A Feature Extraction Method for Distributed Photovoltaic Ledger Information Based on Multi-Source Heterogeneous Data Fusion

Abstract

Solar photovoltaic power generation has become one of the most important means of securing the world’s energy supply strategy, drastically reducing emissions and ensuring sustainable development by taking advantage of the cleanliness and safety of the energy source. The development of photovoltaic (PV) power stations in pieces and the formation of PV power station clusters have become a new issue. Most traditional methods use ledger data from a single source for feature extraction, and it is difficult to achieve complete extraction of their features. In this paper, a distributed PV plant cluster ledger feature extraction method based on multi-source heterogeneous data fusion is proposed to achieve accurate prediction of PV power cluster power. Firstly, the information integration equipment is used to aggregate the power generation status information of the PV power plant cluster, including natural environment information, module parameters, battery energy storage, shadow coverage, etc. to achieve the collection of heterogeneous data from multiple sources. Secondly, the energy conversion calculation is performed based on the above information, and the unit time failure rate information is integrated to optimize the energy conversion through the failure rate calculation model. Finally, the LSTM neural network is used to process the above information, and the neural network is used to predict the PV power to optimize the operation of the PV power system. The comparison with the linear regression model proves that the model proposed in this paper has a higher prediction effect.