Hybrid Wind-Diesel Power Load Forecasting Using Advanced Data Preprocessing and Pso-Optimized Deai-Lstm Model

Abstract

Accurate load forecasting in hybrid wind-diesel energy systems is crucial to ensuring energy reliability, cost efficiency, and environmental sustainability. In this paper, a strong prediction model is suggested based on the PSO-optimized DeAI-LSTM framework that incorporated complex preprocessing and deep learning. The paper starts with preprocessing of the data which involves the imputation of missing values, removal of outliers, normalization, and dimensionality reduction through the Principal Component Analysis (PCA). A deep autoencoder (DeAI) provides feature transformation in a time-cognisant and denoised fashion. The data transformed is fed to a Long Short-Term Memory (LSTM) model where the hyperparameters are optimized using Particle Swarm Optimization (PSO). Benchmark testing against multivariate time-series data indicates better performance in prediction of the proposed model compared to the traditional models including vanilla LSTM, GRU and ARIMA. DeAI-LSTM model has the best RMSE and MAE since it was stable and efficient when it comes to dealing with intermittent renewable inputs and predicting a diesel generation. The paper discusses the applicability of the model in microgrid and hybrid power operational planning and suggests extensions of the model which can be based on real-time deployment, edge computing, and fusion of ensemble models in the future.