Effective energy price prediction using LSTM and ARIMA in the smart grid

In the Smart Grid, demand response plays a fundamental role due to the 70% energy wastage in the current power grid. This paper utilizes Long Short-Term Memory (LSTM) and Autoregressive Integrated Moving Average (ARIMA) models to predict electricity prices. The study incorporates seasonal energy variations in both demand and electricity prices. The forecasting models were implemented and evaluated using various performance metrics. The deep learning LSTM model demonstrated lower error rates compared to the traditional statistical ARIMA model. Several challenges were encountered during the research, including unexpected negative Regional Reference Prices (RRP), model limitations, and unpredictable price fluctuations. In terms of accuracy, statistical models like Simultaneous Perturbation Stochastic Approximation (SPSA) were also implemented, revealing significant prediction variations. Both statistical models and machine learning approaches were used in the prediction process. However, there is potential for improved accuracy by employing different techniques, such as alternative models, hybrid approaches, and dynamic hyperparameter tuning. This research has important implications for future work in enhancing model accuracy, demand response modeling, and Smart Grid optimization.