Load Forecasting Method for Offshore Oil and Gas Fields Considering Multivariate Correlation

Aiming at the problem that traditional load forecasting models are difficult to effectively analyze the correlation between various characteristic variables of offshore oil and gas fields, which makes it difficult to carry out offshore oil and gas field load forecasting, a method for offshore oil and gas field load forecasting considering multivariate correlation is proposed. Firstly, considering the energy consumption characteristics in the production process of offshore oil and gas fields, extracting system level load demand feature quantities, quantifying the correlation degree of feature quantities using Pearson correlation coefficient, and selecting strong feature quantities as input for the prediction model. Secondly, based on Gated Recurrent Unit (GRU) and incorporating Dual Attention (DA) mechanism to optimize parameter values, a load forecasting model for offshore oil and gas fields based on DA-GRU algorithm is established. Finally, a simulation test was conducted using actual offshore oil and gas field data to compare the predictive performance differences of different models. The results show that using the strong feature quantities selected in this paper for load forecasting has higher accuracy, and the GRU prediction model incorporating DA mechanism has an average root mean square error 13.39% lower than traditional GRU.