Permanent Magnet Synchronous Motors (PMSMs) are widely utilized in automotive and industrial applications due to their high efficiency, power, and torque density. Developing an accurate discrete-time model is crucial for implementing advanced control techniques, including direct torque control, model predictive control, and sensorless control. This paper presents the development of a discrete-time model for a PMSM, emphasizing numerical stability and dynamic accuracy. This model facilitates the creation of new current controllers, model-based observers, and parameter estimation techniques for motor drive systems operating at high fundamental frequencies and low sample rates. The model is derived using modified Euler discretization, extended to account for group delays and flux harmonics, and is compared with continuous-time simulations that include delta-sigma analog-to-digital converters. The switching inverter uses pulse-width modulation to generate the required constant voltages, with superimposed pseudo-random binary sequence (PRBS) signals, demonstrating the dynamic model’s accuracy. The same conditions are used for experimental validation. Comparison with simulations shows that the discrete-time model remains stable and accurate, while experimental verification demonstrates a significant increase in accuracy for motor drive systems with saturation.

Reducing the contact bouncing of electromechanical relays is a key ingredient to increase their switch reliability and overall lifetime. For this reason, this paper presents a control system as well as an optimization algorithm to reduce the kinetic energy of internal relay components by suspending acceleration during each switch cycle. A two-level control of the coil supply voltage is used in order to generate a control signal, which is defined by a certain start time as well as a specified duration. Both values are optimized via a computationally lightweight algorithm using a variant of a particle swarm optimization (PSO), which is a hybrid form of a bare-bone PSO and evolutionary PSO, augmented with dense estimation. Results are presented both qualitatively and quantitatively based on real-world experiments with various sample relays. Using the well-known run-to-run (R2R) algorithm as a baseline, the proposed bouncing optimization algorithm (BOA) reduces the switch on bouncing by 79 % ($+$24 % more than R2R) and the switching off bouncing by 59 % ($+$40 % more than R2R) compared to a conventional switch cycle without bouncing countermeasures. In addition, the proposed algorithm’s capability of self-adapting to changing environmental circumstances is validated experimentally.Note to Practitioners—Contact bounce increases relay wear. It is therefore desirable to reduce this. For this purpose, many mechanical optimizations have already been carried out. In this work, however, a software solution (BOA) is presented in which an online optimization algorithm is used to design the control signal of the relays in such a way that bouncing is reduced. The acceleration of the contacts is briefly suspended during the switching process, so that the contacts collide with less kinetic energy and, therefore, bounce less -as in the case of elastic impact. With BOA, it is possible to reduce the duration of bouncing by up to 79% during extensive tests. A transfer to various types of relays seems possible, but two points are critical for productization: First, a microcontroller is needed for implementation, which can significantly increase the manufacturing costs of a product. Secondly, a measurement signal of the load side of the relays is necessary, which can be technically challenging.

With electric power systems becoming more compact with higher power density, the relevance of thermal stress and precise real-time-capable model-based thermal monitoring increases. Previous work on thermal modeling by lumped-parameter thermal networks (LPTNs) suffers from mandatory expert knowledge for their design and from uncertainty regarding the required power loss model. In contrast, deep learning-based temperature models cannot be designed with the low amount of model parameters as in a LPTN at equal estimation accuracy. In this work, the thermal neural network (TNN) is introduced, which unifies both, consolidated knowledge in the form of heat-transfer-based LPTNs, and data-driven nonlinear function approximation with supervised machine learning. The TNN approach overcomes the drawbacks of previous paradigms by having physically interpretable states through its state-space representation, is end-to-end differentiable through an automatic differentiation framework, and requires no material, geometry, nor expert knowledge for its design. Experiments on an electric motor data set show that a TNN achieves higher temperature estimation accuracies than previous white-/gray- or black-box models with a mean squared error of 3.18 K2 and a worst-case error of 5.84 K at 64 model parameters.

Besides precise motor control, a suitable operating strategy is required for a highly efficient operation of an induction motor. Especially due to the present extensive use of these motors in torque-controlled applications, e.g., electric vehicles, efficiency improvements by optimizing the operating strategy are of prime interest. Most model-based operating strategies in the literature use simplified motor models or offline optimized look up tables, so that they are generally not able to adequately consider all loss effects in the entire operating range of the drive. In contrast, this paper presents an adaptive operating strategy that takes into account all relevant losses and changing system conditions (e.g., temperature, DC-link voltage) at runtime. Using a precise electrical-thermal motor model identified offline, the changing system states are adapted online and the resulting nonlinear optimization problem for loss minimization is solved iteratively. Experimental validation demonstrates the efficiency improvement of the adaptive operating strategy – Over the entire constant torque range, the efficiency of an industrial motor can be increased by an average of 0.7 percentage points compared to a standard operating strategy.

Electric motors are used in many applications, and their efficiency is strongly dependent on their control. Among others, linear feedback approaches or model predictive control methods are well known in the scientific literature and industrial practice. A novel approach is to use reinforcement learning (RL) to have an agent learn electric drive control from scratch merely by interacting with a suitable control environment. RL achieved remarkable results with superhuman performance in many games (e.g., Atari classics or Go) and also becomes more popular in control tasks, such as cart-pole or swinging pendulum benchmarks. In this work, the open-source Python package gym-electric-motor (GEM) is developed for ease of training of RL-agents for electric motor control. Furthermore, this package can be used to compare the trained agents with other state-of-the-art control approaches. It is based on the OpenAI Gym framework that provides a widely used interface for the evaluation of RL-agents. The package covers different dc and three-phase motor variants, as well as different power electronic converters and mechanical load models. Due to the modular setup of the proposed toolbox, additional motor, load, and power electronic devices can be easily extended in the future. Furthermore, different secondary effects, such as converter interlocking time or noise, are considered. An intelligent controller example based on the deep deterministic policy gradient algorithm that controls a series dc motor is presented and compared to a cascaded proportional-integral controller as a baseline for future research. Here, safety requirements are particularly highlighted as an important constraint for data-driven control algorithms applied to electric energy systems. Fellow researchers are encouraged to use the GEM framework in their RL investigations or contribute to the functional scope (e.g., further motor types) of the package.

Precise modeling and identification of induction motors is becoming increasingly important due to the extensive use of these motors in torque-controlled applications, e.g., electric vehicles. To achieve high precision, several nonideal motor characteristics including thermal effects have to be modeled and identified. Most thermal models in the literature utilize a loss model which is separated from the motor model considered in the control task leading to inconsistencies between these models. In this contribution, a combined electrical-thermal model is developed and its identification is addressed. Hence, the achieved universal drive model delivers flux, torque, loss and temperature estimations. Thus, the model provides information for three main drive tasks: general control, operating strategy and condition monitoring. With a comprehensive data set recorded at the test bench, the model parameters are optimally identified. On a separate test set, the proposed model is validated to estimate the torque generated by the motor with a root-mean-square error of 0.4 % related to nominal torque as well as the temperatures in the stator and rotor with root-mean-square errors of 1.0 K and 1.1 K, respectively.

For many permanent magnet synchronous motor (PMSM) drive applications (e.g., traction or automation), precise torque control is desired. Classically, this is based on extensive offline motor identification, e.g., by direct mapping of torque-flux-current look-up tables. In contrast, this article proposes a torque estimation method based on online differential inductances identification in combination with a data-driven finite-control-set (FCS) model predictive current control (MPCC). This scheme does not require offline identification or expert motor design knowledge. The required flux maps are determined by integrating the differential inductances in the left id-iqhalf-plane. By considering varying differential inductances, the proposed method is ideally suited for highly utilized PMSM with significant (cross-) saturation effects where estimation models with constant inductances fail. For the identification of the differential inductances, the system excitation, based on the FCS-MPCC working principle, is utilized. Consequently, no additional signal injection is required and the estimation scheme is applicable in the entire speed range. With this method, an open-loop torque control can be realized without knowledge of exact motor parameters except the permanent magnet flux linkage as a datasheet parameter. Extensive experimental investigations on a highly utilized PMSM in the entire speed range including standstill prove the performance of the proposed approach.

This article presents an approach to the controller design for electrical drives, which makes use of methods of deep reinforcement learning. Conventional control methods dominated the field for a long time, since they usually lead to control solutions with very robust and steady results. Yet, it often can be found that the overall control performance heavily correlates with the experience and education of the developing engineer. Moreover, conventional methods strongly depend on the available knowledge of the control system (e.g., plant model accuracy), which often causes the necessity for thorough identification methods. Real-time capability issues are also a present problem of sophisticated control approaches, such as model-predictive methods. Especially, in the domain of electrical drive train control, solving elaborate online optimization problems may be critical when very small plant time constants have to be considered. The methods of deep reinforcement learning will not only enable to acquire a suitable controller structure, but, moreover, the procedure will tune itself, which will allow for a more abstract level of investigation. This article presents a first proof of concept by means of controlling the phase currents of a permanent magnet synchronous motor in a field-oriented framework. The results found are promising and motivate further research in this field.

The main goal of an industrial microgrid during grid-connected operation is maximal cost saving for the microgrid owner. Many industrial companies do not only pay for the amount of electrical energy, but also for the maximum electrical power, which they have drawn from the distribution grid within the billing period. Under these conditions two basic options of cost saving exists utilizing the local energy storage systems inside the microgrid: reduction of the maximal power peak (peak shaving) and increase of self-consumption. For maximal cost saving, an operation strategy which combine both is desirable, but the combination requires information about the further residual power flow. A favorite option is the extrapolation of the residual power flow into the future. Unfortunately, it was found that errors in the extrapolation of the residual power lead to bad results in crucial situations. Therefore, this paper presents an additional artificial neural network (ANN) trained to predict residual power peaks, which will work in parallel to the extrapolation. This application-specific enhancement minimizes the effects of extrapolation errors and improves the original strategy in outcome and reliability. For an exemplary application, the self-consumption of the industrial microgrid is thereby increased by approx. 27 % compared to the original result without peak power prediction.

Monitoring critical temperatures in permanent magnet synchronous motors (PMSMs) is crucial to ensure safe operation and maximum device utilization as well. In this work, the application of recurrent neural networks featuring memory blocks (LSTMs/GRUs) are investigated upon their suitability to accurate temperature time series prediction inside PMSMs or similar motor types, which is the first time in literature to the author's best knowledge. Considered motor components are stator yoke, teeth and winding as well as the rotor's permanent magnets of a highly-utilized PMSM for electric vehicle applications. Having benchmark data available, numerous neural networks are trained and optimized with the aid of the Chainer framework and particle swarm optimization is conducted for finding suitable model hyper-parameters (e.g. number of hidden neurons or layers) on a computing cluster. It is found, that the Euclidean norm performance (in the range of 1-3 K) is similar but the worst-case predication errors (in the range of 9-14 K) are significantly higher compared to established modeling techniques like lumped-parameter thermal networks (LPTNs). This initial investigation motivates future research to increase ANN-based estimation accuracy by taken other ANN topologies, training methods or hyper-parameter optimization approaches into account.

The permanent magnet synchronous motor (PMSM) is widely used in highly utilised automotive traction drives and other industrial applications. With regards to the device life-time, safe operation and control performance, the magnet temperature is of great interest. Since a direct magnet temperature measurement is not feasible in most cases, this contribution gives a review on state-of-the-art model-based magnet temperature determination methods in PMSM. In this context, the existing publications can be classified into thermal models, flux observers and voltage signal injection approaches. Firstly, brief introductions of these methods are given, followed by a direct comparison regarding drawbacks and benefits. Finally, this contribution concludes with an outlook of potential further investigations in this research field.

Interior permanent magnet synchronous motors (IPMSM) are preferentially chosen as traction drives for electric vehicles due to their high power and torque density. In a field-oriented control scheme an operation point selection strategy is required to choose appropriate current set points to generate the requested torque with high precision and optimal efficiency. Although the currents are controlled in terms of a closed-loop control, the operation point selection is an open-loop torque control. Thus a precise motor model considering the impact of saturation effects and iron losses is required to estimate appropriate current set points. This paper proposes a lookup-table (LUT) based operation point selection. The lookup-tables are generated offline using the Maximum Efficiency (ME) strategy considering saturation effects and iron losses.