Recurrent Neural Network Based Short-Term Load Forecast with Spline Bases and Real-Time Adaptation

Short-term load forecast (STLF) plays an important role in power system operations. This paper proposes a spline bases-assisted Recurrent Neural Network (RNN) for STLF with semi-parametric model being adopted to determine the suitable bases constructing RNN model. To reduce exposure real-time uncertainties, interpolation is achieved by adapted mean adjustment and exponentially weighted moving average (EWMA) scheme finer time interval adjustment. circumvent effects of forecasted apparent temperature bias, temperatures issued weather bureau are adjusted using errors over preceding 28 days. The proposed trained 15-min data from Taiwan Power Company (TPC) has been used operators since 2019. Forecast results show that RNN-STLF method accurately predicts short-term variations demand studied period. calibration can help accommodate unexpected changes patterns shows great potential applications.