Comparative Study of Time Series Analysis Algorithms Suitable for Short-Term Forecasting in Implementing Demand...

Area of Science:

• Energy Systems Engineering
• Data Science
• Electrical Engineering

Background:

• Accurate short-term load forecasting is crucial for effective demand response (DR) strategies in smart grids.
• Advanced Metering Infrastructure (AMI) provides real-time data essential for improving forecasting models.
• Evaluating diverse time series algorithms is necessary to optimize energy management and grid stability.

Purpose of the Study:

• To compare the performance of four time series forecasting algorithms: ARIMA, SARIMA, LSTM, and SVM.
• To assess their applicability and effectiveness for short-term load forecasting using AMI data.
• To provide guidelines for selecting the best forecasting model based on data characteristics and application requirements.

Main Methods:

• Comparative analysis of ARIMA, SARIMA, LSTM, and SVM algorithms.
• Evaluation based on predictive accuracy, computational efficiency, and scalability.
• Utilizing a dataset of real-time electricity consumption from AMI systems.

Main Results:

• Support Vector Machines (SVM) excelled in predicting nonlinear patterns and high volatility.
• Seasonal AutoRegressive Integrated Moving Average (SARIMA) effectively captured seasonal electricity consumption trends.
• Long Short-Term Memory (LSTM) demonstrated potential for complex temporal dependencies but required significant data and tuning.

Conclusions:

• The choice of forecasting model significantly impacts the efficiency of demand response strategies.
• Each algorithm possesses unique strengths and weaknesses, necessitating careful selection based on specific data and application needs.
• Integrating advanced forecasting techniques into smart grids enhances reliability and supports dynamic energy management.