Daily Power Plant Operation Prediction Using Adaptive Filter Based on Wavelet Symlet

Abstract

Purpose – This study aims to develop an accurate method for predicting the daily operation of power plants to support optimal scheduling of generation and maintenance activities.

Methodology – An adaptive filter based on wavelet symlet (adaplet) is applied using the Normalized Least Mean Square (NMLS) algorithm. The model adjusts its coefficients dunamically based on historical operational data to minimize prediction error.

Findings – The method was tested on Indonesian power plant operation data and achieved a Mean Square Error (MSE) of 0.079. Segment-based evaluation confirmed the model’s ability to provide consistent prediction accuracy across different time frames.

Originality – This research introduces a novel approach by combining wavelet symlet and adaptive filtering in the context of power plant operation prediction, which allows accurate forecasting using limited data.

Research limitations – The study focuses on short-term prediction (up to 3 days ahead) and does not include external influencing factors such as weather or system demand. Only the NLMS algorithm was utilized, without comparison to other adaptive methods.

Practical implications – The proposed method enables operators to generate more accurate and reliable schedules, improving overall system performance and reducing outage risks.

Social implications – Enhancing the reliability of power plant operations contributes to a more stable electiricty supply, indirectly supporting public services and economic activities.