Grey System Theory Based Net Load Forecasting for High Renewable Penetrated Power Systems

Abstract

Net-load is the imbalance between aggregated load and renewable generation. System operations like economic dispatch necessitate accurate forecasting of net-load. There has been significant progress in load and renewable generation forecasting, however, little research focus has been there on Net-Load Forecasting (NLF). Therefore, this paper proposes three Direct Grey index net-load forecasting models (DGM (1, 1), DGM (1, 2) and DGM (1, 3)) using Grey System Theory (GST). GST based models are suitable for accurate and fast very short-term (five minutes ahead) forecasting due to their momentum transfer behavior. Proposed NLF models are implemented for the Bonneville Power Administration (BPA) balancing area. Forecasts obtained from proposed models are compared with actual net load data and forecasts obtained from reference Artificial Neural Network (ANN) model. Comparison with actual net load shows that proposed models have strong potential for very short-term NLF. At very short time frames, net load shows a very high correlation with previous time steps data. Proposed models utilize such characteristics for forecasting, compared to continuous error reduction procedure in ANN. Continuous error reduction in a very short time frame can lead to under/overestimation of ANN weights and that lowers the forecasting accuracy. Proposed NLF models, especially DGM (1, 3) show superior performance over ANN.