinct and different from what is needed at the global (i.e. feeder) level. There are two methods of implementing secondary side control injection of shunt VARs or injection of series voltage (Figure 5). The utility industry has typically used shunt VARs on the primary side to manage feeder primary side voltage profiles and to compensate for average feeder-level reactive loading and voltage drops due to line impedance. 6

7 A major contributing factor to the root cause of grid edge voltage volatility is the voltage drop across the distribution transformer leakage inductance (Figure 5). Leakage drop Primary Line Leakage drop V c... Series Solution Service Transformer I c Shunt Solution Time varying Loads Service Transformer Time varying Loads Figure 5: Shunt solution (left) and series solution (right) for Volt/Var control A secondary side shunt VAR device must first compensate for the voltage drop across the transformer shown in Figure 5. If a 1 volt increase in voltage is required to bring a 50 kva load into compliance, the transformer may need 300 kvars or more of shunt VAR injection on the primary side. The same level of voltage change can be achieved at that specific node by injecting as little as 3 kvar on the secondary side, a 100:1 reduction in level of control effort! Distributed VAR controllers allow adjacent nodes, which would be at the same voltage on the primary side, to see different levels of control action at each location and to realize targeted voltage regulation at multiple points along the feeder. This cannot be achieved with conventional primary side shunt VAR control the standard method for utilities today. An alternate method for achieving voltage control at a specific node is to insert a controlled voltage in series to add or subtract the needed voltage to bring the node into compliance. For the same example where a 1 volt increase is needed, the rating of the series injection device could be less than 1 kva for a 50 kva load. For a +/-5% control range, the rating could be as low as 5 kva, a fraction of the load rating. In both cases (series or shunt injection), the fractional rating of the control effort as compared with the load impacted, is a key reason for why this control can be done cost effectively. While this trade-off in the favor of pure series devices seems obvious, it should be noted that the voltage drop across the transformer is much worse when the load power factor is poor. In this case, a small rated shunt VAR compensator has the most beneficial impact. The trade-offs between shunt and series injection can be quite complex and need to be fully understood before a decision can be made. It appears that both types of control would be needed to achieve full distributed Volt/VAR control at the feeder level. Shunt devices install easily with no load interruption, and do not have to handle fault currents caused by downstream load-side faults. For a 50 kva transformer, this could be as high as 4,000 Amperes, almost 20X the full load current. Fault current handling can add substantially to the cost and complexity of a series solution. Series device installation also results in customer interruption. These devices need to manage high fault currents without interrupting customers. Series devices also worsen the voltage profile on the primary side for adjacent customers without local devices for voltage correction, while shunt devices actually improve the voltage for adjacent customers on the same feeder, even those with no devices connected at their respective nodes. This suggests that a best approach would deploy distributed shunt VARs to first improve the feeder voltage profile, and to add series devices for those locations where the voltage deviations still require additional correction. It should be noted that a purely series device approach would not improve the feeder voltage profile or voltage fluctuations due to distributed PV solar, as PV solar inverters are constant power devices that will push real power into the grid independent of the AC voltage at their terminals. Thus, if the objective of deploying series devices is to reduce operating cycles for traditional 7

8 LTC and LVR grid assets, it is unlikely to be realized in a pure series solution. Also, deploying a pure series solution for long feeders will cause the primary side voltage to drop even more sharply with distance, reducing the effectiveness of the distributed control. A balanced Volt/VAR approach will allow a longer flat voltage profile over the length of a feeder and manage load/source dynamics. In comparing methods, factors such as cost, weight, installation ease, and level of VAR and voltage correction available all come into play. A secondary side, pure series solution is likely to be more expensive than a pure shunt solution. Also, a series device does not offer any benefits on the primary side, and can often cause degradation in voltage profiles. An approach that combines series and shunt in an integrated Volt/VAR solution can be much larger, heavier, difficult to install and more expensive than a pure series or shunt solution. Experience shows that as much as 20-30% of the load kva rating is needed in sizing shunt kvars, while only +/-5% correction in voltage is needed for series compensators. Finally, losses can become a critical part of total cost of ownership for the two types of solutions. For instance, a series regulator for a 50 kva load with 1% losses could dissipate as much as 40 MWHr $50/MWHr) in losses over 10 years per device, as compared with a 3.2 MWHr ($160) for a shunt compensator device rated at 10 kvar, typical of what would be connected for a 50 kva load. Such losses deduct from the operational and financial benefits of the device. It is clear that the choice of series versus shunt control, or the use of an integrated device that offers both, requires the consideration of a complex set of parameters and considerations. Trying to integrate these two types of compensation into one unit can often increase the cost and size of the integrated device. Taking into account system performance and cost, it is likely that VAR (shunt) and Volts (series) control will be needed to fully manage the grid edge, although not necessarily in an integrated device. V. REAL EXAMPLES OF GRID EDGE CONTROL Varentec has deployed a distributed, dynamic solution called Edge of Network Grid Optimization (ENGO) at multiple utilities. The grid edge regulator device used is the ENGO-V10 shown in Figure 6, first installed in 2012 and now with over 300 units in the field. This device provides 0 10 kvar (variable) of fast distributed VARs to help stabilize and regulate the voltage at select points on a feeder, collectively able to help bring the entire feeder voltage into a tight range. The device can act in less than a cycle, responding to load and distributed PV dynamics, while incorporating unique algorithms that prevent multiple units from working against each other. The devices operate with no peer-to-peer communications, but feature slow communications to enable feeder-level optimization, while providing grid edge visibility and analytics. Sensing & Regulation Decentralized & Distributed Diagnostics & Analytics Figure 6: ENGO-V10 pole-top unit includes sensing, communications, analytics and actual voltage control 8

9 The ENGO solution has been designed and architected with utility input, incorporating key features such as 15 minute install without load interruption, pole or pad mount versions, flexible communication protocols, redundant and resilient operation, no impact on load in the event of unit failure, visibility to feeder-level voltage profiles, and extensive diagnostics capability at the individual unit and feeder level. The ENGO-V units are typically located at points along a feeder where AMI data or analysis shows the possibility of low voltage points, PV solar installations or dynamic disturbances. The grid edge VARs complement fixed and switched capacitor banks, and reduce the duty on existing LTC and LVR operation. Significant impact to feeder voltage consistency has been observed in field data. Figure 7a shows the voltage profile along a feeder over an entire day. The voltage is measured on the secondary side of distribution transformers by ENGO-V units with the VAR compensation turned off. This clearly shows that the voltage varies significantly from node to node over the course of the day, and does not decrease from the beginning of the feeder to the end as expected from the primary side voltage profile. Figure 7b shows the same feeder on a similar day with the ENGO-V units operating with a voltage set point of 245 volts. Each unit will inject VARs to try to regulate the voltage at its node to the set point, but will do so in a manner that avoids interactions with other neighboring units. Through the combined operation of over 40 units the voltage profile over this part of the feeder is dramatically improved. Not only is the incidence of voltage levels below the set point dramatically reduced, but also the feeder voltage volatility is reduced and the overall voltage more closely conforms to the set point at all times. The voltage regulation also indirectly improves node and system power factor. (a) ENGO-V Units OFF (b) ENGO-V Units ON Figure 7: (a) 3-D voltage profile of a feeder over a day. ENGO-V units are connected at 40 nodes that are maintained OFF. Voltage volatility shows that minimum voltage node migrates significantly as discussed, (b) Voltage profile over a day with ENGO-V units ON. Dramatic flattening of voltage profile is seen, with 5 volts of additional control range in a CVR or demand application. Figure 8 shows an example of the instantaneous voltage profile across a feeder showing a tightening of the voltage band. The red line also shows a minimum voltage of 236 volts, as opposed to 230 volts in Figure 2, an improvement of about 2.5% in voltage control compared with primary side control. Figure 8 also shows the level of VAR injection at each node, represented by green bars, showing the need and ability for decentralized control. 9

10 0.1 % 0.5 % 2.3 % % Time 2.6 % 3.7 % 4.6 % 6.3 % 7.4 % 7.4 % 7.4 % 7.4 % 6.9 % 6.9 % 7.6 % 8.4 % % Time Figure 8: Voltage profile (dark blue) over a day along the feeder. Red line shows minimum voltage seen over a day, while green bars shows VARs injected at each node at the specific time the data was sampled. With the ENGO-V units, the operating incidence of voltages below a band of (240-2) volts is reduced from 68.2% to 0.6% on similar days, as shown in Figure 9. Further, it is seen that when implementing a CVR program, the ENGO-V units enable about 5 volts in additional LTC-driven voltage reduction. This can provide as much as 2X the CVR-based energy savings as compared with typical primary side control strategies. 60 ENGO OFF Volts 60 ENGO ON Volts Figure 9: Histograms showing voltage spread over a day with ENGO-V units OFF (red) and ENGO-V units ON (green). Low voltage regions are substantially reduced. Examining the impact on distributed solar, Figure 10a shows a feeder with 2 MW of total PV operated on a partly cloudy day, showing significant volatility in the voltage as insolation levels change. Figure 10b 10

11 shows the same feeder on another partly cloudy day with ENGO-V units that provide noticeably reduced volatility due to the combined action of slow primary and fast secondary side assets. ENGO-V units are currently deployed on many feeders, with as many as units on individual feeders. In every case, the impact of distributed, decentralized, dynamic, and decoupled grid edge control can be verified. Fast sub-cycle response has been demonstrated, even as over 100 units provide feeder level support with no instabilities. (a) Figure 10: (a) 3-D plot of feeder with 2 MW of distributed PV solar on a partly cloudy day with ENGO-V units OFF, (b) on a partly cloudy day with ENGO-V units ON The ability of shunt VAR devices to regulate individual node voltages has been confirmed, even as conventional thinking questions whether such control can be done. This is possible because of the 2-10 volt drop across distribution transformers, a condition that had not been measured or observed previously. Even with ENGO-V support, there continue to be a few overloaded transformers where the ENGO-V units cannot provide complete correction. In such cases it is likely that additional VAR capacity can be added, or a series voltage injection device may be deployed. Conventional thinking also suggests that distributed shunt VAR devices will interact with each other and cause system instability. This traditional thought has been disproved by hundreds of ENGO-V devices operating in the field, in a coordinated fashion, without any peer-to-peer communications. (b) VI. CONCLUSIONS Utilities are beginning to understand the implications of lack of visibility and control at the grid edge, resulting in significant impact on key objectives of grid modernization. Grid optimization initiatives such as CVR and peak demand management, integration of distributed renewables, and improving reliability and quality for weak feeders, are challenges that cannot be addressed using conventional, centralized command and control techniques, and for which utilities have no tools in their toolkit. This white paper has identified a major contributor to grid edge problems voltage drops across distribution transformers resulting from load dynamics, and has shown that this can dramatically impact the level of benefits that utilities can achieve from many of their grid modernization initiatives. The problem is distributed and dynamic in nature and requires decentralized solutions that decouple the volatility at the grid edge from the slow and measured response that conventional grid assets can provide. Both distributed shunt and series solutions have been discussed, with an assertion that shunt VAR controllers support the feeder voltage profile and provide much of the decentralized grid control needed. In addition, it is anticipated that series voltage control devices are needed at select locations to fully gain control of the voltage along the grid edge. 11

12 Varentec s ENGO-V has demonstrated the ability to realize dramatically improved grid edge control through the use of distributed, decentralized and dynamic shunt VAR devices. Multiple pilots with over 300 devices in the field have validated the grid edge control issue and the ability of ENGO devices to address this rather complex problem. Gaining control of the grid edge will allow utilities to fully realize the benefits of many grid modernization programs, including advanced VVO/CVR schemes, peak demand management, line loss reduction, as well as reliability/quality enhancement. IMPACT OF GRID EDGE CONTROL Grid Optimization - Realize 2X Better CVR/Peak Demand Reduction - Improve Volt/VAR Optimization - Reduce Line Losses & Improve Power Factor Grid Integration (Dynamic Sources & Loads) - Improve Grid Integration of PV Solar - Reduce Impact of Load Dynamics - Reduce Operation of Primary Control Assets Grid Voltage Support & Visibility - Mitigate Low Voltage Pockets - Improve Power Quality Sags & Momentaries When radically new concepts are proposed, such as massively distributed grid edge control, there is often healthy skepticism about whether such technologies can actually work outside the laboratory in the complex environment that is the grid. Varentec, through field installations at numerous utilities in the US and internationally, has clearly established a new paradigm in how grid control can be implemented. Field data from these deployments has demonstrated that distributed dynamic grid edge control can be cost effectively realized and deliver key benefits to utilities. For further information, please contact Varentec at info@varentec.com. Varentec, Inc Atteberry Lane San Jose, CA Varentec, Inc. 12