Electricity Consumption Prediction Using a Multi-Model Machine Learning Framework

Abstract

Electricity demand prediction is crucial for efficient energy management in modern societies. In this research paper, we explore the efficacy of machine learning models in forecasting electricity consumption. Leveraging established ML models such as Long-Short term memory (LSTM), Random Forest, Support Vector Regression (SVR), we initially conduct predictions. However, to enhance predictive accuracy and address specific challenges inherent in prediction, we proposed modification in these traditional models. Our approach involves the development and utilization of novel ML architectures tailored to the electricity prediction task. We introduce three enhanced models: Deep LSTM, Random Forest combined with Neural Networks, and SVR integrated with Fuzzy Systems. These models are designed to capture intricate patterns, non-linear relationships, and uncertainties present in electricity consumption data. Our goal is to determine the best model for predicting power usage through empirical analysis and comparative review. Additionally, we provide insightful information on the advantages and disadvantages of each approach concerning energy management strategies and sustainable resource allocation.