A Dynamic Nonlinear Autoregressive Exogenous Model to Analyze the Impact of Mobility during COVID-19 Pandemic on the Electricity Consumption Prediction in Jordan
NOPR - NISCAIR Online Periodicals Repository
View Archive InfoField | Value | |
Title |
A Dynamic Nonlinear Autoregressive Exogenous Model to Analyze the Impact of Mobility during COVID-19 Pandemic on the Electricity Consumption Prediction in Jordan
|
|
Creator |
Shbool, Mohammad A
Altarazi, Farah AlAlaween, Wafa' H |
|
Subject |
ARIMA
Electricity demand LSTM NARX Recurring dynamic neural network |
|
Description |
164-173
Due to the global COVID-19 pandemic, governments have adopted regulations and restrictions to prevent spreading the disease. Changes in socioeconomic status, lifestyle, mobility and consumer consumption behavior have resulted due to these restrictions. These changes caused the amount and pattern of electricity consumption to be affected during and after the pandemic. In this study, we developed a data-driven model of electricity consumption based on machine learning techniques to analyze the effect of Mobility during and after the pandemic on electricity consumption prediction, which has been considered along with other factors that typically affect electricity consumption, including historical load profile, weather measurements, and timing information. The Nonlinear Auto Regressive Exogenous (NARX), a recurring dynamic neural network with feedback, establishes the model. The model performance results show improved prediction performance when considering the mobility factor; the error residuals between the actual and forecasted max load values were lower than when not considering the Mobility. The test dataset's least Mean Square Error (MSE) was decreased by 43%. In addition, the regression values between actual and predicted values have improved when considering the mobility factor. The same applies to the R-value and Root Mean Squared Error (RMSE), with an improvement of 6.0% and 7.6%, respectively. For comparison purposes, two additional models were developed to verify the results using the Auto-Regressive Integrated Moving Average (ARIMA) and Long-Short Term Memory (LSTM), as well known models. These models also demonstrated improved prediction performance when considering the mobility factor. However, the NARX model exhibited the best results, with lower MSE and higher R values. The models considered in this study can be used to predict the electricity consumption values of other pandemics or another wave of COVID-19 to assist decision-makers in having higher consumption visibility, thus better planning resources, capacity, and costs. |
|
Date |
2024-02-15T09:39:52Z
2024-02-15T09:39:52Z 2024-02 |
|
Type |
Article
|
|
Identifier |
0022-4456 (Print); 0975-1084 (Online)
http://nopr.niscpr.res.in/handle/123456789/63335 https://doi.org/10.56042/jsir.v83i2.84 |
|
Language |
en
|
|
Publisher |
NIScPR-CSIR, India
|
|
Source |
JSIR Vol.83(2) [February 2024]
|
|