Multistage Artificial Neural Network Short-Term Load Forecasting Engine With Front-End Weather Forecast

摘要

A significant portion of the operating cost of utilities comes from energy production. To minimize the cost, unit commitment (UC) scheduling can be used to determine the optimal commitment schedule of generation units to accommodate the forecasted demand. The load forecast is a prerequisite for UC planning. The projected load of up to seven days is important for the allocation of generation resources. Hour-ahead forecast is used for optimally dispatching online resources to supply the next hour load. This paper addresses the systematic design of a multistage artificial-neural-network-based short-term load forecaster (ANNSTLF). The developed ANNSTLF engine has been utilized in a real utility system. The performance analysis over the past year shows that a majority of the forecast error was detected in a consistent period with a large temperature forecast error. The enhancement of ANNSTLF is proposed to improve the forecasting performance. The comparison of forecasting accuracy due to this enhancement is analyzed.