Pumping Unit Energy-Saving Algorithm Based On Neural Network

In China, most oil fields have entered the middle or late period of exploitation at present.Pumping units are in light or idle run, which leads to pumping unit seriously worn, and at thesame time a large number of electrical energy has been consumed due to most of wells withdeficient-liquid supply. In addition, power consumption covered a large part of the proportionin exploitation costs. Therefore, a pump saving algorithm is needed to control the pumpingunit start-stop time dynamically according to the changes of the amount of oil for reducingthe cost and improving the efficiencies of the oil field.Based on the current oilfield production efficiency and the research of oilfield in Chinaand overseas, firstly, the dissertation studies the principle of pumping system and existingproblems of several energy saving technology to determine the energy saving method.Secondly, for the determination of the network model adopted in this algorithm, the neuralnetwork has been studied. Then based on neural network, a pump energy saving algorithm isproposed. Such algorithm aims at the problem that pumping unit start-stop time can notmatch the quantity variance of oil. By combining genetic algorithm with fuzzy neuralnetwork, the algorithm selects current related parameters as the sample to train the networkand obtains forecast state of pumping unit to ensure the reasonable time of interval pumping.In experiment study, the algorithm is simulated and energy efficiency is analyzed. Theresults show that the proposed algorithm can not only decrease power consumption of oilwells effectively during the late stage but also raise the oilfield benefits economically.