Hybridising neurofuzzy model the seasonal autoregressive models for electricity price forecasting on Germany's spot market

Electricity price forecasting has become an area of increasing relevance in recent years. Despite the growing interest in predictive algorithms, the challenges are difficult to overcome given the restricted access to relevant data series and the lack of accurate metrics. Multiple models have been developed and proven to work in the area of EPF. This paper proposes a new univariate hybrid model, trained, and tested on German electricity market data, based on the Seasonal Auto-Regressive Integrated Moving Average (SARIMA) and the NeuroFuzzy-Local Linear Wavelet Neural Network (LLWNN). Although a series of complex challenges create difficulties in refining the model, the proposed algorithm significantly narrows the gap between predictions and actual prices. The ability to predict the dynamics of the price of electricity on the spot market is an important asset for both suppliers and consumers, with a view on prophylactic calibration of supply-demand ratios. The model can be extended and applied to any energy market with a stable structure.