This paper presents a practical usability investigation of recurrent neural networks (RNNs) to determine the best-suited machine learning method for estimating electric vehicle (EV) batteries’ state of charge. Using models from multiple published sources and cross-validation testing with several driving scenarios to determine the state of charge of lithium-ion batteries, we assessed their accuracy and drawbacks. Five models were selected from various published state-of-charge estimation models, based on cell types with GRU or LSTM, and optimisers such as stochastic gradient descent, Adam, Nadam, AdaMax, and Robust Adam, with extensions via momentum calculus or an attention layer. Each method was examined by applying training techniques such as a learning rate scheduler or rollback recovery to speed up the fitting, highlighting the implementation specifics. All this was carried out using the TensorFlow framework, and the implementation was performed as closely to the published sources as possible on openly available battery data. The results highlighted an average percentage accuracy of 96.56% for the correct SoC estimation and several drawbacks of the overall implementation, and we propose potential solutions for further improvement. Every implemented model had a similar drawback, which was the poor capturing of the middle area of charge, applying a higher weight to the voltage than the current. The combination of these techniques into a single custom model could result in a better-suited model, further improving the accuracy.

An attractive approach to increasing the output power and bandwidth of converters is through the parallel connection of multiple, subrated converter cells via mutually coupled inductors. Achieving a balanced contribution to the output current by all paralleled converter cells is critical; good reference tracking performance, however, should also be maintained. Since these two control tasks are coupled, tuning of the controller parameters is challenging. This article demonstrates a methodology for decoupling the output current reference tracking control task from the current balancing control. We show that once balancing and tracking controllers are decoupled, the controller parameter tuning can be conducted independently, simplifying design and analysis of controller performance. Experimental results are presented, which validate the proposed methodology, applied to a three-cell, parallel interleaved converter with cyclic cascade coupled inductor filter network, operating at 100 V _RMS and 9 A _RMS .

This study presents a field‐programmable gate array implementation of an advanced pulse‐width modulator which combines a first‐order hold (FOH) with phase accumulator carrier pulse‐width modulation (PACPWM) for significantly reduced phase delay, and bandwidth extension in multi‐level applications. The FOH block supports arbitrary resampling rates and operates in a single clock cycle for simple integration into multi‐rate or asynchronous systems. The proposed modulator is validated experimentally, with performance matching that predicted for both FOH and a zero‐order hold (ZOH). For PWM with 8 kHz carrier frequency and 50.4 kHz resampling frequency, a 15 degree phase delay advantage is demonstrated for FOH compared with ZOH across the 5–10 kHz modulation band.

Grid connected battery systems play an important role in present day power systems. In this paper, a DC-side sensoreless control of battery interfaced three-phase full bridge cascaded H-bridge multilevel grid-connected inverter is proposed. The key achievement of this paper is the use of the DC-link voltage and current estimation proposed in [1] for a three-phase battery storage system and achieving battery SOC balancing using the estimation algorithm. Coulomb counting (Amp-hour balancing) method is used for SOC estimation because of its simplicity and it can utilize the value of the estimated battery current directly. One disadvantage of this method is the need to identify the value of the initial SOC value which was assumed to be known. The balancing between the batteries has been done using the estimated SOC values. Simulated results are presented to show the efficacy of the proposed algorithm.

This study presents a newly developed modular system for control of power electronic amplifiers in high‐power, high‐frequency power hardware in the loop experiments. The proposed design comprises computational nodes connected via multi‐gigabit fibre‐optic links. Nodes are modular and reconfigurable, allowing interfacing with a range of sensors and actuators, typically voltage and current sensors, and switching converters. The system has been designed for scalability to permit arbitrarily high power ratings and fidelity to be achieved through paralleling and interleaving of multiple power converter modules, targeting an ultimate sample and control rate of 1 MHz. Experimental validation is presented using a two‐node configuration to facilitate a power electronic experiment operating at 120 V _rms and 10 A _rms , with a fundamental frequency of 50 Hz and frequency content up to the 11th harmonic.