Short-Term Load Forecasting Using Convolutional Neural Network for Chhattisgarh State

Load forecasting is an important energy management and operation process that can predict future energy needs. In other words, it is the process of predicting how much electricity will be used at a certain time and how the demand will affect the grid. Various factors, including historical consumption, weather conditions, economic growth, etc., influence load forecasting. The study evaluates the performance of the convolutional neural networks (CNN) for Short-Term Load Forecasting (STLF) in Chhattisgarh State. Convolutional neural networks can extract characteristics from input data in predictive activities. In the experiment, we use one-dimensional CNN (1D-CNN) for STLF, which is trained and tested on an input dataset that includes load demand data from the past 7 days as historical data, with the present-day load demand as the target data. The data values for the input dataset vary from 25 Dec 2021 to 30 June 2024 for both the training and testing process. Lastly, to analyze the overall performance of CNNs on selected datasets for Chhattisgarh State, the evaluation matrix is determined and calculated as Mean Absolute Percentage Error (MAPE) and Mean Absolute Error (MAE). It is usually used to measure the difference between actual and estimated data. A comparative analysis of CNN’s different structures is conducted to find an accurate model for STLF.