An open-set fault diagnosis framework for MMCs based on optimized temporal convolutional network

Online DMD-based Kalman filter for current sensor fault-tolerant control of MMC-HVDC transmission systems

Improvement of local outlier factor algorithms for lithium-ion battery fault diagnosis

Modular multilevel converter (MMC) is an important part of high voltage direct current (HVDC) transmission system. With the increase of the number of submodules in series, the open circuit failure of IGBT is the most common fault type of MMC. Therefore, it is necessary to study IGBT open circuit fault diagnosis of MMC. The diagnosis method of IGBT open circuit faults for MMC is proposed based on probabilistic neural network (PNN) and Markovian distance (MD) algorithm to further improve the reliability of MMC. The fault diagnosis model based on PNN-MD theory is established by using Matlab/Simulink software. Then the MMC circuit model is established by using PSIM circuit model. The voltage and current characteristic values of each sub module circuit in normal working state and IGBT open circuit state are collected. Meanwhile, these values are used as training samples and input into pnn-md for training. Through simulation, compared with the application of BP neural network in fault diagnosis, the PNN-MD algorithm can diagnose IGBT open circuit fault of MMC faster and more accurately. The innovation of this paper is to combine machine learning with MMC fault diagnosis, and has a broad application prospect.

The widely used Line Commutated Converter (LCC) suffers from commutation failure which might cause severe economic and secure consequence. The recently proposed Modular Multilevel Converter (MMC) HVDC is expected to be an alternative solution. However, the half bridge MMC (HBMMC) cannot interrupt the short current caused by DC transmission line fault, which have stunted the engineering practice of MMC in the long-distance overhead transmission line situation. A hybrid LCC-FBMCC system was proposed in this paper. Firstly, based on the parameter of a practical ±500kV HVDC project, the electromagnetic simulation model of this LCC-FBMCC system was established in PSCAD/EMTDC. Secondly, the electromagnetic transient process under DC line fault was simulated. The transient over-voltage characteristics of the LCC-FBMMC system were compared with that of the LCC-HBMMC system. Finally, the DC fault ride ability and the impact of DC short fault on the power transmission was discussed, which can present reference to the development and application of MMC in remote and bulk power transmission area.

Battery health prognosis using improved temporal convolutional network modeling

Submodule (SM) fault diagnosis is necessary for the reliability of modular multilevel converters (MMC) and the complex operating environment of MMC may cause missing data and noise in signal detections, where some of the existing open circuit fault diagnosis methods can be disabled. An artificial intelligence (AI) diagnosis method for open circuit faults in MMC based on extreme gradient boosting system (XG-Boost) is proposed in this paper which can diagnose the open circuit fault quickly and accurately in the mentioned conditions. Compared with neural network diagnosis method, the proposed method has higher accuracy with the same non-ideal data. Faster speed and better processing ability of the proposed method are also validated by simulation of a three-phase MMC in MATLAB/Simulink.

This paper proposes a new strategy for balancing the Capacitor Voltages (CVs) for Modular Multilevel Converters (MMCs). The balancing is one of the main challenges in MMC applications and it is usually solved by adopting a global arm control approach. For performing such an approach, a sorted list of the SubModules (SMs) according to their capacitor voltages is required. A common way to accomplish this task is to implement a sorting algorithm in the same controller used for the modulation technique. However, the execution time and the computational efforts of these kinds of algorithms increase very rapidly when the number of SMs grows. A novel idea is presented in this paper by using a mapping strategy that directly stores in a ranked list the SMs according to the measured CVs. Avoiding the use of sorting algorithms leads to a considerable reduction of the execution time as well as the computational efforts.

Buildings are subject to faults in their heating, ventilation and air-conditioning (HVAC) systems that can lead to excessive energy wastage, poor indoor climate, equipment failures and high maintenance costs. Field studies have shown that employing fault detection, diagnosis and prognosis (FDDP) tools followed up with equipment services and corrections can help achieve up to 40% of energy savings within the HVAC system and improve indoor climate, increase equipment lifecycle and reduce maintenance costs. The increasing adoption of building automation systems (BAS), Internet of Things (IoT) and other smart technologies in recent years have allowed large amounts of data to be continuously collected from building systems. This data-rich environment, along with the surge in data analytics and machine learning tools, has made cost-effective data-driven FDDP strategies possible. Compared to purely physics-based methods, data-driven methods require less explicit knowledge of the underlying physical system, thus are often easier to develop, and can learn certain intricate relationships that exist among data. Within the reported data-driven FDDP methods, there exists a few research gaps: 1) data imputation methods that leverage mutual information from correlated measurements to defy poor data quality from BAS have not been utilized efficiently; 2) there lacks a systematic and scalable fault diagnosis framework that incorporates probabilistic temporal relationships to track fault evolution; 3) existing fault diagnosis strategies typically focus on traditional rule-based control strategies and their scalability for advanced control strategies such as Guideline 36 have not been explored yet; 4) active threats information such as cyber-attacks, are typically not incorporated in an FDDP framework; 5) fault prognosis strategies to preemptively identify gradual faults for predictive maintenance have rarely been studied. This research attempts to address the above-mentioned research gaps through the following: Data Imputation: reported data imputation methods that are suitable for handling and repairing multi-source BAS data are evaluated. Data collected from a medium-sized, mixed-use institution building situated in Stockholm, Sweden and a small commercial building simulated in a laboratory setup is used to evaluate five different data imputation methods. Results demonstrate that incorporating time-lagged cross-correlations within the k-Nearest Neighbor (kNN) method helps to significantly improve the imputation accuracy and minimize the impact of repaired data on data-driven algorithms. Dynamic Bayesian Network (DBN)-based Framework for Cyber-Physical Fault Diagnosis: a DBN framework with discretized conditional probabilities parameters to represent the temporal relationships among building measurements is developed. Both domain knowledge and machine learning methods are used to develop the DBN structure and parameter model. The developed framework is evaluated for both traditional rule-based and Guideline 36 controls using datasets from a real building, a laboratory building, and a virtual testbed. Results show that the developed DBN framework is effective in diagnosing and isolating faults in systems even with different control strategies. The framework also successfully distinguishes whether system abnormalities originate from cyber-attacks or naturally occurring physical faults. Potential future direction to improve fault isolation using modified DBN topological structure is also reported in this study. DBN-based Framework for Fault Prognosis: an extension of the DBN framework in conjunction with Robust Multivariate Temporal (RMT) variate selection is proposed for fault prognosis. The RMT variate selection is used to extract localized temporal features from high dimensional datasets to determine the best inputs for training forecasting models. The expected fault-free behavior of multiple target variates, selected using domain knowledge, is forecasted using incoming data. The prediction errors generated from the forecasting phase are used as evidences in the DBN inference to estimate future fault probabilities. Gradual faults simulated in the virtual testbed are used to evaluate the prognosis framework. Results show that the developed framework is effective in prognosing gradual faults by leveraging the trending growth on the prediction errors. The research presented in this thesis contributes to the overall objective of developing a robust and cost-effective DBN-based framework for fault diagnosis and prognosis of building HVAC systems. Potential solutions to other existing challenges of implementing data-driven FDDP strategies, such as obtaining high-quality datasets, handling and repairing missing data, establishing a baseline model for detecting abnormalities despite other disturbances such as weather and internal conditions changes, and extracting temporal features from timeseries data are also examined.

To ensure the reliable operation of modular multilevel converters (MMCs) in their widespread applications, the isolation of the faulty components from the system is mandatory. Power switching devices are the components in MMCs with the highest failure probability. Among different failure types of switching devices, the diagnosis of open circuit (OC) faults is extensively instigated because it is the most common fault type of MMC. In this paper, the fault diagnostic techniques presented in the literature for MMC in case of switch OC failures are analyzed and summarized. The fault diagnostic strategies are studied from model-based perspectives, which has the advantages of easy implementation and lower cost.

Summary Electrical type tests are a key step to verify the design of voltage source converter valves for a specific modular multilevel converter (MMC) high-voltage direct current (HVDC) project. For a high-power handling capability of modern insulated gate bipolar transistor (IGBT), it is very difficult to balance between type test quality and test rig investment. In compliance with IEC 62501 and CIGRE 447 requirement, this paper presents a kind of synthetic test circuits (STCs), consisting of a steady-state test circuit, an IGBT overcurrent turn-off test circuit, and a short-circuit current test circuit, used for operational tests on MMC-based IGBT valve sections. Using the STCs, the current stress, the voltage stress, the junction temperature, di/dt, the performance of the valve's electronic circuits, and the protection capability of the parallel thyristor can be validated. Further, the parameters of the test circuits can be adjusted with the number of submodules of the MMC test valve. Thus, the STCs are an economical alternative for testing a representative part of an MMC valve. The paper describes and discusses the type tests, completely performed with the three test circuits, on MMC valves for a Xiamen MMC HVDC scheme, which is rated at 1000 MW/±320 kV. Copyright © 2016 John Wiley & Sons, Ltd.

The expansion of renewable energy systems has led to the widespread use of large-capacity converters, and the application of modular multilevel converters (MMC) is expanding. In large-capacity MMCs, the carrier frequency is limited by the switching device characteristics. Therefore, in steady-state response, the characteristics of MMCs output a smooth sinusoidal waveforms, but there is still a problem with the response to load fractuation. In this paper, a new space vector PWM (SVPWM) method that improves the transient response by 1MHz sampling between carriers and compensating the output vector in real time. The output vector is recalculated every 1us by using an FPGA based hardware controller, and the recalculation is immediately reflected to the current vector, which improves the output characteristics of MM Cs with low carrier frequency. The effectiveness of the proposed method is verified.

Although the application of modular multilevel converter (MMC) is becoming more and more widespread, due to its working mode, there will be low-frequency ripple voltage in the sub-module (SM) capacitors. The most significant ones are the fundamental frequency ripple and the second-order frequency ripple, which will increase the volume of the SM capacitors and affect the application of MMC. This paper intends to change the traditional MMC SM topology to two capacitors with symmetrical structure, let the ripple voltage in the symmetrical structure capacitor the same in magnitude and opposite in direction. Then, the voltage ripple of the SM can be completely eliminated, and the capacitance value used can be reduced. Experimental results prove the validity of the method.

The model of a modular multilevel converter (MMC) determines the extent of critical circuit information that electromagnetic transient simulations can reveal. In this paper, two MMC models are proposed for efficient real-time hardware-in-the-loop (HIL) emulation on the field-programmable-gate arrays (FPGA). The nonlinear switch-based model employing the insulated-gate bipolar transistor (IGBT) dynamic curve-fitting model considers factors affecting its transient performance so that device-level behavior such as power loss and junction temperature can be reproduced accurately in the electromagnetic-thermal simulation of a power converter for its design evaluation. Meanwhile, regarding the MMC submodule as a transmission line stub achieves faster computation speed and enables the formation of a hybrid arm to save FPGA hardware resources. As the large network that the MMC presents is burdensome for real-time execution with a small time-step, circuit simplification based on partitioning and merging is conducted. Hardware implementation of a three-terminal high-voltage direct-current (HVDC) system containing an MMC-based DC-DC converter is carried out and the efficacy of proposed models is validated by comparing HIL emulation results with the off-line simulation tool PSCAD/EMTDC.

As the modular multilevel converter (MMC) becomes more used in medium voltage applications, its usage along with wide band gap devices, with low conduction losses and high switching frequency capability, seems promising. Currently it is not rare to see these converters fully equipped with SiC-MOSFETS in their structure, even tough the costs still can be prohibitive. To investigate this issue, this paper presents an analysis and discussion of the application of different devices technology for the cells of a Modular Multilevel Converter, e.g. SiC MOSFETS and IGBTs. The study addresses the different loading of the devices and how to better apply and select them considering different MMC applications. To further increase the efficiency of the topology, an analytical comparison is realized to evaluate the losses in different technology devices in parallel. At the end, a cost/benefit analysis is presented.

Modular multilevel converter (MMC) is a recently emerged multilevel topology for high-voltage high-power applications. However, in wind power application the performance of the MMC has not been deeply investigated. In this paper the application of MMC in wind energy systems is studied. The converters used to connect the wind turbine to the grid having the rated active powers of 2MW and 10MW are designed and investigated. Electrical losses and thermal loading of the power devices in the proposed converter solutions are analyzed. The efficiency of the MMC converter under different P/Q boundaries defined by grid codes is investigated and compared with two-level and three-level NPC converters. It is concluded that it is possible to use the MMC in wind power application and the losses are distributed evenly between the sub-modules of the MMC converter. However, inside a sub-module the losses of the power devices are not equal that can lead to the de-rating of the converter.