Study Of Fault Diagnosis For Power Installation In Heating Boiler House Based On Neural Network, SVM And Their Improved Algorithms

In the north of China, central heating has become the main heating way in winter. Boiler room as the heating source, its economic and effective, safe and reliable heating is very important to thousands of people’s daily work and life.Ventilator and water pump is the main energy consumption equipment of heating boiler room, having been in a state of high automation, continuous operation, the state monitoring and inquiring, fault diagnosis and processing being very important. However, the current monitoring and diagnosis is still on the basis of manual operation and experience, so in this paper, based on the urban heating GIS system platform, fault diagnosis for power installation in heating boiler house based on neural network, SVM and their improved algorithm is developed,which can manage equipment in real time, convenient to diagnose fault and maintenance decision-making.By analyzing a heating companies heating period boiler breakdown maintenance data, heating boiler room has fault diversity, fault frequency imbalance and destructive characteristics of vibration. The ventilator and water pump have 11 kinds of common failures.Combined with the characteristics of the ventilator and water pump fault in heating boiler room, heating boiler room power plant intelligent fault diagnosis system is put forward. Arc GIS is chosen as the system development of GIS platform, using the Geodatabase data model to manage the spatial data of the heating network and equipment information. Using advantages of Microsoft Visual c + + 2010 and MATLAB, the fault monitoring and diagnosis system is developed.Aiming at the ventilator and water pump faults with uncertainty and complexity characteristics in heating boiler house,combining the adaptive and learning capabilities of neural network and the language description to get knowledge of fuzzy system, a two stage fault diagnosis model of ventilator based on T-S fuzzy neural network is proposed,the first model for the adaptive fuzzy neural network ANFIS model, the secondary model for multiple input multiple output fuzzy neural network model. which can diagnose the types of faults and identify the causes of faults, according to the changes in the characteristics spectrum values of the vibration signal of common faults.With the simulation tests by MATLAB software,through the example comparison of the fuzzy neural network and the BP neural network,the results illustrate thatthe fault diagnosis method of the fuzzy neural network can recognize the faults rapidly, accurately and steadily, which provides a efficient way for the diagnosis.Based on the basic theory of support vector machine(SVM) classification problem, Linear classification, the approximate linear classification and nonlinear classification algorithm being derived in detail, and the choice of concept of kernel function, And analyzing the effect of punishing parameters c and RBF kernel function g on nonlinear model of support vector classification,heating boiler room ventilator fault diagnosis models are established based on grid search algorithm, genetic algorithm, particle swarm optimization algorithm and parameters without optimization, through the comparison of the result of the fault diagnosis, it is concluded that the genetic algorithm and particle swarm algorithm accuracy is relatively high, which is more suitable for occasions in the present study on the algorithm of search optimization.The genetic algorithm to optimize the SVM based on binary tree and particle swarm algorithm to optimize the SVM based on binary tree of heating boiler room ventilator fault diagnosis model are established. Through the comparison of fault diagnosis, it is concluded that the two having his strong points, and according to the specific equipment heating failure, deciding the optimization method. Binary tree SVM parameter optimization of the ventilator fault diagnosis is compared with parameter optimization SVM, binary tree SVM parameter optimization of fault diagnosis is more suitable for occasions in thepresent study on the structure and algorithm of search optimization, which is worth to be popularized and applied.Binary tree SVM parameter optimization of the ventilator fault diagnosis is compared with the fuzzy neural network fault diagnosis, binary tree SVM parameter optimization of fault diagnosis has the advanced nature in study of high dimension and small sample, and applicability on the algorithm.