Research On The Speed Estimation For Submersible Motor Based On ELMAN Nerual Network

As a large oil equipment, the submersible motor is becoming one of the important facilities in the oil field at home and abroad during middle later stage. To ensure its reliable, efficient and stable operation, the real-time monitoring of parameters such as motor speed, motor temperature is needed. Because of its own structure and poor working conditions, it's difficult to monitor the motor speed by using traditional speed generator or photoelectric digital pulse encoder. Therefore, a identification of speed sensorless estimation program is proposed.In this paper, the research object is submersible motor provided by DaQing powerlift pump Ltd. We analyze the importance and necessity of proposing the speed sensorless thought, comprehensively introduce the research achievements of speed sensorless at home and abroad. Based on a comparison of strengths and weaknesses of the methods, such as Kalman Filter, Model Reference Adaptive System, and Rotor Shaft Harmonic Etc, we proposed sensorless speed estimation program based on Neural Network.Firstly, according to the induction motor's state equation in the two-phase stationary coordinate, we obtain the nonlinear dynamic relation between motor speed, stator voltage, and stator current. secondly, the theory of Neural Network is introduced, and the ELMAN Neural Network which also have nonlinear dynamic performance is chosen. After training, the multilayer ELMAN Neural Network can approach any nonlinear dynamic process with any accuracy. Thirdly, a dynamic model of three-layer ELMAN Neural Network is established, among which the input is stator current and the output is speed. From the simulation results using the MATLAB/SIMULINK, it can be seen that the estimation program based on ELMAN Neural Network not only has high accuracy at steady state, but also has effective tracking result at dynamic state. Lastly, using IPC as experimental research platform, the dynamic estimation of motor speed based on Neural Network is implemented, which uses VB language to write the whole algorithm and the user interface for speed dynamic showing.As the motor's extremely poor running condition, the stator current has high-frequency noisy, and the stator current inherent has harmonic, to improve the estimation accuracy of Neural Network, using wavelet analysis to filter the sample stator current, only leaving stator current fundamental component. Experiments show that this method can improve the speed estimation's stability and accuracy based on Neural Network.Taking into account the submersible motor iron loss, the copper loss and the stray, they will destroy motor insulation, accelerate motor aging. For the safe and reliable operation of motor, the temperature monitoring is needed. In this paper, the DC inject method is proposed to calculate the stator resistance, according to the linear relationship between resistance and temperature in large area, we can estimate the temperature of submersible motor. The result of simulation and experiment show that this algorithm has high accuracy.