Research On Fault Diagnosis Of Heat Exchange Station Based On Improved GMM Of PSO-EM

With the popularization of central heating,the heat exchange station plays a very important role in the heat supply.In case of fault,the economy and safety of heating system will be affected.The most common fault in the heat exchange station is the leakage of pipe network.At present,a large number of sensors and valves are usually added to identify and diagnose the fault between pipe sections.However,due to the limited use of sensors in the actual heating field,and the sensor itself is also inevitable to fail in the harsh environment.Therefore,based on this research background,a fault diagnosis method based on detection algorithm and fault model fusion is established for leakage fault and sensor fault of heat exchange station.This paper studies as follows:First of all,in view of the problem of how to judge whether there is abnormality in the heat supply data of the heat exchange station,the Gaussian mixture model(GMM)algorithm is used as the evaluation method of fault abnormality detection to cluster the data of the heat exchange station and identify the outlier in the data as the abnormal data of the heat exchange station.Because the expectation maximization(EM)algorithm in the Gaussian mixture model has the disadvantage of falling into the local optimum,the particle swarm optimization algorithm is used to improve the EM algorithm.EM algorithm is given a suitable initial value by particle swarm optimization algorithm,which makes EM algorithm more easily jump out of the local optimum when it is iterative optimization in the global,and improves the convergence efficiency.GMM algorithm is compared with PSO-EM-GMM algorithm by experiment,and the improved algorithm is proved to have better results by two indexes of detection rate and false alarm rate.Secondly,because of the lack of fault model in the heat exchange station,the anomaly detection algorithm is difficult to distinguish the fault types.An experimental simulation platform is built to simulate the steady-state fault and the fault characteristics in the process of dynamic temperature rise and fall by using the temperature control system.By calculating the Euclidean distance between the state vectors,the fault of the simulation design is judged.The modal differences between the normal state and the fault state verify the rationality of the fault design.Then the SAPSO-LSSVM algorithm is used to model the leakage fault and the different sensor fault under the mode of the heat exchange station,and the fault model when the heat exchange station has different faults is obtained.The accuracy of the model is verified by the fitting degree of the index.Different fault models correspond to different faults in the actual industrial field,and realize fault classification diagnosis and model induction.Finally,based on the established fault models of different types and different modes,PSO-EM-GMM algorithm and GMM algorithm are used to carry out anomaly detection based on the fault model.Through the comparison of detection rate and false alarm rate,the effectiveness of the improved algorithm is proved,and the application of fault detection and fault model diagnosis of heat exchange station is realized.The data from the heat exchange station can be compared with the model to determine the fault type.