Search

Power electronics (PE) and high-frequency switching circuits are key to superior performance of electric vehicles. It is vital to monitor the condition of the PE components in real-time for safety and reliability. In this article, we propose two anomaly detection methods based on a combination of data preprocessing to suppress noise and outliers, multivariate regression models to predict signals of interest under nominal operation, and sequential analysis of residuals. In particular, the methods utilize median filtering to extract on-state medians in each switching cycle in nonlinear autoregressive exogenous neural network models or filtered on-state data in partial least squares-based models to represent the nominal circuit behavior. Optimal and approximate dynamic programming-based feature selection methods are developed to select the most informative signals or their transformations. Predictions from the learned models are used to generate the residuals for anomaly detection by Page's cumulative sum test. The proposed models and anomaly detection methods are validated on three accelerated aging experimental datasets, comprised of 60 power mosfet devices with low-frequency and high-frequency switching under disparate operating conditions. Due to the simplicity and efficiency of the data-driven anomaly detection schemes, the proposed methods can potentially be embedded in real-time digital platforms.

DC-link capacitors are at the heart of every power electronic inverter. A major drawback of DC-link capacitors is their aging factor which is due to various thermal and electrical stresses. This results in reduced system efficiency and complete system failure if left untreated. This paper proposes a data-driven condition monitoring system for the DC-link capacitor in grid-tied solar inverters. Compared to the literature, the condition of a DC-link capacitor is estimated without directly predicting the value of its capacitance or ESR. By treating the solar inverter system as a black box and by using the existing system sensors, the proposed condition monitoring method outputs the condition of the capacitor and classifies it as healthy, half-degraded or fully degraded. The proposed method is non-invasive and low-cost, where no additional sensors are required. By relying on classification techniques, this work addresses the difficulty of collecting real-world datasets with small incremental capacitance steps used in training regression models that predict capacitance values. Features not previously used in the literature are explored and result in very high classification accuracy. The proposed method is tested on a single-stage, single-phase grid tied PV inverter simulation model. The proposed method achieves 99% accuracy, a value similar to that found in the state of the art. The feasibility of the proposed method is tested on a low-cost microcontroller.

Permanent magnet synchronous motors are widely used in elevator applications. This is due to their high efficiency and reliability, whereby motors are directly connected to the elevator ropes without the need of having a gearbox. This efficient system comes with some new challenges regarding the control of the motor. Since no gearbox is added to the system, the motor speed oscillations are directly transferred to the elevator cabin. This urges the need of having a smooth control algorithm to suppress the measurement noises generated by the absolute encoder especially at low speed. On the other hand, whenever the mechanical brakes are released, the speed controller needs to react quickly to prevent the cabin from rolling back. In order to overcome these challenges, this paper presents an improved speed control algorithm that reacts quickly in transient state to reduce the rollback of the cabin and slowly at steady state to ensure smooth ride. The proposed controller is based on the combination of a Model Predictive Controller (MPC) along with a torque observer. It also benefits from being computationally less demanding unlike other MPC techniques. The proposed controller is experimentally tested on an elevator and then compared to classic controllers. Finally, conclusions are drawn.