Research On Leakage Discharge Analysis Model Of Water Distribution Network Based On Algorithm For Signal Processing

Water loss in the water distribution network(WDN) causes a huge waste of water resources and reduces the earnings of water supply company, and physical loss is the main component of the total water loss. Studying leakage discharge analysis model of WDN has important theoretical significance and application value for guideing the work of leakage control, especially for calculating the efficiency of pressure management. However, now the models, mainly exponent-type models, don’t perform satisfactorily in practical applications due to the lack of data to check them. Further, that how to use new theories and methods to enrich and improve the WDN’s leakage discharge analysis model is an important part of the research on water loss control. The research was carried out from local to global, the experiment of operating conditions was introduced and leakage exponent laws of single leak were studied firstly, then the leakage discharge analysis model of WDN was established by using the algorithms for signal processing.The impact of leakage stability, leak’s characteristics, pipe viscoelasticity and outside flow boundary condition on leakage was studied experimentally. By stress analysis of pressurized pipeline and comprehensive analysis of leak’s plastic deformation and elastic deformation, an improved fixed area and variable area discharge(FAVAD) model was established by derivation. The parameters in the improved FAVAD model has clear physical meanings, so it provides effective guidances to reduce physical loss. For the phenomenon that leakage discharge in pressure rising process is less than that in pressure falling process, the drop-restore pressure leakage control technology was proposed, then the drop-restore pressure leakage control model was established and its mechanism was explained by viscoelastic theory. From the experimental study and model’s analysis, the value of leakage exponent is as follows: 0.5~0.6 for circular, rectangular, cambered leaks, 0.5~2.5 for longitudinal cracks and 0.5~1.5 for circumferential cracks; for all longitudinal cracks and the circumferential cracks on the pipe without immobility, it is larger when pipe diameter is larger, pipe wall thickness is smaller and the elastic modulus is smaller. The relationship between leakage discharge and pressure head is complex, and there is uncertainty because of leakage stability and viscoelasticity.The total water supply flow was divided into two parts, consumed water flow and leakage discharge flow, and the conclusion was that consumed water flow and leakage discharge flow are correlative. In order to analyze leakage discharge in the WDN, applicability effectiveness of blind source separation(BSS) theory and filtering theory were analyzed comparatively, and three types of algorithms for signal processing had been choosed, they were Fast ICA algorithm, CICA algorithm and Kalman Filtering algorithm. Accordingly, three types of WDN leakage discharge analysis models, Fast ICA model, CICA model and Kalman Filtering model, were established based on the above-mentioned three types of algorithms. All the models were optimized with the data obtained in the WDN laboratory by the experiment of operating conditions, and their similarity coefficients are in order of 0.7834, 0.9454, 0.3351. After a comprehensive analysis of 11 evaluation indexes, it showed that CICA model has the best best performance.CICA model took the total water supply flow and head pressure at the entrance as two input parameters, its reference vector(RV) was generating with the input parameters, the input parameters were the hydraulic parameters got easily from WDN with district metered areas(DMA), and the higher-order statistics information of consumed water flow and leakage discharge flow was used, so the model solved the problem of complexity and uncertainty between leakage discharge and pressure heads. After argument, for the CICA model’s application, the precondition is that the total supply water flow is not gaussian, or though the total supply water flow is gaussian but the head pressure at the entrance is not gaussian. Because the simulation results of CICA model provide waveform informations only, and the the size of the amplitude is uncertain, the approach was to liste dualistic overdetermined linear equations to solve two standard deviations of consumed water flow and leakage discharge flow according to flow balance principle, the equality constraints of two unknown parameters and the constraints of leakage discharge’s standard deviation value range were studied, then the real amplitude was obtained. Further, the convergence of CICA model was well, and the parameters optimized had a relatively broad range. Various application methods of CICA model for different topologies of WDN were studied, and the model had been verified in the laboratory WDN and actual WDN. It seems that the CICA model performs well in the WDN with one entrance, in which the average error is less than 5.0%.The application method about the coupling of CICA model and research results of single leak was studied, where research results of single leak provided the constraints of leakage exponent’s value range mainly and CICA model provided leakage discharge as the verification data, so the parameters in the exponent-style global models had realized being checked by using the data of multi-conditions and the application effect was improved. The CICA model was applied to correct the pattern of water for the node demand distribution in TJ city’s WDN hydraulic model, and to estimate the degree of leakage discharge reduction by pressure management, both obtained good results. By the evaluation of the model, compared with other model methods and measurement methods, CICA model showe d a very prominent advantage when it was used to estimate the degree of leakage discharge reduction by pressure management.