Study On Leakage Fault Diagnosis For Heating Network

The large-scale modern heating system is a complex technical system with abundant components and complex operation conditions. The fast-expanding and aging heating supply network is its weakest part with increasing reported faults and breakdowns. Leakage is commonest problem. Leakage diagnosis, an essential component of the reliability of the heating system, guarantees efficiently the heating network's economical and secures operation and improves efficaciously the heating network's management. Studying and exploring leakage diagnosis methods is a new question to heating supply enterprises and researchers.The paper probes into the leakage of the hot water supply network, attending particularly to the investigation and statistical records of hot water supply network, simulation of the hydraulic leakage conditions based on spatial network, and the leakage diagnosis method.First, an investigation is conducted among the first stage hot water supply networks of 12 heating power companies in Heilongjiang Provinces, into the faults of pipelines and components, and the quality of maintenance personnel. The investigation finds that most of the faults are not accidental but the results of gradual erosion in the pipelines, valves or compensators. Severe faults can be forecasted by little leakages which if are well taken care of will be removed and never deteriorate. However, the current leakage detection, mainly conducted by manpower, is subject to all kinds of factors, which bungles the best chances of repair and increases the hazards. As a result, it is valuable to diagnose the little and moderate leakages of the hot water supply networks.Against the space asymmetry of parameter and topological structure in the heating networks of supply and return water pipelines, which is caused by the leakage condition, the graph theory is used to depict the space topological structure of the hot water supply network, and to establish a hydraulic computation model of the leakage condition for the spatial hot water supply network. The model is applied to analyzing the operational parameters of circulating pumps, pressures of nodes, flow of pipelines, available head and flow of heat users, considering respectively the leakage amounts on the same node, and the same leakage amount on different nodes. It finds that the parameters of the branch-shaped heating network change regularly with leakage, and that the circumstances of the loop-shaped heating network are more complicated due to its unique structure, though some patterns could still be found.One pipeline and all its components are taken as one unit and the hot water supply network is abstracted into a system consisting of pipeline units. Thereupon, the BP neural network is applied to diagnose the pipeline leakage and thus a nonlinear identification model is built, which identifies the leakage pipeline according to the changes of hydraulic pressures of the monitoring points. Verified by a branch-shaped heating network and a loop-shaped heating network, the model can detect reliably the leakage pipeline, but stricter demands for neural network structure and training approximation are needed for the loop-shaped heating network. Applied to real heating systems, this model will improve the capabilities of problem solution.Base on the pipeline leakage diagnosis model, which is also called the first-stage leakage diagnosis model, hierarchy and modularization is introduced to the leakage diagnosis, which produces the second-stage diagnosis model for the hot water supply network to predict the particular location of leakage and its amount. BP neural network and support vector machine are used respectively to the construction of the second-stage leakage diagnosis model. Computed results show that the two methods can both solve the problems of nonlinear identification and that the model based on the support vector machine is more generalizable and predictable.A leakage condition regression model of the spatial hot water supply network is established on the basis of support vector machine with nonlinear function fit. With the pressure changes of the monitoring points, the model can work out the pressure changes of other nodes on the heating network. With this method, more information on the heating network is gained when leakage happens, which will effectively solve the problems of insufficient monitoring points and malfunctioning monitoring equipment, and guarantee the establishment of the leakage diagnosis model of the heating network.An empirical simulation method is provided against the incomplete faults records and insufficient data reserved, and is simulated under leakage conditions on a multi-function experiment platform. The two-stage leakage fault diagnosis model and leakage condition regression model are verified with observed data. It finds that the methods proposed in this paper have high practicality and can fulfil the needs of diagnosing leakage.