Load forecasting based on grasshopper optimization and a multilayer feed-forward neural network using regressive approach

This paper introduces a proposed model for mid-term to short-term load forecasting (MTLF; STLF) that can be used to forecast loads at different hours and on different days of each month. The combined MT-STLF model was investigated to aid in power generation and electricity purchase planning. A hybrid model of a multilayer feed-forward neural network (MFFNN) and the grasshopper optimization algorithm (GOA) was introduced to obtain high-accuracy results for load forecasting using the combined MT-STLF model. The MFFNN is prepared by processing the input layer and output layer and finally selecting a suitable number of hidden layers. The main steps in developing the model from the MFFNN include entering the data into the network, training the model and finally implementing the prediction process. The accuracy of the model obtained before using the GOA was lower than that after applying the GOA. Weather factors such as the temperature were used as inputs to the MFFNN during MT-STLF modelling to ensure high accuracy. In the proposed model, the temperature had a clear effect on the forecasted load. Additionally, there was a difference between the maximum and minimum loads in winter and summer months. A regressive model was introduced to determine the relations between the dependent variable (the load) and the independent variables that affect the load, such as the temperature. The regressive model used in the paper highlights the effect of the temperature on the hourly load. The accuracy of the hybrid model is satisfied with deviation error varied between -0.06 and 0.06. Moreover, the performance of the proposed forecasting model has been assessed by three indices; Root Mean Square Error (RMSE), Mean Absolute Error (MAE) and Mean Absolute Percentage Error (MAPE) then, compared with other forecasting models considering other optimization algorithms. (C) 2020 Elsevier Ltd. All rights reserved.