Robust deep auto-encoding network for real-time anomaly detection at nuclear power plants
Shi Yong,Zhang Linzi
DOI: https://doi.org/10.1016/j.psep.2022.05.039
2022-05-31
Process Safety and Environmental Protection: Transactions of the Institution of Chemical Engineers, Part B
Abstract:Detecting anomaly conditions in nuclear reactor is a critical issue in safety management of Nuclear Power Plants (NPPs). Conventionally, the operating status are monitored in transient data with pre-designed labels by human operators or basic diagnosis systems. Nowadays, continuous time series data from multi-sensors are increasingly collected and emerging unlabeled abnormal status are monitored during the operation, making it challenging to capture both spatial and temporal dependency at each time steps without supervised labels. In this paper, a robust unsupervised Multi-Variate Convolutional GRU Encoder-Dncoder (MVCGED) method is proposed to perform anomaly detection and fault diagnosis in multi-sensor operation time series data. Specifically, MVCGED first construct each time steps into signature matrices to maintain both spatial and temporal features via sliding windows with inner-correlation and forget mechanism. Subsequently, A CNN feature extraction network , CNN-based GRU encoding network and CNN decoding network are implemented successively to capture and reconstruct the hidden patterns of the signature matrices. Finally, the reconstruction loss are further utilized to detect anomalies and diagnose faults. Extensive empirical studies based on PCTRAN nuclear power plant operation data demonstrate that MVCGED outperforms commonly-used baseline methods. Due to a wide range of benefits and advantages over conventional measurement devices, the phasor measurement units (PMUs) garnered huge attention worldwide by the electricity grid operators in the past decades. Many electric utilities have already deployed PMUs in their grids for network observability. This chapter presents an improved optimal PMU placement (OPP) formulation that reduces the number of PMU requirements and increases the number of measurements by ensuring complete network observability. The presented formulation considers the existence of zero injection buses in the network. It also incorporates the channel limitations of the available PMUs. Moreover, the OPP formulation is extended to encompass power system contingencies for example, single line outage and single PMU loss cases. Then, this chapter employs the grey wolf optimization (GWO) algorithm, a recently developed metaheuristic algorithm, to solve the developed OPP formulation on several IEEE benchmarked transmission and distribution networks. Besides, it compares the results obtained from the GWO and the mixed-integer linear programming to confirm the superiority of the metaheuristic technique. Finally, a comparative result analysis with the referenced work demonstrates the effectiveness of the developed OPP formulation and the employed solution methodology. This chapter presents the use of synchronized phasor measurements of voltage and current acquired by the phasor measurement units (PMUs) for online identification of the transmission line parameters and system Thevenin equivalent at a particular node. It illustrates three different methods for identifying the transmission line parameters depending on the number of PMU measurement sets available at the two ends of the line. The performances of these approaches are evaluated and quantified without measurement noise and in the presence of random noises and biased errors. Besides, a minimum of three PMU measurement sets at three different loading conditions is used to identify system Thevenin equivalent at a particular node. The presented approaches are tested on a 115 kV transmission system chosen from the Saudi Electricity Company network and verified with PSCAD/EMTDC and MATLAB/SIMULINK simulations. This chapter addresses fault diagnosis in the series-compensated lines (SCLs) of the transmission network. It presents an adaptive fault location algorithm for SCL using synchronized phasor measurements collected from the phasor measurement units (PMUs). The algorithm is model-free, and as such, it can be applied to transmission lines regardless of the type of series compensation. Moreover, the algorithm utilizes PMU measurements for online computation of the system Thevenin equivalent and line parameters to tackle the parameter uncertainty that occurs due to the operating history and environmental conditions. The adaptive fault location algorithm is implemented on a 400 kV transmission network with an SCL, and the study is carried out on PSCAD/EMTDC and MATLAB/SIMULINK platforms. Simulation results confirm the satisfactory accuracy and reliability of the adaptive fault location algorithm. The efficacy of the adaptive algorithm is investigated under fault information (resistance, inception angle, location, and type), prefault loading condition, and line compensation degree uncertainty. This chapter presents a phasor measurement unit (PMU) based adaptive fault location algorithm for three-terminal power transmission -Abstract Truncated-