Research On Energy Consumption Prediction And Optimization Of Combined Station Based On Machine Learning

At present,China's resource utilization rate is low,and energy consumption continues to rise.In the oilfield gathering and transportation system,the combined station is the most important production processing unit.It is urgent to save energy and reduce consumption in the combined station.The basis of realizing energy saving and consumption reduction in the combined station is to analyze the energy consumption in the combined station,optimize the combined station and achieve the purpose of energy saving and consumption reduction.Therefore,under the background of the development of the ' digital ' oilfield,it is necessary to carry out the energy consumption prediction and optimization of the combined station based on machine learning.This paper takes the production and operation monitoring data of a combined station in Daqing Oilfield as the support,preprocesses the actual monitoring data,and analyzes the energy consumption characteristics of the combined station and the correlation between energy consumption and various parameters.On this basis,machine learning is applied to the energy consumption prediction of the combined station.Four energy consumption prediction models based on a support vector machine algorithm,random forest algorithm,XGBoost algorithm,and BP neural network algorithm are established,and genetic algorithm and particle swarm optimization algorithm are used to optimize the energy consumption of the combined station.In this paper,the concept of machine learning and related algorithms are first proposed,and the monitoring data of production and operation of the combined station are cleaned.The abnormal data are identified through the box diagram.Based on the jupyter in Anaconda platform,the Python programming language is used to reconstruct the missing data by using the Lagrange interpolation method.After verification,this method can accurately and reasonably complete the repair of abnormal data monitored by the combined station,which provides a basic guarantee for the establishment of subsequent prediction models.The process flow of a combined station in Daqing Oilfield is analyzed.The comprehensive energy consumption per ton of liquid,gas consumption per ton of liquid,power consumption per ton of liquid,heat energy utilization rate,and power energy utilization rate are selected as the energy consumption indexes of the combined station to analyze the energy consumption operation characteristics of the combined station.The results show that the thermal energy utilization rate and electric energy utilization rate of the combined station are the highest in summer,and the comprehensive energy consumption per ton of liquid,gas consumption per ton of liquid,and power consumption per ton of liquid are the highest in winter.Based on Python language,the Pearson correlation coefficient method is used to analyze the correlation between the energy consumption of the combined station and basic parameters such as processing liquid volume and incoming liquid inlet temperature.On this basis,four energy consumption prediction models of support vector machine algorithm,random forest algorithm,XGBoost algorithm,and BP neural network algorithm are established based on machine learning theory,and the comprehensive energy consumption per ton of liquid,gas consumption per ton of liquid,power consumption per ton of liquid,heat utilization rate and power utilization rate of the four seasons and the whole year are predicted respectively.The results show that the predicted values of the four prediction models are highly fitted to the real values of energy consumption,and all reach the confidence level.According to different accuracy,different prediction models are used for different indicators.The comprehensive energy consumption per ton of liquid,heat energy utilization rate,and energy utilization rate is predicted based on BP neural network algorithm.The gas consumption per ton of liquid and electricity consumption per ton of liquid are predicted based on random forest algorithm.The prediction model with the highest fitting degree of each index is selected for MSE,RMSE,and MAPE analysis.The error of the four models is small.This study has a certain guiding role in the energy consumption prediction of the combined station.Finally,combined with the theoretical basis of the combined station energy consumption prediction model,the particle swarm optimization algorithm and genetic algorithm are used to optimize the combined station energy consumption respectively.Among them,the particle swarm optimization algorithm has a better optimization effect,and the optimized combined station energy consumption and operation cost are significantly reduced.The research content of this paper provides theoretical guidance for the prediction and optimization of comprehensive energy consumption per ton of liquid,gas consumption per ton of liquid,and power consumption per ton of liquid in the combined station provides technical support for energy saving and green operation of the combined station and further provides the theoretical basis for promoting the rapid development of 'digital' oilfield.