758 Applications of the Renewable Energy Network Optimization Tool (ReNOT) for use by Wind & Solar Developers: Part II

Wednesday, 26 January 2011
Randall Alliss, Northrop Grumann Corporation, Chantilly, VA; and R. Link, D. Apling, H. Kiley, M. Mason, E. Martin, and G. Higgins
Manuscript (1.9 MB)

Handout (2.0 MB)

As the renewable energy industry continues to grow so does the requirement for atmospheric modeling and analysis tools to maximize both wind and solar power. Renewable energy generation is variable however; presenting challenges for electrical grid operation and requires a variety of measures to adequately firm power. These measures include the production of non-renewable generation during times when renewables are not available. One strategy for minimizing the variability of renewable energy production is site diversity. Assuming that a network of renewable energy systems feed a common electrical grid, site diversity ensures that when one system on the network has a reduction in generation others on the same grid make up the difference. Part one of this paper introduced ReNOT and its capabilities. This paper presents two case studies on applying ReNOT to the wind and solar farm industry, respectively. As part of collaboration with Rocky Mountain Supercomputing Center (RMSC) and the State of Montana a study was performed to estimate the optimal locations of a network of wind farms. Comparisons were made to four existing wind farm locations in Montana including Glacier with a 210 MW name plate capacity, Horseshoe Bend with a total capacity of 9 MW, Diamond Willow with a capacity of 20MW and Judith Gap with a total capacity of 135 MW. The goal of this study was to see if ReNOT could find a four site network that made more effective use of the existing four site network of wind farms' 374 MW nameplate capacity. We developed three different metrics in which to pick sites. Metric 1 (M1) picks sites that converges on the single best location for power production, on average. Metric 2 (M2) picks sites that maximizes geographical diversity, even at the expense of very poor aggregate power. Metric 3 (M3) picks sites based on the previous day's mean power, and accounts for short-term variability (i.e., 1 hour). In a sense M3 attempts to approximate usable power by minimizing ramping events which are so important to industry. In addition we investigated several performance metrics including Mean Power, Usable Power, and ramping event frequency. A ramping event is defined as an increase or decrease in power production over the course of one hour. Of interest was the frequency of ramping events that exceeded 10% of total capacity for the network. Networks with few ramping events are markedly superior to networks producing otherwise identical aggregate power. The optimization was run over the 15 year period (1995-2009) of hub-height wind data (40 meters AGL). Figure 1 indicates the existing wind farms in white while the optimized network is shown in yellow. The ReNOT derived network produces 58% more usable power than the four existing and operating wind farms denoted in white. In addition, the optimized four site network produces three times fewer significant ramping events. The solar study was performed over south Florida. The goal of this study was to see if ReNOT could find four optimal locations that exceed the peformance of a randomly picked four site solar farm network. The optimization was performed over a 15 year period (1995-2009) using a GOES derived cloud analysis at 4km and 15 minute resolution, respectively. As with the wind study we developed a cost function that emphasizes network stability, total power and day ahead forecastability. Networks with more consistent day to day cloud cover and are more accurately forecastable by a day in advance will be favored by ReNOT. Figure 2 shows the results of the optimization. The white x's represent the randomly picked solar farms while the red, yellow and white dots represent the top three networks from the ReNOT run. ReNOT prefers sites on the west coast of Florida due to favorable cloud conditions. However the optimized networks only produce 10% additional power than the existing/proposed sites. This is due to the limited area within which the optimization was artificially constrained (yellow box). The runs performed in this study were made with out regard to other practicle restrictions for example, building in state parks, population centers or within proximity to electrical grid infrastructure. However, we are currently adding this capability into ReNOT. Additional detailed results will be presented at the conference.

- Indicates paper has been withdrawn from meeting
- Indicates an Award Winner