Study On Real-time Fault Diagnosis For Boiler In Power Plant

The boiler in power plant is a complex system, with strong coupling among process variables. Due to most of its equipments under high temperature and pressure environment, it is the most instable unit in power station. Therefore, it is important for safety in production to monitor operating parameters of boiler and diagnose fault according to the abnormal operating parameters. According to the characteristic of operation mode in production process, theories and approaches of fault detection and diagnosis (FDD) are mainly studied in this paper and applied to real design of FDD for fouling of heat transfer surfaces.Diagnosis based on dynamic causal model (DCM) is a dynamic fault diagnosis method combining qualitative reasoning and quantitative computation, which can express complicated cause-effect relationship and has capacity of containing large-scale potential information and good completeness. On the basis of summarizing the existing literatures of fault diagnosis based on DCM method, DCM is first presented to apply in fault diagnosis of boiler in power plant. Two kinds of residuals provided by DCM, which one is local residual and the other is upstream residual, are insensitive to fault detection, furthermore upstream residual of fault node in causal graph must be reconstruction during fault isolation. In this paper, another kind of residual, which is more sensitive to fault, is constructed to substitute for upstream residual. The improved method avoids the shortcoming of DCM, and improves the diagnostic speed.For process and systems with complex non-linearity, eliminating all unknown variables is not trivial, e.g. in the presence of algebraic loops, implicit equations, non-invertible functions, etc. However, most symbolically non-resolvable relationships can be numerically solved, diagnostic bond graph model (DBGM) is used to derive numerical residual. In this paper, the idea of combining DBGM with DCM is first presented to FDD of complex non-linearity systems. DBGM is used for detecting faults, and DCM is used for explaining the residual by diagnostic reasoning in causal graph. The advantage is the diagnosis time using DBGM-DCM can be reduced and the automatic degree of diagnosis process can be increased greatly.A fault diagnosis method based on DBGM-DCM is used to apply for fouling monitoring on heat transfer surfaces of boiler. A method similar to fuzzy control principles is used to deal with the uncertainty of thermal parameters. Sensor failures are very common in harsh industrial environments, and based on DBGM-DCM method can easily detect and isolate sensor failure. Operational condition variation, such as load reduction, load increase and attemperator putting into operation, can't have an obvious effect on diagnostic result of this method. In addition, this method can solve the problem of strong coupling among process variables.Combustion in furnace is a complicated physical chemistry process, so it is difficult to set up accurate model of combustion. Radial basis function (RBF) network is used to determine the local heat flux of membrane water wall when the tube surface is cleanliness. By analyzing an approximate analytical solution of membrane wall's temperature field, a linear model can be set up between the temperature difference of two key points on rear side and the local heat flux of membrane water wall. The degree of slagging is judged by comparing the local heat flux with the local heat flux when tube surface is cleanliness. The diagnosis result is good and practicable.