The improved integrated Exponential Smoothing based CNN-LSTM algorithm to forecast the day ahead electricity price

Area of Science:

• Energy Economics
• Artificial Intelligence
• Time Series Analysis

Background:

• Deregulation of electricity markets has spurred the growth of short-term trading, necessitating accurate day-ahead price forecasting for effective bidding by generators and consumers.
• Electricity prices fluctuate significantly due to dynamic consumer bidding patterns, highlighting the need for robust predictive models.

Purpose of the Study:

• To propose and evaluate a modified Exponential Smoothing-CNN-LSTM (Convolutional Neural Network-Long Short-Term Memory) method for enhanced day-ahead electricity price forecasting.
• To assess the performance of the proposed forecasting model using real-world data from the Indian Energy Exchange (IEX).

Main Methods:

• The study integrates Exponential Smoothing for capturing trend and seasonality with CNN-LSTM for modeling complex spatial and temporal dependencies in time series data.
• A hybrid Exponential Smoothing-CNN-LSTM model was developed and tested on day-ahead electricity market data.

Main Results:

• The proposed Exponential Smoothing-CNN-LSTM model achieved superior forecasting performance, evidenced by a Mean Absolute Error (MAE) of 0.11, Root Mean Squared Error (RMSE) of 0.17, and Mean Absolute Percentage Error (MAPE) of 1.53%.
• The hybrid model demonstrated improved accuracy compared to individual Exponential Smoothing, LSTM, and CNN-LSTM techniques.

Conclusions:

• The developed Exponential Smoothing-CNN-LSTM method offers a significant advancement in day-ahead electricity price forecasting for short-term market participants.
• The model's effectiveness suggests its potential applicability to time series forecasting challenges in diverse sectors such as finance, retail, healthcare, and manufacturing.