Comparative Analysis of different Machine learning Models for Load Forecasting

Load Forecasting helps the utility to make important decision such as load scheduling, load shedding, etc. The main objective of load forecasting are control, operation and planning of power system. With increasing complexity of power system, the proper choice of machine learning techniques also becomes challenging. This paper presents a comparative analysis of nineteen machine learning models such as Linear Regression, Bagged Tree, Cubic Support Vector Machine, Gaussian Process Regression with four different kernel function e.g. Squared-exponential, Rational Quadratic, Exponential, Mattern 3/2, Fine tree, Coarse tree, Quadratic support vector machine, Interaction regression, Medium tree, Robust linear regression, Stepwise linear regression, Linear support vector machine, Fine Gaussian support vector machine, Coarse Gaussian support vector machine, Medium Gaussian support vector machine, and Boosted tree. For short term load forecasting, a dataset of July 2022 of Phata region of Maharashtra, India is considered. The simulation result shows that Exponential Gaussian Process Regression gives the best prediction of load compared to other models. The validation results indicate that it has the lowest RMSE (Root Mean Square Error), MSE (Mean Square Error) MAE(Mean Absolute Error) and their values are 1.2, 1.44 and 0.77 respectively.

Automated summary

Load forecasting is important for utility companies to make decisions regarding load scheduling and load shedding. This paper presents a comparative analysis of 19 machine learning models and finds that Exponential Gaussian Process Regression performs the best in load prediction.