TODAY’S STUDY: WIND AND SOLAR ON THE GRID
PJM Renewable Integration Study, Task Report: Review Of Industry Practice And Experience In The Integration Of Wind And Solar Generation
Kevin Porter, Sari Fink, Jennifer Rogers, Christina Mudd, Michael Buckley, Cali Clark, November 2012, (by Exeter Associates, Inc. and GE Energy for PJM Interconnection, LLC.)
Best Practices In Integration Of Variable Generation
The previous sections consisted of an extensive literature review on industry practice and experience with integrating wind and solar generation. This section summarizes the GE team’s views on the ‚best practices‛ in integrating wind and solar generation. Our format will be to state a best practice, followed by a short description of why it is considered a best practice. A section on additional options is also included to consider new and innovative practices that do not have sufficient operating experience to be fully classified as a best practice but should be monitored.
Energy Market Scheduling
Sub-hourly scheduling and dispatch, for both internal (within-RTO and within-utility) and for scheduling on external interconnections with other balancing authorities, is a best practice.
Sub-hourly scheduling and dispatch is considered a best practice as it allows grid operators to follow net load variability over finer time scales than hourly, as is the case in most non-RTO regions in the U.S. In addition, sub-hourly scheduling unlocks existing generation flexibility that would not be as available in hourly scheduling and dispatch. Sub-hourly scheduling and dispatch is common among most of the RTOs in the U.S. Sub-hourly scheduling among external interconnections with other balancing authorities is not as common. Most external schedules between RTOs and between utilities not in RTOs tend to be hourly schedules. That said, some RTOs such as PJM are instituting sub-hourly scheduling on some external interties. PJM, for example, has adopted sub-hourly scheduling on external schedules with MISO and started sub-hourly scheduling at NYISO’s Keystone proxy generator bus on June 20, 2012.
Visibility of Solar Distributed Generation
Install telecommunications and remote control capability to clusters of solar DG in PJM’s service area. Alternatively, have distribution utilities install such capability and communicate data and generation to PJM.
Include solar in variable generation forecasting.
Account for the impacts of non-metered solar DG in load forecasting.
Follow and/or participate in industry efforts to reconcile provisions in IEEE-1547 and Low-Voltage Ride-Through Requirements.
PJM is likely to have significant amounts of additional distributed generation, particularly distributed solar, over the next several years, particularly in response to the solar requirements that are part of many state RPS requirements in the Mid-Atlantic. The increased amount and the lack of visibility of distributed generation could negatively affect load forecasting. In addition, the lack of reconciliation between low-voltage ride-through requirements and IEEE-1547 could lead to potential future reliability issues for PJM.
Absent such reconciliation, PJM can undertake various measures such as installing telecommunications and remote control capability in areas with large amounts of solar DG. Alternatively, the transmission owners and/or distribution utilities can install telecommunications and remote capability and communicate data to PJM. Solar will need to be accounted for in both generation forecasting and load forecasting. Although the emphasis in this section has been on solar DG, utility-scale solar is also growing. PJM has 3.6 GW of solar in its interconnection as of the end of 2011, although only 40 MW of solar was on-line at that time.346 Utility-scale solar can be integrated into wind forecasting systems. Solar forecasts can be prepared by metering a fraction of utility-scale solar on the grid and scaling up the forecast and production data by participating solar plants. In addition, solar plants can be required to participate in a variable generation forecasting system as a condition of interconnection, and provide solar resource and production data as they come on-line.
Non-metered solar DG can affect the accuracy of a load forecast if there is a significant amount of non-metered solar and it is not accounted for in load forecasting. Non-metered solar DG can be estimated through measuring the difference between full sun output and a forecast based on current weather conditions.
Consider separating regulation requirements into regulation up and regulation down if there is a shortage of regulation for certain hours, if there is a disproportionate need for a certain type of regulation (up or down), or if there is a desire to more finely tune regulation requirements.
Have operating reserve requirements set by season or by level of expected variable generation, instead of a static requirement that changes infrequently.
Rely on demand response to provide some reserves.
Require wind and solar generators to be capable of providing AGC.
Most RTOs require that regulation is a combined service, namely that both regulation up and regulation down are provided as a single service. Two RTOs – CAISO and ERCOT – have separated regulation requirements into regulation up and regulation down. Doing so recognizes the different demands that variable generation places on regulation (such as on regulation down during high wind/minimum load events). Furthermore, it would allow wind generators to potentially provide regulation down if called upon. That said, RTOs and other grid operators that have ample quantities of regulation and can draw upon look-ahead unit commitment tools and sub-hourly markets to identify the need for regulation may not need to separate regulation into regulation up and regulation down. That may be the case for PJM.
Some grid operators also adjust their regulation requirements to vary when higher levels of variable generation are expected, such as by season or by month, or to adjust regulation requirements based on changes in installed variable generation capacity, as ERCOT does.
Turning to demand response, recent variable generation integration studies such as the WWSIS have noted that demand response could be called upon to provide contingency reserves instead of requiring additional generation. ERCOT allows up to half of its non-spinning reserves (1,150 MW) to provide 30-minute non-spinning reserves. Other RTOs also use demand response to provide ancillary reserves, although the amount is still relatively small. Because demand response is still relatively new, some RTOs limit the amount of demand response they will use as they gain greater operating experience. Demand response in PJM, for instance, can supply regulation, synchronized reserves and day-ahead reserves, but demand response is limited to 25% for each category, and demand response can serve only two of the three categories. Other RTOs and grid operators have comparable limits for demand response. These limits will likely be eased as greater experience with demand response is gained.
Variable generation can provide various types of reserves if the reserves can meet the technical requirements. Only MISO and CAISO specifically prohibit wind power from providing reserves. Wind generators can accept AGC signals but would require wind generators to spill a portion of wind energy to do so. Grid operators should require wind and solar generators, as a condition of interconnection, to be capable of providing AGC. Wind generators, for example, can provide down regulation during nighttime hours or when there is a risk of minimum load.
Wind and Solar Forecasting
Implement a centralized forecasting system for wind and utility-scale solar that offers day-ahead, very short-term (0-6 hours), short-term (6-72 hours), and medium or long-term forecasts (3-10 days).
Incorporate estimates of non-metered solar production into day-ahead and short-term load forecasts if there is a significant amount of solar DG either already installed or predicted.
Ensure that short-term wind and solar forecasting systems can capture the probability of ramps, or implement a separate ramping forecast.
Institute a severe weather warning system that can provide information to grid operators during weather events.
Monitor the use of confidence intervals and consider adjusting them periodically.
Integrate the wind and solar forecasts with load forecasts to provide a ‚net load‛ forecast.
Institute requirements for data collection from wind and solar generators. A growing number of utilities, transmission providers and RTOs are implementing wind forecasting, and a smaller number are expanding into solar forecasting. The different time frames of forecasts serve a unique purpose and should be part of any forecasting system. The medium to long-term forecasts provide a climatological overview of what to expect, such as the possibility of storms or high wind events. The day-ahead forecasts, as the name implies, will provide day-ahead wind and solar forecasts. Short-term and very-short-term forecasts will provide near-term updates of wind and solar forecasts.
The short-term and very-short-term forecasts should be evaluated to see whether they are adequate in predicting wind ramps, or whether a separate ramp forecast is needed. There is some controversy among variable generation forecasters and the wind power industry as to whether a separate ramp forecast is necessary as some maintain, or whether continual and frequent updating of forecasts is sufficient for predicting ramps. PJM, for example, is relying on its short-term wind forecast as part of its Intermediate Security-Constrained Dispatch to identify potential wind ramps in the near future. Regardless of whether a separate ramp forecast is adopted or not, a severe weather warning system should also be added to advise grid operators of the potential for extreme meteorological events.
Except for CAISO, we are not aware of any utility, RTO or transmission provider that is forecasting for non-metered solar. Because of the rapid growth of solar DG systems, some at least rudimentary forecasts may be needed for non-solar DG. Otherwise, load forecasts may be inaccurate on high solar days.
A number of utilities and RTOs use a confidence interval as part of their forecasts. These confidence intervals should be periodically evaluated and perhaps changed if the forecasts are proving sufficiently accurate and grid operators are comfortable with the forecast.
Previous variable generation integration studies have suggested that incorporating variable generation forecasts directly into reliability commitment schedules, such as using a load net wind and solar forecast, would result in reduced total system operating costs through decreased fuel consumption, operation and maintenance costs, and more efficient plant dispatch overall. Few, if any, grid operators use their wind and solar forecasts in this manner, and are not capturing the full benefits of forecasts. A combined ‚load net wind‛ forecast could be used after clearing the financial day-ahead market in the reliability commitment process (usually considered the first stage of the next day’s real-time market). The ISO and RTO process to commit sufficient resources to supply anticipated load may have to account for the increased uncertainty around the wind power forecast. That said, the ‚load net wind‛ forecast should contribute to more efficient market operation and dispatch, should improve overall operating reliability, and should not financially benefit wind and solar generators over what they would otherwise receive as price-takers in the real-time market. It is comparable to the use of an improved system load forecast that is created by the ISO or RTO.
With regard to data for forecasts, FERC Order No. 764, issued in June 2012, requires transmission providers to amend their pro forma Large Generation Interconnection Agreements to institute data requirements for wind and solar generators over 20 MW. More specifically, FERC is requiring wind generators to provide site-specific meteorological data including, but not limited to, temperature, wind speed, wind direction, and atmospheric pressure. For solar generators, FERC is also requiring site-specific meteorological data including, but not limited to, temperature, atmospheric pressure, and irradiance. FERC-jurisdictional entities that have forecasting systems may want to evaluate whether additional data is needed and include such requirements in their compliance filing to FERC. In addition, some RTOs impose penalties on wind and/or solar generators if they fail to provide data, or do not provide quality data. FERC-jurisdictional entities may want to decide whether they want to include such a provision as well.
Intra-Day Unit Commitment
Consider establishing intra-day unit commitment, if one is not already in place, and incorporate short-term wind and solar forecasts.
Wind and solar forecasts are more predictable and more accurate the closer they are to real-time as compared to day-ahead forecasts. Running intra-day unit commitment algorithms, in addition to day-ahead unit commitment, and using the results to inform forecasts – or using a more stochastic approach to unit commitment with frequent rolling updates – are useful strategies for taking advantage of short-term variable generation forecasts. As noted earlier, PJM uses short-term wind forecasts when it runs Intermediate Security-Constrained Dispatch. ERCOT is considering whether it can review wind and solar forecasts six hours ahead of real-time operations.
Consider Establishing a Look-Ahead Dispatch for Very Short Time Frames
A variation of intra-day unit commitment is MISO’s Look-Ahead Unit Dispatch System for even shorter time periods, such as two hours ahead. Look-ahead dispatch could result in more precise scheduling of variable energy generation and less need for manual actions by grid operators in response to changes in system conditions, load demand or generation, whether from variable energy generation or other generation. PJM has something comparable with Intermediate Security-Constrained Dispatch that looks ahead from 15 minutes to two hours, with grid operators able to adjust the look-ahead time.
Capacity Value of Wind and Solar
Conduct an ELCC Study of Wind and Solar at Regular Intervals
Recent work from the NERC IVGTF concluded that the ELCC approach is superior to time-based approximation methods (e.g., the capacity of wind or solar during peak hours). Time-based approximation methods have the disadvantage of assuming the hours included are weighed equally, while ELCC methods put greater weight on higher-risk hours. That said, approximation methods are often used if data is unavailable. Such methods can be reasonable if they are regularly benchmarked to an ELCC study. MISO conducts an ELCC study of wind every year. That may be too frequent for some grid operators, but ELCC studies should not be considered a one-time occurrence, as wind and solar production can vary from year to year.
Require Wind Generators to be Equipped with Ability to Limit Ramps
ERCOT, BPA and AESO are among the grid operators that impose ramp rate limits on wind generators. Such ramp rate limits are reasonable in smaller balancing areas with limited interconnections and high amounts of wind. For larger balancing areas, simply requiring wind plants to have the ability to limit the rate of power increase, to be enabled or disabled by instruction from PJM, is sufficient. Such ramp rate limits are not necessary under all operating conditions but can be useful in certain circumstances. If dispatched down or knocked off-line for reasons such as cutting out for high wind speeds, wind generators can ramp up to full capacity very quickly. In that case, the use of ramp limits on wind plants may be useful. In addition, grid operators should not impose ramping down limits due to decreases in wind speed, although such limits due to curtailment, shut-down sequences, or other control measures can be reasonable.
Do Not Impose Frequency Response Requirements on Wind Generators Unless it is Absolutely Necessary
ERCOT, BPA and AESO are among grid operators who are requiring wind generators to provide frequency response. However, this requirement incurs a power efficiency penalty for wind generators. Frequency response can also be obtained more economically from other generation sources than wind.
Other Potential Best Practices
There are new innovations and practices that show promise but have not garnered enough operating experience to be classified as a best practice. These are discussed further below.
Short-Term Dispatch and Scheduling Requirements for Wind
Consider Including Wind in Short-Term Dispatching and Scheduling
Several RTOs have instituted short-term scheduling requirements for wind generators to follow, with some RTOs imposing penalties for non-compliance. The details vary by RTO.
MISO’s Dispatchable Intermittent Resources require the submission of 5-minute forecasts by node, or the acceptance of MISO’s default wind forecast. MISO can levy an Excessive or Deficient Energy Deployment Change if an 8% tolerance band is exceeded for four or more consecutive 5-minute intervals within an hour.
ERCOT may impose penalties on wind plants if wind plants have been given an economic dispatch below their high dispatch limit, and wind plants deviate more than 10% from that base point.
NYISO requires wind resources to bid a price curve in real-time, then uses those price bids to determine reduced base points for each wind plant during economic dispatch. If wind is dispatched down, NYISO can levy over-generation if wind generators do not follow the dispatch signal. Wind is exempt from under-generation charges.
PJM economically dispatches wind plants based on a wind plant’s offer in real-time.
With the help of wind forecasts, short-term scheduling and dispatch requirements for wind generation can help improve overall scheduling and perhaps reduce the need for short-term reserves. It can also serve as a transition towards incorporating wind and solar forecasts into day-ahead scheduling and dispatch, which is classified as a best practice.
Contingency Reserves and Variable Generation
Consider Using Contingency Reserves for Very Large but Infrequent Wind Ramps
To date, variable generation integration studies and the experience of a small sample of RTOs and utilities discussed earlier have found that higher amounts of wind capacity do not lead to an increased need for contingency reserves, as wind facilities tend to be smaller in size and an instantaneous drop in wind capacity is highly unlikely. However, as noted earlier, the loss of DG solar capacity could potentially be considered a contingency event should DG solar capacity reach projected capacity additions, and if IEEE-1547 and low voltage ride through requirements are not reconciled.
Separately, the electric power industry is debating whether contingency reserves can be called upon for large but infrequent wind ramps. Few do so currently. As noted earlier, NERC indicates that it may be appropriate to use contingency reserves in response to large but infrequent wind ramps, while allowing time for other resources to cover the wind ramp. This issue bears further monitoring, but using contingency reserves to at least partly cover very large wind ramps may be an economical means of addressing very large wind ramps, instead of building new generation that would operate for only a small number of hours per year.
New or Revised Reserves
Consider Establishing a Slower Responding and Longer-Lasting Reserve to Cover Wind Ramps
Monitor Industry Initiatives to Acquire or Encourage More Flexible Reserves
Present NERC requirements dictate that contingency reserves be restored no more than 105 minutes from the start of a contingency event. Most wind ramps last longer than that, and this has given rise to discussion as to whether a slower responding and longer lasting reserve may be needed. Alternatively, layers of different operating reserves can respond to a wind ramp and at different time intervals, or the energy market itself, if the market is deep enough and flexible.
Some RTOs and utilities are considering whether to introduce new reserves in anticipation of higher levels of variable generation, such as CAISO’s Flexible Ramping Constraint or the MISO Ramp Management project. These initiatives are quite new, thus not enough operating experience has been obtained to evaluate these initiatives and to recommend a ‚best practice.‛ However, the differences in approach between CAISO and MISO is noteworthy, as CAISO has developed a new service and cost structure, while MISO is essentially holding back more generation in its commitment stack to ensure ramping needs are met.
Monitor Industry and Regulatory Discussions on Integration Charges
Load pays for most types of reserves, but there is increasing discussion as to whether variable generators should pay for part or all of the costs of certain reserves. In its final rule on variable generation, FERC recently decided not to require transmission providers to add a generator regulation service but instead to consider any proposed charges on a case-by-case basis. We note that there is considerable disagreement within the electric power industry and among academic researchers over how to craft integration charges. Therefore, no recommendation is being made at this time other than to monitor industry developments on this issue.
Do Not Rely Upon Virtual Bidding to Cover for Forecasting Errors
Some in the electric power industry have suggested that virtual bidding can address scheduling and dispatch inefficiencies from wind forecast errors. There is little academic or industry research to support or contradict this view. Our view is that the use of a state-of-the-art forecasting system, coupled with incorporating the forecast into unit commitment and operations, should drive forecasting errors as low as possible and should not leave consistent and sustained opportunities for virtual bidding. However, if the forecast is not state-of-the-art and is not incorporated into unit commitment and operations, then more reserves will likely have to be committed. More opportunities will be available for virtual bidders who use state-of-the-art forecasts to capture the financial gains resulting from poor operating practices. The energy market will settle satisfactorily but costs will be higher than necessary.