Photovoltaic generation power prediction research based on high quality context ontology and gated recurrent neural network

Abstract

This paper proposes a new method to improve the prediction accuracy of photovoltaic power generation. By improving the accuracy of photovoltaic generation prediction, the grid can reduce the restriction on photovoltaic power and thus improve the return on investment of the photovoltaic industry. This paper innovatively obtains high-quality contextual information through high-quality ontology and improves the accuracy of final prediction from the perspective of improving the quality of input data. By building a high quality context ontology model, the context is classified according to its different sources. Then the quality of the classified context is scored. Finally, the high quality context is selected to replace the low quality context. Simulation results show that this method can represent the context quality more flexibly while increasing the ontology in a small scale. Besides, this paper also used the gated recurrent neural network as the prediction model. Experimental results show that the prediction accuracy of photovoltaic power generation based on high quality context ontology and Gated Recursive Neural Network is about 5% higher than that of Long Short Term Memory model. When the number of hidden layers of the prediction network is set to 4 and the number of iterations is set to 100, the accuracy is the highest and the mean square error is 0.0037. In conclusion, this method can effectively improve the prediction accuracy and has a high application prospect in the industry.