Optimization of Load Forecasting in Power Systems Using Hybrid AI Models

Abstract

Load forecasting is useful in efficient running and planning of power systems, predominantly in situations involving more and more renewable energy sources. The previous forecasting techniques tends to overlook the complex non-linear trends in load data. This paper proposed a hybrid AI solution, which combines Long Short-Term Memory networks, Random Forest and ARIMA to improve the precision of load forecasting. The two models together enhance the ability to foretell results and be insusceptible to changes. Various experiments are carried out by loading data of the past and the suggested model is tested against other common methods of forecasting. It can be seen that there are significantly improved forecasts because of lower MAPE and RMSE in the results. What is more, the hybrid model is more efficient and able to support real-time requirements. The work establishes a decent avenue of future developments in load forecasting that helps in the reliability of operating power systems.