Smart metered distribution system load forecasting using statistical approaches

To ensure a steady supply of electricity and optimize power system management, accurate load forecasting is crucial. Using load data from the mixed residential and academic grid at NIT Patna, this study investigates the use of the ARMA, ARIMA, and SARIMA models for long-term load forecasting (LTLF). While ARIMA uses differencing techniques to represent non-stationary data, ARMA is utilized for stationary data, capturing autocorrelation and short-term trends. By taking seasonal variations into account, SARIMA, an extension of ARIMA, improves forecasting accuracy for data exhibiting cyclical patterns. The findings demonstrate that SARIMA performs better than ARMA and ARIMA in terms of forecast accuracy, particularly when it comes to identifying seasonal variations and long-term trends. This study emphasizes the significance of model selection, data preparation, and performance assessment using measures such as Mean Absolute Percentage. By the comparative analysis of the proposed statistical techniques, around 7% MAPE was obtained which is considerably good for a long-term load forecasting model.