Hybrid Machine Learning Model for Electricity Consumption Prediction Using Random Forest and Artificial Neural Networks

Abstract

Predicting electricity consumption is notably essential to provide a better management decision and company strategy. This study presents a hybrid machine learning model by integrating dimensionality reduction and feature selection algorithms with a backpropagation neural network (BPNN) to predict electricity consumption in Thailand. The predictive models are developed and tested using an actual dataset with related predictor variables from public sources. An open geospatial data gathered from a real service as well as geographical, climatic, industrial, household information are used to train, evaluate, and validate these models. Machine learning methods such as principal component analysis (PCA), stepwise regression (SWR), and random forest (RF) are used to determine the significant predictor variables. The predictive models are constructed using the BPNN with all available variables as baseline for comparison and selected variables from dimensionality reduction and feature selection methods. Along with creating a predictive model, the most related predictors of energy consumption are also selected. From the comparison, the hybrid model of RF with BPNN consistently outperforms the other models. Thus, the proposed hybrid machine learning model presented from this study can predict electricity consumption for planning and managing the energy demand.

Comparison of the Accuracy of Predicting Electricity Revenue by Predicting Electricity Quantity and Predicting Electricity Revenue by Historical Received Electricity Fees

“…Based on the mining and analysis of regional data over the years, the following important conclusions have been drawn: the sales of electricity in each quarter over the years have both a gradual growth trend and seasonal periodic changes in time series; Except for January and February, the quarterly proportion ζ of all other calendar years is very stable, with slight fluctuations around a stable value. When establishing mathematical models, factors affecting the Spring Festival should be considered in January and February [6].In order to fully consider the characteristics of seasonal cyclical changes and gradual growth of electricity sales, and to reduce errors caused by random fluctuations in a single month due to short time, the Winston method is used to predict the quarterly electricity sales of a certain month.The Winston method is a seasonal prediction method that combines factor analysis of time series with linear trends, seasonal changes, and regular changes, and exponential smoothing method. This method has three smoothing equations, which perform exponential smoothing on long-term trend S , trend increment b , and seasonal variation F .…”

Predicting Residential Electricity Consumption Using CNN-BiLSTM-SA Neural Networks

“…Additionally, Saxena [33] proposed a hybrid model based on ARIMA, logistic regression, and artificial neural networks to forecast peak load days for a billing period, achieving significant savings for an American university during a oneyear testing period. Similarly, Kesornsit [34] presented a hybrid model integrating dimensionality reduction and feature selection algorithms with a backpropagation neural network to predict electricity consumption in Thailand. This model consistently outperformed others in their study.…”

Forecasting Water Demand for Istanbul by Applying Different Machine Learning Algorithms

“…Karamaziotis et al (2020) conducted an estimation study using autoregressive integrated moving average and multilayer perceptron methods to better plan daily water consumption. Kesornsit and Sirisathitkul (2022) estimated water consumption using the multilayer perceptron, multiple linear regression, and support vector regression methods using the data between 2006-2015. Kühnert et al (2021) applied a deep learning approach, a long short-term memory method, to predict water demand and provide optimal pump control.…”