Research On Data-driven Fault Diagnosis Of Sewage Treatment Plant

The normal operation of the sewage treatment plant is related to the treatment effect of urban sewage.Due to the complexity and uncertainty of the sewage treatment process,the failure of the sewage treatment plant often occurs,which brings challenges to the normal operation of the sewage treatment plant and affects the quality of the discharged water.In the past,the fault diagnosis of the sewage treatment plant is mostly from the perspective of system level,based on a deep learning algorithm to monitor and analyze the relevant data of the treatment process,to judge the operation of the sewage treatment system.An important precondition of this kind of fault diagnosis process is to assume the correctness of the collected data,so it is very important to detect and diagnose the operation status of the sensor.In this paper,a sewage treatment plant in Heilongjiang Province is taken as the research object,and the fault of the sewage treatment plant is deeply studied.The main research contents are as follows:The related processes and key steps of the activated sludge process in wastewater treatment were studied,and the differences between the research object and BSM2 model were compared.The research object was simplified to facilitate the follow-up research.Then,the relevant fault diagnosis methods are studied.By comparing the advantages and disadvantages of each algorithm,the combination of different algorithms is tried to improve the accuracy of the algorithm.In order to detect and diagnose the condition of sewage treatment system more comprehensively,the rule of fault diagnosis hierarchy is proposed,which is divided into sensor level,component level,equipment level and system level.In this paper,sensor and component level fault diagnosis will be the main research object to complete the first two levels of fault diagnosis application research of sewage treatment plant.Gearbox and bearing are the key parts in the main equipment of sewage treatment plant.Therefore,for the fault of key components in the sewage treatment plant,a GA-ACO improved BP neural network algorithm is proposed.By optimizing the combination of the parameters of the neural network,we try to speed up the training speed of the algorithm and improve the accuracy of the final results.To verify the effectiveness of this method,a fault diagnosis scheme for key components of the sewage treatment plant is designed,and the final classification results are compared with the unimproved BP algorithm.The results show that the training speed of the above method is faster and the diagnostic accuracy is higher.In this paper,a sensor fault diagnosis method based on KPCA and SPE is proposed to detect and classify the common faults.The location of the sensor is determined according to the effective index SVI of the sensor,and the fault data is reconstructed by using the cyclic iteration method.Design the sensor fault diagnosis scheme,collect the data of flow and liquid level sensor of a sewage treatment plant in Heilongjiang,verify the design scheme and analyze the results.The experimental results show that the amount of data required in the training is small,only a small amount of sampling data can complete the training of the algorithm,and the fault recognition rate is high,which can effectively locate the fault sensor,and the reconstructed data obtained by the reconstruction method is close to the real measured value.