Water chiller fault diagnosis method based on hybrid neural network

Abstract

The invention discloses a water chiller fault diagnosis method based on a hybrid neural network, and aims at the problem that the single BP neural network is difficult to perform accurate prediction of the water chiller and the BP neural network has its own disadvantages. According to the method, a hybrid neural network model RBF-BP based on wavelet transform is put forward to perform fault diagnosis of the water chiller through combination of the characteristic that RBF neural network can approximate any function. The RBF network and the BP network are connected in parallel as one neural network, which is called the RBF-BP implicit strata for short. The algorithm has the advantages of both the RBF network and the BP network. Its learning process converges quickly and avoids the problem that the training process is liable to fall into the local minimum value. The method can be effectively applied to fault diagnosis of the water chiller and improve the performance of fault diagnosis.

Description

technical field [0001] The invention relates to the technical field of data-driven fault diagnosis, in particular to a method for fault diagnosis of a chiller based on a hybrid neural network. Background technique [0002] With the vigorous development of social economy, HVAC has become an indispensable important equipment for high-rise buildings. Chillers are the most important energy-consuming equipment in HVAC, and the energy consumption accounts for about 70% of the total energy consumption. Therefore, through the fault diagnosis of chillers, timely detection and resolution of faults are of great importance to the reliable operation of the HVAC system and energy saving. significance. [0003] Chillers are essentially nonlinear, coupled, time-varying parameters, and random working processes. There is a relatively complete selection of characteristic parameters for chiller fault diagnosis, including chilled water supply temperature, condenser inlet water temperature, con...

AI Technical Summary

Problems solved by technology

[0005] At present, most of the knowledge-based chiller fault diagnosis uses neural networks, most of which use a single neural network and other techniques such as wavelet decomposition, fuzzy system, genetic algorithm, etc. The most commonly used method is BP Neural network, these are just a point in the feature extraction analysis and network parameter optimization problems to make up for the defects of the BP network itself, it is difficult to solve the two problems of the neural network at the same time

Images