Supply and Load Forecasting. The energy sector uses near-term (1 hour to 2 weeks) and medium-term (monthly and seasonal) load forecasting models to adjust electricity generation to meet demand and to optimize generation and purchase of lower cost power sources. Load forecasting tools use inputs such as existing and historical 71 U.S. EPA, Heat Island Effect, Energy Savings, xxxx://xxx.xxx.xxx/heatisland/about/energysavings.html. Accessed July 21, 2005. 72 U.S. EPA, Heat Island Effect, Research, xxxx://xxx.xxx.xxx/heatisland/research/index.html. Accessed July 21, 2005. loads, meteorological data including real-time and weather forecast data, economic data such as the cost of energy, and available energy sources. Potentially Relevant NASA Resources ▪ Meteorology data, including temperature, cloud cover, and visibility ▪ Solar radiation ▪ Wind resources ▪ Reservoir height and rainfall ▪ Seasonal climatology models Current Status/Relevant Issues: Several basic load models exist, one of the most common types being artificial neural network models, which use historical loads and predicted temperature. Statistical regression models are also used to estimate load. Statistical economic models are used in conjunction with load forecasting models, although they are more difficult to predict than load.73 Supply and load forecasting models have also been developed for specific types of energy, such as wind power, that incorporate relevant resource parameters.74 A variety of nonprofit organizations and private companies have developed load forecasting software marketed to the energy industry. For example, the Electric Power Research Institute (EPRI) has developed the Artificial Neural Network Short-Term Load Forecaster, which forecasts hourly system loads from 1 hour to 35 days ahead with errors less than 3%.75 NASA’s contribution would be greater in medium-term forecasting, since NOAA as an operational agency provides the real-time data needed for short-term forecasting. NASA is establishing a partnership with EPRI focused on the possibility of integrating NASA data and models into EPRI’s load forecasting decision support systems.76 73 Xxxxx, X., Electricity Load and Price Forecasting Using Statistical Methods and Models, Second Moment, xxxx://xxx.xxxxxxxxxxxx.xxx/articles/electricity.php. Accessed August 2, 2005. 74 Neural Network-Based Forecasting for Wind Generation, Strategic Science and Technology, EPRI, September 2003.
Supply and Load Forecasting. Short- to medium-term load forecasting is typically performed using software-based decision support tools that require numerous near-real time inputs for decision-making. NASA weather and energy resource data could be integrated into these forecasting models relatively easily, provided that specific products were developed and available on a timely basis. NASA is already pursuing a partnership with EPRI who has an existing load forecast decision support system. Partnerships with NOAA would likely be required to ensure continued operational datasets and a smooth transition from research to operations. Load forecasting is closely aligned with GEOSS and has potential global application. With existing data, nascent partnerships, and alignment with GEOSS, supply and load forecasting would be an optimal subject for further development.
