Fault Diagnosis Of Heating Networks Based On BP Neural Network Optimized By Gentic Alforithm

With the constant improvement of the municipal infrastructure construction,urban central heating is received more attention increasingly. As a part of the lifeline of city, urban central heating is a important aspect of eusuring urban fuction worked properly. Under the influence of many factors, such as materials of heating pipe, environment and operation time, heating network will inevitabile resulting some faults and bring some loss of urban and production, so finding the problems and locking fault location can improve the cost effectiveness and safety of heating network.The most common fault of heating network is leakage fault. heating networks leakage fault has uncertainty characteristics, such as randomness, fuzziness and variability.Most of domestic heating companies generally use artificial leak detection and the acoustic leak detection method, which are susceptible to human factors and the impact of outside intereference and have a misjudgment rate, so these traditional methods can’t meet the requirements of heating network modern security monitoring. For now, the diagnostic technique based on mathematical model and sensor are the mature fault diagnosis methods of heating network at home and abroad,but both of them exist some limitations. The leakage fault diagnosis based on mathematical model needs to establish accurate mathematical model of heating networks, and general actual heating network is difficult to meet the requirements. The leakage fault diagnosis based on sensor directly processes testing signal,so it don’t need to build mathematical model. Because of the general water temperature of urban heating network is above90, so the ordinary sensor is difficult to meet the requirements of the actual heating network system.With the rapid development of artificial intelligence, the emphasis of heating network leakage diagnosis research gradually shifted to the automatic and intelligent fault leakage diagnosis technology, especially BP neural network which has the ability of self-learning and highly nonlinear fit, has become a hot research topic. BP neural network fault diagnosis of heating network obtains diagnosis model by a series of learning, and the learning samples of BP neural network is leakage pressure changes in different working condition. Because BP neural network uses gradient descent learning method, it has Salivary gland pitfalls of slow convergence speed and falling into local minima.In order to make up for the defects, this paper presents a fault diagnosis model of heating networks based on two-stage BP neural network optimized by genetic algorithm, which has a powerful search optimization ability. The model includes diagnosis model of leakage pipe and diagnosis model of leak locator. The model not only can quickly and accurately diagnose the leakage of concrete pipe, can also locate the leakage point.At the end of this paper, the above model is applied to an actual heat ing network by MATLAB, The results show that the model has the better function than the fault diagnosis model of the traditional BP Neural Network, and with high accuracy,So this model can be effectively applied to the heating pipe network fault intelligent diagnosis system.