Optimal Solution for Load and Price Forecasting in Smart Grids

Abstract

With the advancement of technology, people need to know the things in advance so that they can make appropriate decisions for it. Similarly, when it comes to energy, we are curious to know that how much energy we are going to use in the next hour and how much it will cost them. There are many accurate price and load forecasting algorithms already working but they ignore the convergence rate. Data Science helps in evaluating the huge amount of data to predict future values. Executing any algorithm on such large amount of data needs high computational time. In the case of predicting the next hour or one-day value if the algorithm takes too much time in results formulation, then it becomes useless for us. We incorporated deep learning techniques as they process a large amount of data very fast and can predict fairly accurate results with a fast convergence rate. The proposed solution LSTM-BiGRU is formed in combination of LSTM and GRU, both are RNN variations and capable of forecasting best results. LSTM and GRU are combined in a best possible way to achieve the maximum accuracy with a fair convergence rate. The proposed solution is showing MAPE in load forecasting from 3.12% to 4.07%. The proposed solution is showing MAE in January 2019 for price forecasting from 2.35 to 3.02 and the execution of proposed solution in every scenario is recorded <1 min. So, a fair tradeoff is maintained between forecasting results and computational time. In future, the proposed method can be improved by other techniques i.e., block chains, optimization of proposed hybrid algorithms with evolutionary algorithms, and the use of GPUs and TPU can further decrease the computational time.