A hybrid intelligent algorithm based short-term load forecasting approach

In this paper, a new two-step algorithm is proposed for short-term load forecasting (STLF). In the first step of the method, a wavelet transform (WT) and an artificial neural network (ANN) are used for the primary forecasting of the load over the next 24 h. Inputs of this step are weather features (include the daily mean temperature, maximum temperature, mean humidity, and mean wind speed) and previous day load data. In the second step, a WT, the similar-hour method and adaptive neural fuzzy inference system (ANFIS) are used to improve the results of primary load forecasting. In this study, a WT is employed to extract low-order components of the load and weather data. Furthermore, the number of weather data inputs has been reduced by investigating the weather conditions of different cities. To evaluate the performance of the proposed method, it is applied to forecast Iran’s load and New South Wales of Australian’s load. Simulation results in four different cases show that the proposed method increases load forecasting accuracy.

► A new hybrid intelligent algorithm is proposed for STLF.
► Proposed forecasting algorithm consists of two steps to increase forecasting accuracy.
► Wavelet Transform and an ANN are used for the primary forecasting step.
► Wavelet transform, similar-hour method and ANFIS are used in the second step.
► Proposed method improves forecasting accuracy of Iran and Australian.