Research On Stroke Optimization Frequency Of Submersible Reciprocating Pump Based On Dynamic Liquid Level Forecasting

The submersible reciprocating pump belongs to a rodless pump oil production equipment,which has the advantages of good control-ability,stop-start at any time and high energy utilization rate.As a new type of pumping unit,the submersible reciprocating pump has much more room for improvement in stroke control and fault diagnosis.It is of great importance to improve the control method of submersible reciprocating pump to give full play to the advantages of this new technology.This improves the oil recovery rate and makes it widely used in middle and late stage wells.Therefore,based on the analysis of the composition and working principle of the submersible reciprocating pumping unit,this paper proposes an optimization method for the stroke optimization frequency of the submersible reciprocating pump which based on the dynamic liquid level forecasting.The core driving equipment of the submersible reciprocating pumping unit is a submersible linear motor.The ground control cabinet can adjust the submersible motor's stroke at any time according to the height of the underground moving oil surface to achieve the purpose of optimal control,and through dynamic planning to make crude oil output or the economy is optimal,and the real-time and accuracy of the measurement and prediction of the dynamic fluid level becomes the core element of optimization.In this paper,it takes a lot of manpower and material resources for manual measurement,and it is easy to produce errors.The soft measurement method itself has high complexity and is prone to over-fitting.It is proposed to establish a dynamic prediction model of particle swarm optimization extreme learning machine.The historical dynamic liquid level information predicts the dynamic liquid level for a period of time in the future.This model can avoid the occurrence of over-fitting and the generalization ability is strong,which lays a foundation for the research of the optimization.Based on the results of dynamic liquid level prediction,the dynamic programming stroke frequency method and fuzzy matching stroke frequency method are put forward respectively.The dynamic programming stroke frequency method looks for the optimal solution of the objective function according to the dynamic liquid surface prediction results in the total interval matching,and the fuzzy matching stroke frequency method matches the punching time according to the function relation between the dynamic liquid surface and the stroke frequency and the artificial experience.Through the experimental comparison and analysis,the results obtained by the dynamic programming stroke frequency method are better,but the calculation amount is large,The results obtained by the fuzzy matching stroke frequency method are slightly lower than the dynamic programming stroke frequency method,but the calculation amount is small,which is convenient for practical application.Finally,a downhole data acquisition device based on power line carrier is designed to solve the problem that a large amount of data is needed to construct the time series of moving liquid level.It can be powered by submersible cable and transmit downhole acquisition information to the ground control cabinet in real time.The control cabinet can be connected and interacted with remote group-well control terminal through 4G network to support the realization of the optimal control method of impulse.