Research On The Prognostics And Condition-based Maintenance Policy For The Hydroelectric Generating Unit

Condition-based maintenance (CBM) is hydroelectric generating unit is a kind of maintenancestrategy which can evaluate the equipmentcondition and predict the failure time and trend, then make the final decision.As an advanced maintenance technology the application of condition based maintenance (CBM) for the hydroelectric generatingunit has being a research topic of research value and economy contents.Prognosticsareone of the most important tasks for CBM. For the research on the reliability centered prognostics, these models are mainly derived from the failure event data and accompanied condition monitoring data, while the information value of the condition monitoring data before failure event are omitted.Maintenance decision making is one of the important tasks. Nowadays these CBM model for hydroelectric generating unit are modelled as single-component system, without considering the complex dependencies among the components. To make the CBM policy for generating unit much further accord with the industrialapplication, the research about the CBM policy for the generating unit as multi-component system is indispensable in terms of the sophisticated dependency among the components. As a result, this dissertation presents the research about the prognostics and CBM policy of complex system for the generating unit. Followings are the mainly about the research work and results:First of all, the thesis studies the method for failure hazard prediction. The classical proportional intensities model (PIM) is mainly used to for data analysis combining event data and accompanied condition monitoring data. To improve the accuracy of hazard forecasting, the monitoring data during the event interval is analyzed additionally as well as the accompanied condition monitoring data. As a result, a generalized proportional intensities model (GPIM) is proposed, and maximum likelihood estimation (MLE) is applied for the estimation for parameters of the proposed GPIM. The comparative results from the case study of Gezhouba Power Station also highlight the effectiveness of the proposed hazard forecasting model. Secondly, the thesis contributes three studies on the CBM policies for multi-component system. For the traditional CBM policies for generating unit, the system is normally modelled as single-component system for simplicity. For the Proportional hazards model based control-limit policy for the multi-component system with economic dependence, the analytical treatment of such cost evaluation issues becomes unfeasible but one possibility of solving the problem is to resort to Monte Carlo simulation. Additionally, a comparative case study highlights the effectiveness of the Monte Carlo simulation for multi-component system.Thirdly, for the engineering practice, the cost involved in the maintenance cannot be necessarily non-constant. For this sake, an extended control-limit policy is proposed from the constant control-limit policy for multi-component system, and the proposed control-limit is proportional to downtime-cost for CBM. The aim of the proposed policy is to reduce the maintenance cost rate by reducing (resp. improving) the possibilities of performing high (resp. low) downtime periods. The idea of the proposed policy is the peak-shift of maintenance timing. Additionally, two case studies highlight the effectives of the proposed downtime-cost dependent control-limit policy compared to the constant control-limit policy.Finally, to generalize the control-limit policy in terms of the non-constant downtime cost, a generalized downtime-cost based control-limit policy is proposed. The generalized control-limit policy is none-liner to the downtime cost. For the generalized control-limit policy, the high (resp. low) threshold cannot be necessarily assigned to the high (resp. low) price periods. In the case study, the forecasting electricity price f is derived by the SAMRIMA model based on the historical electricity price from the American PJM power market. A series of typical forecasting electricity price scenarios, derived by the Scenario Generation and Scenario Reduction method, are analyzed by the optimal thresholds, and then some relations between the optimal threshold and the electricity price series characteristics, such as the number of continuous periods and the number of cumulative periods for different price levels. The case study also highlights the effectiveness of the generalized non-constant control-limit policy. Further, the sensitive analysis of the proposed policy is discussed for different downtime cost ratios, reliabilities of component, and degradation process to conclude the scope of the policy.For the CBM for hydroelectric units in the power system, the related maintenance model and analytical methods are complicated, so there still are many opportunities and challenges. More research work should be done to put these theoretical models into industrial application. The research work of the dissertation only focuses on a small part of prognostics and CBM policy for hydroelectric generating unit as much component system and may provide some preference for the future work.