|SDG&E will purchase 315 MW of output from the Ocotillo Wind Energy Facility under a 20-year power purchase agreement. Photos courtesy: Pattern Energy|
San Diego Gas & Electric tries to model the impact of intermittency in 2020, today
By Gail Freddo, SDG&E Transmission Planning
California’s Renewable Portfolio Standard (RPS) requiring investor-owned utilities, electric service providers and community choice aggregators to increase procurement from eligible renewable energy resources to 33 percent of total procurement by 2020 is the most aggressive in the nation, compelling providers to meet the challenge head on. Large-scale integration of distributed energy resources such as wind and solar with electric power systems presents operational and planning challenges that all regulated utilities will eventually face in the years ahead as more renewable capacity comes on line.
No stranger to innovation, San Diego Gas & Electric (SDG&E), a Sempra Energy utility, wanted to get ahead of the renewables curve by commissioning a study in late 2010 examining how intermittent wind and solar projects planned for Southern California’s Imperial Valley would affect its transmission and distribution systems. The obvious questions were:
- 1. How do we predict system performance issues that may arise as power sources shift to accommodate the variable output levels of renewable generation and increased loading variability of distribution circuits?
- 2. What are the potential risks to our system and what additions may be required in order to maintain compliance with the NERC/WECC/ and regional California ISO (CAISO) planning standards?
A CAISO report issued at about the same time predicted that overall, current and planned renewable sources in the long-range plan’s regional grid expansion portfolio would be adequate to accommodate the 33 percent RPS. SDG&E, however, wanted to take a closer look at how their system would hold up with a higher percentage of intermittent resources supplying energy to their service territory.
Cause for concern
Distribution systems the size of SDG&E’s represent a multi-billion dollar capital investment over a period of several decades, with the original design being highly reliable when operated in the intended fashion. The addition of new, intermittent sources of generation at various locations creates potential reliability impacts which the original system designs were not intended to address. As is fairly standard, across SDG&E’s system, 12 KV distribution lines are arrayed in a tapered format, where the wires are of a higher gauge at substations and become progressively smaller with more numerous branches as they fan out to various loads. Because photovoltaics, or PV, could be connected anywhere along this complex distribution infrastructure, the reliability of previously well-balanced sections can get called into question. As a typical example, there was reasonable concern about interference due to the intermittency from, say, a 2-MW project at the extreme end of a circuit and how it would impact reliability.
Presently, the only renewable energy project in SDG&E’s service territory is a 50 MW wind project that came on line in 2005 at the extreme periphery of their 69 KV system. This interconnection has caused many issues with voltage swings (signatures) at the point of interconnection. That in itself is not extreme but grid stability gets called into question when you multiply that by a factor of 10. Given utility focus on reliability of service, the potential impacts have to be considered so as to determine when and where the matter can become more serious. Similar conditions with larger projects at other utilities have shown this to be true.
Equally important was the matter of voltage regulation and accommodating renewables output as a “must take” source as power is being produced, thus taking priority over conventional sources. This requires having enough spinning reserves available in the interconnection across the system to continue output in case a wind farm and or a PV project goes out of service (ie. a cloud passes overhead or the wind suddenly drops off). Cost accounting methods are designed to account for things that are common to any generation source, as well as the very real additional cost impacts associated with addressing the higher unpredictability of wind and solar sources (e.g. there is an absolute requirement that a certain percentage of standby generating capacity be on line in order to maintain system reliability in the case of unforeseen contingencies.).
|The Ocotillo Wind Energy Facility reached commercial operation in December 2012.|
The Generator Interconnection Group at SDG&E routinely studies the cumulative impact of all generation resources as part of the CAISO Generator Interconnection Process (GIP).These impact studies are very specific in scope and require that all generators in the interconnection queue be modeled. SDG&E, however, was looking ahead several years to the 2015-2018 timeframe when contemplating its research criteria as a number of projects were slated for completion in that window and they had to forecast power purchase agreements for roughly the same period of time. Regionally, SDG&E is expecting at least 150 MW of power from a proposed wind farm and an additional 150 MW of rooftop solar.
Along with needing to address reliability impacts, renewables programs need to address financial impacts, ensuring long-term viability (see box).
Achieving an average of 33 percent of total renewable energy served over a year’s time means that, on any given day, there could be a wide range of actual associated renewable capacity being utilized at any particular time, given the fluctuations in output involved from these power sources. The steadiness of a power source is an important benefit cost-wise, with the more unpredictable and volatile sources creating added costs for the system planner. An initial hypothesis held that even with an average of a certain level such as 33 percent for renewable injection, the quality of that energy would in all likelihood not be of the same quality as conventional generation because it didn’t have that same system inertia.
The investigation performed with DNV KEMA was a non-standard study to divine, with a relatively high degree of granularity. Assessing the reliability of the grid while interacting with accumulated intermittent sources proved to be a challenge.
Modeling grid behavior
The study was conducted over nine months using commercially available software tools GridView from ABB, GE’s PSLF (Positive Sequence Load Flow), and combining their data with DNV KEMA’s own proprietary modeling software, KERMIT.
GridView performs regional energy market simulation and analysis for modeling of large-scale transmission networks by factoring in network topology, hourly load profiles, transmission constraints and generator data (heat rate curves etc.). It was used in this exercise to measure generator dispatch values and branch flows over a period of time. PSLF provided snapshots of time during simulations of dispatching large blocks of power across the transmission grid.
What interested SDG&E was that DNV KEMA’s KERMIT tool could incorporate the economic day-ahead studies and power flow data gathered from the other systems. With GridView and PSLF calculating intermittency of renewable plants, KERMIT could show the results of any mitigating actions taken to cope with intermittency. Such initiatives to better integrate renewables into the grid included the need to analyze the benefits of energy storage, optimal capacitor placement, integration of demand reduction programs, and requirements for new transmission capacity, to name a few. The solution was then employed to extrapolate the data to reflect system behavior in the year 2020.
The study began by analyzing a years’ worth of data from GridView to identify eight sample days throughout 2010 with data mined from SDG&E OSIsoft Pi historian (Table 1). These eight days were selected based on the level of renewable generation penetration, the SDG&E load and generation dispatch so that any drop in voltage may have the worst possible impact. Within this grouping, KERMIT simulations isolated April 26th (referred to as “2020 Off Peak” case) and September 4th (referred to as “2020 Peak” case) as the two days with the largest frequency deviation. April 26th was selected over January 18th because it had more renewable generation dispatched.
Expunging this data was the most time consuming portion of the nine-month study, as DNV KEMA required two- to four-second interval data over a calendar year. This translated into many gigabytes of data transferred out of the historian and then feeding it into KERMIT. Fifty-six Pi tags were necessary to acquire all of the aggregated data of information requested. On a two- or four- second interval, this protocol entailed utilization of a great deal of data, even though it did not yet include the frequency and interface data from CAISO that was also entered. For the purposes of the study, SDG&E and DNV KEMA later decided for a slightly coarser sampling to speed up the process.
Characteristics of renewable resource behavior in a time domain of high intermittency can dramatically affect the control functions and ancillary services required in the system to operate reliably and in compliance with NERC standards (e.g., frequency response/droop, regulation/AGC, spinning reserves and balancing services). The time domain modeling features of KERMIT are closely correlated with these critical power system performance requirements.
Results of the KERMIT simulations shown in Fig. 2, where the renewable event is defined and the response provided. To study the worst impact, outages of the San Onofre Nuclear Generating Station (SONGS) operated by Southern California Edison are considered as an additional contingency. This plant provides the bulk of base load to the region. A description of the renewable event is indicated as:
- The Disturbance starts at 11:15:18 am
- Solar power drops 1053 MW in 1-sec
- Wind power drops 493 MW in 5-sec
In this case, only 13 conventional SDG&E power units are in service and the total load is 2,580 MW. The SONGS contingency is applied one second after the renewable event has ended and the system frequency is at its lowest (at time stamp 11:15:24). The system frequency is at its lowest value, 59.56 Hz, at 11:15:30, six seconds after the SONGS contingency. The frequency recovers in two stages. The first recovery is seen at 11:16:06 due to governor action. The second recovery is seen at 11:20:00 when the real time balancing market kicks in. The time instant corresponding to Point Z is 11:20:16.
In addition to uncovering system behavior during a renewable event, the study also tested mitigating actions to address system disruptions from intermittency.
Energy storage can reduce the extent of frequency decay, but to be effective, the size of energy storage must be sufficient to cover the supply deficit that arises due to a drop-off in renewable power. It is best to view the role of storage as a very fast-acting generator providing short-term energy injection onto the grid. For example, if there was a drop-off in wind resulting in an 870 MW deficit for five minutes, a conventional plant takes time to ramp up in order to respond to this disturbance while energy storage can respond almost instantly. The main barrier to energy storage, however, is the current cost. Prices will have to drop substantially for it to be competitive.
Demand reduction is an option but it would have to exceed more than 10 percent of load to be effective as a solution. More detailed investigation is necessary on demand reduction parameters such as ramp and dispatch rates. Fig. 3 shows the settle down frequency (Hz) between having no load reduction to 10 percent.
Other solutions considered in the study included capacitor placement, dropping load, line upgrades, generator re-dispatch and reconfiguring existing protection schemes.
More work needs to be done on specific renewable resource behavior impacting the SDG&E grid, but the information from this initial study achieved through a credible simulation created valuable insight on a very complex issue. The risk depends on system conditions including network topology, available reserves, loading and the penetration level of intermittent generating resources. It will be interesting indeed to actually see the penetration numbers in 2020, but current tools are most assuredly providing a reasonably accurate landscape.
Power Engineerng Issue Archives
View Power Generation Articles on PennEnergy.com