| Hydropower generation unit is an important guarantee for the power system to adapt to the large-scale and high proportions of new energy as a stable and flexible energy supply.Due to the impact of power load demand and its complex coupling of water,machine,and electricity factors,the units need to face more frequent transient processes with dramatic changes in stability indicators such as unit vibration and pressure pulsation,and more complex coupling relationships among parameters,which seriously threaten the operational reliability and health of the equipment.In this work,the key scientific issue is to reveal the coupling mechanism between operating state variables of hydropower units and quantify the potential risk laws of multi-source index parameters.A new perspective for unified research on dynamic safety evaluation of hydropower generation unit is proposed,and the unit state maintenance strategy is further optimized based on the revealed mechanism.The following research results have been achieved:(1)From the perspective of information transfer,the coupling mechanism between the operating state variables of the unit is revealed.It is proved that the information causality between the two subsystems can be expressed by the information transfer of the two system intuitive variables in the transient process.Taking the load rejection condition of a hydropower generation unit as an example,it is shown that the information transfer relationship between variables is bidirectional and asymmetric,and the path and direction of information flow will not change with the change of the rejected load,the path of information flow is always unit output → operating head → unit speed → guide vane opening → flow.In addition,it is revealed that the information transfer relationship is across scales,which exists between the same and different time and frequency scale variables.(2)To improve the accuracy of feature extraction of multi-source coupling fault signals of hydropower generation unit,a novel strategy of noise reduction for multi-source coupling fault isolation of the units based on information causality(IFC)and variational mode decomposition(VMD)is proposed around the key issue of revealing the mechanism of multisource coupling fault signals of hydropower unit.The proposed IFC-VMD is applied to the fault feature extraction and noise reduction of Unit 3 of the S Power Station,and the redundancy and interference components are accurately eliminated.The results show that the proposed method has obvious advantages over the existing mature methods in reconstructing multi-source fault signals,and the noise reduction fidelity is increased by 3%.The actual effective fault characteristics are well preserved,which provides effective theoretical support and an analysis method for feature separation and noise reduction of multi-source fault coupling signals of hydropower units.(3)To accurately diagnose the early weak signal characteristics of unit faults,a novel process diagnosis model based on dynamic information transfer is proposed.This method innovatively realizes the quantitative description of the dynamic information transfer(DIT)characteristics between the unit signals and breaks through the traditional fault diagnosis mode paradigm of hydropower units based on signal feature extraction and analysis.The applicability and effectiveness of the method are verified by the measured data of a 250 MW hydropower unit.The results show that the proposed diagnostic model is more sensitive to detecting small amplitude anomalies.By comparing with PCA and KICA-PCA methods,the proposed model improved the diagnostic success rate by 84.7% and 90.6% respectively,advanced the detection delay by 43 and 33 test samples respectively,and the false alarm rate was as low as 1.2%.The research results provide a new perspective and technical support for the state process diagnosis of multi-sensor systems such as hydropower generation unit.(4)Based on the information causality between the parameters and variables,the complex state data of hydropower generation unit are deeply mined,and the auxiliary variable selection method based on the information causality and Page Rank algorithm is proposed.Through deep learning and fuzzy comprehensive evaluation,the unit state trend prediction and operation state risk quantification are realized.The results show that(1)The dimensionality of the whole state variables of the system can be effectively reduced by mining the causality between the information among the variables,and when the dominant variable and its dependent variable are used as the prediction input,the prediction errors of the dominant variable are within 0.1,and the lowest is 0.0006.By considering the information transfer sequences between the predicted object variable and its causal variable,the prediction input dimension in the case of lack of samples is increased to improve the prediction accuracy.The mean absolute percentage errors(MAPEs)for the prediction of unit vibration and pressure pulsation are 0.0043 and 0.3809,respectively.(2)For the unit of Z Power Station is in the Class D risk zone within the test load range of 130MW～250MW,and the maximum cumulative risk at 200 MW is 1.6265.Among them,the index with the highest risk contribution is the pressure in the blade-free zone,which accounts for up to 50% of the maximum risk contribution in all test load ranges.Based on the field test results,it is determined that the risk of the unit comes from the dynamic and static interference of the runner components.It is recommended that the power station reduce the abnormal vibration phenomenon during operation by increasing the diameter of the movable guide vane distribution circle or optimizing the runner.(5)Given the lack of a unified theoretical framework for condition-based maintenance of hydropower generation unit at the present stage,an integrated framework for reliability prediction and condition-based maintenance(CBM)strategy of hydropower generation unit based on generalized proportional hazard model(GPHM)and Semi-Markov decision process(SMDP)is proposed.The effectiveness and applicability of the proposed method are verified by a case study of a unit of Gezhouba power station.The results show that the proposed method can achieve unit reliability prediction and maintenance cost optimization under variable operating conditions.Compared with the mean time cost model(MTCM),the proposed method avoids 599 ￥/d operation cost and 15 day detection interval deviation in unit single equipment system analysis,respectively.Compared to the(non)constant maintenance threshold model,34 days and 12 days of maintenance waiting time are avoided in the analysis of multiple equipment systems for the unit.In addition,the effects of operating conditions and discount rates on the performance of the framework are discussed,which improves the completeness of the method.The research results provide theoretical and practical guidance for highly reliable hydroelectric generating unit or their wider application. |
PDF Full Text Request