Medium Term Load Forecasting for an Industrial Factory Using Bi-LSTM

This paper introduces an improved method of medium-term load forecasting in industrial factories, utilizing Bidirectional Long Short-Term Memory (Bi-LSTM) networks. While LSTM models are traditionally employed for time series prediction, this study highlights the superior accuracy of Bi-LSTM, which processes data sequences in both directions, capturing complex temporal relationships more effectively. Our methodology is validated through an empirical analysis based on actual load data from an industrial factory, showcasing significant improvements over conventional LSTM models. This breakthrough has substantial implications for energy management within industrial sectors, enabling more precise forecasting and, consequently, more efficient resource utilization and cost reduction. Although the use of Bi-LSTM is established in various domains, our study delineates specific adaptations and optimizations like hyperparameters tuning that amplify its efficacy in load forecasting. Moreover, results from proposed Bi-LSTM network are compared with LSTM network. The results show that Bi-LSTM predicts load with higher accuracy as compared to LSTM.