Renewable Energy Generation Prediction using Structure Embedded based Gated Recurrent Unit

The renewable energy provides several environmental and economic benefits that compared to other energy resources such as nuclear and fuel-based energy. The data acquired and utilized for predicting renewable energy generation that includes strong-volatility, significant randomness, and intermittent behavior. However, the variation in data obtained from various energy generation resources lead to significant challenge and results in inaccurate prediction results. To overcome this problem, a Structure Embedded based Gated Recurrent Unit (SE-GRU) method is proposed for predicting the renewable energy generation and consumption accurately. The proposed SE-GRU model solves the issue of variation in data due to factor like climate conditions by capturing long term dependencies and immediate fluctuations that enhanced the prediction results. Initially, the data obtained from renewable resources are preprocessed by using techniques like normalization and filling missing values. These preprocessed data are forwarded to proposed prediction model SE-GRU to predict the energy generation and consumption accurately. The experimental results of proposed SE-GRU method achieved Mean Squared Error (MSE) of 0.19 which is less than existing prediction approaches such as Echo State-Network-Convolutional Neural Network (ESNCNN).