In this paper, a novel rotor position estimator for an interior permanent magnet synchronous motor is presented and evaluated. The proposed estimator lies on one of the most popular methods in the field of artificial intelligence : Machine learning-based neural networks algorithm. The main interest is to introduce a cost-efficient position estimator which is comparable to classic methods in terms of functional performances. The estimator model is built by learning from a dataset that associates phase currents and voltages to the rotor position. Learning signals are generated using a simulation model. This is primarily intended to save the resources invested in testbench trials. In this work, off-line training steps and results are described and commented. The efficiency of the proposed position estimator is first verified by functional simulations. Second, real-time experiments are conducted on an actual scale testbench. The NN-based estimator covers a wide speed range and is implemented in the context of IPMSM-based EV traction drives. More broadly, these findings can also be applicable to the AC-based electric drives for the position estimation purpose.

Model predictive control (MPC) is a model‐based control strategy that introduces many advantages but also some drawbacks. The biggest challenge that faces MPC controller is its sensitivity towards parametric variation. This study aims to robustify MPCs by adding a least mean square (LMS) identification algorithm within their structure. The feasibility of this method is tested on a simple example which is the MPC current control of a permanent magnet synchronous motor driven by a two‐level power converter. Many tests are firstly performed without including inside the MPC algorithms the LMS online tuning procedure. These tests are later compared with the ones obtained with the same control technique but where the identification algorithm is turned on. Results clearly show all the benefits of using this approach on the controller performances.

This article suggests a new approach for designing embedded real-time (RT) simulators for power electronic converters. The main concept is to approximate the voltage/current characteristics of each power switch by dedicated coefficient varying transfer functions. The latterare obtained through system identification. The potential of such approach is the possibility to make this identification from real measurements and considering the end-user electrical/thermal environment. The converter is then structured into independent switching cells, represented by dedicated RT models that are executed in parallel. This is the same inside each switching cell model where the inherent transfer functions are simultaneously activated. Furthermore, each transfer function has been implemented with a parallel form giving the possibility to achieve high identification accuracy without compromising the timing performances. Then, with such full-parallel organization and using field programmable gate array devices, it is possible to achieve very low latencies and consequently, a short simulation time step. This article presents the developed proof of concept, applied to power converters with different levels of complexity, namely, a half-bridge dc–dc, a full-bridge dc–ac, and multilevel cascaded H-bridge (5-level and 9-level) power converters. For each case study, RT results are provided, analyzed, and compared withtheir offline counterparts.

Model Predictive Control (MPC) algorithms are computationally intensive optimization based control techniques. Their complexity significantly increases when the cost function prediction horizon increases. This problem is more pronounced when applying this techniques to systems having fast dynamics, in this case the prediction horizon length is limited by the calculation power of the processor. The use of long prediction horizons should provide healthier results than single prediction horizon yet prediction horizon equal to one are widely used for the control of electrical system and are achieving beneficial results. This paper proves mathematically that in some special cases of MPC applied to first order systems the use of a prediction horizon equal to one is similar to the use of a prediction horizon greater than one. An MPC algorithm is then applied for the control of the current of an RL load driven by an H-Bridge, multiple prediction horizon are used and results are compared.

Model Predictive Control is an optimization based control technique. It has the benefits of being flexible and achieving a quick time response. However this controller suffers from its computational burden and its sensitivity toward measurement noises, parametric variations and model inaccuracies. Nowadays, the computation of MPC is no longer a problem thanks to the new digital devices that exist on the market. However, robustness is still a major problem that complicates the usage of MPCs in applications where the system model and parameters are not accurately known such as the control of outer loops in cascaded control schemes. This paper presents multiple attempts to robustify the speed outer loop of a Synchronous Motor driven by a two-level power converter. An intuitive MPC speed outer loop algorithm is firstly introduced. This first algorithm is weak against parametric variation, model inaccuracies and measurement noises. Thus, a second more robust algorithm based on a torque balance expression is presented. This algorithm is robust toward parametric variations and model inaccuracies but weak against measurement noises. Therefore, an ultimate improvement is added in a third algorithm in order to achieve a robust MPC controller. Moreover, a Least Mean Square identification algorithm is added to all control laws in order to identify the dynamical model parameters. These MPC controllers were experimentally tested and compared with a classic PI controller. Finally conclusions are drawn.

Their importance and reliability make resolvers key building blocks in flight control and navigation dynamics. This paper gives an overview of the different synchronous demodulation methods, for such devices and lists advantages and drawbacks for each one. A particular attention is given to implantation in the same device of both analog to digital converter (ADC) and algorithm. This true system on a chip (SoC) solution is realized by the fusion field programmable gate array (FPGA) from Actel Company.