Research on ultra-short-term load forecasting method of oil and gas field integrated energy system based on hybrid neural network

Abstract

Abstract Oil and gas fields have a large amount of distributed new energy. In order to improve the utilization rate of new energy and respond to the dispatching needs of China's State Grid, it is necessary to study the use of ultra-short-term load forecasting algorithms to improve the load forecasting accuracy of oil and gas fields and support the coordinated interaction of source, grid and load in the integrated energy system of oil and gas fields. This paper proposes an ultra-short-term load forecasting algorithm based on a hybrid neural network called Convolutional-Bidirectional Long Short Term Memory-Skip (CNN-BiLSTM-Skip). Using the operating load data of an oil and gas field in Northeast China as a data set, we first constructed a cooling, heating and power system architecture model with wind turbines, photovoltaics, power grids and natural gas as “source and grid loads”; Secondly, we used an improved hybrid multi-time scale algorithm and unit A prediction model was constructed based on the operating load data, and the prediction results of the nonlinear part and linear part of the model were output and integrated to obtain the final prediction result; Finally, the prediction error evaluation index of the algorithm proposed in this article was compared with algorithms such as BP, LSTM, and CNN-LSTM. The results show that the algorithm proposed in this article has stronger robustness and higher accuracy. The proposed CNN-BiLSTM-SKIP algorithm improves the prediction accuracy. Compared with the BP neural network algorithm, the MAPE evaluation index has an average accuracy increase of 3.78%, compared with the LSTM prediction algorithm, the accuracy has increased by 1.63% on average, and compared with the CNN-LSTM prediction algorithm, the accuracy has increased by 0.74% on average; and the proposed prediction algorithm is compared with the BP neural network algorithm, LSTM prediction algorithm and CNN-LSTM algorithm, the RMSE and MAE evaluation index values are both the smallest, which can support the collaborative interaction of oil and gas field source, network and load and realize the planning and dispatching needs.