Integrating Swarm Intelligence and Deep Learning for Enhanced Accuracy in Electrical Load Prediction

Abstract

In this modern era, energy became an essential need for human’s daily life. This fluctuation can be solved only with accurate load forecasting. In this paper feature engineering has divided into feature extraction and selection parts, this part has been used to select the most important features and reject irrelevant features. To perform the accurate load forecasting a combination of feature engineering, data-mining, and machine learning techniques has been used. In feature engineering combination of Decision Tree, XGBoost and Relief-F techniques have been proposed. Most relevant features forward to machine learning techniques for classification. To improve the accuracy, rate a combination of Particle Swarm Optimization (PSO) and Convolutional Neural Network (CNN) has been implemented for classification. Forecasting also has been performed with the proposed technique and some conventional machine learning techniques such as; Convolutional Neural Network (CNN), Support Vector Machine (SVM), Random Forest (RF), Linear Discriminant Analysis (LDA) and Linear Regression (LR). Comparative analysis has been done with the help of Mean Square Error, Root Mean Square Error, Root Mean Absolute Error, F1_Score, and Recall values. The comparison shows that the proposed technique has fewer error rates and higher accuracy rates.