A Comprehensive Synergistic Framework Integrating Autocorrelation Analysis with Advanced Seasonal-Trend Decomposition Using LOESS (STL) for Robust and Accurate Short-Term Peak Electricity Demand Forecasting in Tamil Nadu’s

Abstract

rapidly developing regions like Tamil Nadu the reliable and accurate short-term forecasting of peak electricity demand is very important to improve the Operational efficiency, improving the sustainability of power system and also increases the economic management. This work proposes an inclusive synergistic framework which combines effectively the Autocorrelation analysis and seasonal-trend decomposition using LOESS (STL) to enhance the accuracy of Future peak demand prediction in Tamil Nadu’s dynamic electricity grid. This Autocorrelation which systematically captures the intrinsic temporal dependencies, periodic consumption patterns and effects persistence’s embedded with the high-resolution data captured by Tamil Nadu government on hourly basis. By Parallel, the decomposition method like STL decomposition method which decomposes the complex peak demand signals into interpretable seasonal, trend and residual components, thus it leads to more precise and targeted modeling for each constituent. This hybrid approach as combination of autocorrelation and decomposition which overcomes other issues like non-stationarity, seasonality and irregular fluctuations in the time series data. This study also comprises of combination strategies like classical ie., traditional statistical insights method along with the flexibility of modern data-driven data’s in the historical dataset which spanning a period of years provided by the SLDC, Tamil Nadu which also consists of enriched meteorological data along with the calendar variables. The Experimental results are framed to enhance the connection between hybrid framework’s supremacy over the statistical and Machine learning models such as isolation, which includes the significant reductions in the Prediction of errors and generalization of non-availability of data ie., generalization to unseen the periods. Which also includes the exogenous variables like temperature fluctuations leads to impact on holiday effects, on improving the forecast accuracy by improving the contextual relevance of the model. This framework also useful for real time grid management, load balancing and demand responses the application to increase the efficiency on energy enhancement and cost of the power sector in Tamil Nadu. This leads to adopt the conventional time series in to modern machine learning method step towards the success and offering a replicable blueprint for similarly complex energy systems worldwide.