Research On On-Line Fan Performance Monitoring System Based On Improved Support Vector Machine

Fan, the power source of the smoke ventilation, is one of the large-scale rotary devices and system in power plants. The running station of fan has very important influence to the safety and economy of power plant unit, so it is very necessary to monitor the fan performance and running state in real time. An accurate measurement of fan flow is the key of real time monitoring. According to the basic theory of fan, there is a stable, good nonlinear mapping relation between the inlet pressure difference and flow. The theory has proved that it is possible to monitor the flow using the mapping relationship. The relationship of running parameters can be presented by performance curve, so the modeling of performance curve is the basic of online monitoring of fan.Neural network and support vector machine (SVM) method are applied to establish the fan’s differential pressure-flow curve model which accuracy is imprecise. Parameter optimization of support vector machine is optimized by using swarm intelligence algorithm and he threshold is optimized via genetic algorithm BP neural network initial weights. Through the comparison of improved neural network model and SVM model by the experimental data, we find that the improvement of the SVM model is better than the neural network model no matter the way both square correlation coefficient or mean square error and relative error are applied.In this paper, the fan of on-line monitoring system software are established by using Visual Basic language mixed with Matlab programming under the basement of improved SVM model algorithm. Visual Basic language has excellent interface development function and convenient database operation function and Matlab has powerful numerical calculation function. Combine the advantages of two programming languages to establish fans on-line monitoring system. Through the actual fan performance curve model and the real-time data acquisition software fan differential pressure signal, we can calculate the relevant flow value. Then the points are showed on the performance curve, which can reflect the fan run-time operating conditions. Therefore, we can accurately grasp the fan operation, and the safe and economic operation of the fan system play an important role.