Short-Term Tariff Prediction for High Penetration New Energy Electricity Market Based on Singular Spectrum Analysis and CNN-GRU Combination Model

Accurate electricity price forecasts are essential for electricity market participants to make informed choices. The complexity of forecasting electricity prices continues to rise in new power systems that consist mainly of new energy generation. In order to improve the accuracy of electricity price prediction in power markets with a high proportion of new energy sources, this paper proposes an electricity price prediction technique based on singular spectrum analysis after combing the existing electricity price prediction techniques. In this paper, a short-term electricity market price prediction model combining Singular Spectrum Analysis (SSA), Convolutional Neural Network (CNN), and Gated Recurrent Unit (GRU) is constructed. The method takes electricity price, load and new energy generation as input, decomposes and reconstructs the original data by using SSA, and constructs CNN architecture consisting of convolution layer and pooling layer for trend component and oscillation component of each array, respectively, to extract high-dimensional features reflecting complex dynamic changes of electricity price. The feature vectors proposed are constructed in the form of time series, which are used as inputs to the GRU network, and the internal dynamics of the features are modelled and learnt as a rule. Eventually, the final electricity price prediction results were obtained by accumulating the prediction sequences. The experimental results confirm that this method can significantly improve the accuracy of short-term electricity price prediction.