The increasing integration of renewable energy sources and power electronic devices is changing the electricity grid, leading to widespread harmonic and supraharmonic distortions. Accurate measurement and calibration of current transformers (CTs) up to the 150 kHz range are essential for reliable power quality assessment and grid monitoring. However, traditional calibration approaches are limited in both bandwidth and practicality, particularly for high-current and high-frequency conditions.

This thesis develops and validates a broadband calibration methodology for CTs, enabling ratio and phase error characterization from 50 Hz to 150 kHz using a high-precision digital sampling ammeter (from a power analyser) as the core measurement instrument. The proposed system eliminates the need for auxiliary equipment and thus reduces component count, ultimately allowing for simplified broadband calibrations. An uncertainty budget is established with combined expanded uncertainties (k=2) for the measurement system of less than 10 ppm up to 10 kHz, and less than 100 ppm at 150 kHz for the secondary-to-secondary comparison method. This is an improvement over the previous state of the art for this setup, which had an uncertainty of 50 ppm and a maximum frequency of 10 kHz. For primary-to-secondary calibration, uncertainties remain below 110 ppm at the highest frequency, allowing for the further development of a reference current transformer.

The thesis systematically examines the influence of critical experimental factors, such as grounding configuration, shunt selection, conductor positioning, cabling, and measurement duration, on overall calibration accuracy and repeatability. Key findings include the importance of instrument warm-up, the impact of earth-loop currents, and practical considerations for shunt and cable selection for high-frequency application. The demonstrated approach provides a metrological foundation for future implementation of wideband CT accuracy classes and supports ongoing international efforts to establish traceable measurement infrastructure for power quality applications.

This work, carried out at the Dutch national metrology institute (VSL), aims to contribute to the goals of the European ADMIT project.

The creation of effective computational models that function within the power limitations of edge de- vices is an important research problem in the field of Artificial Intelligence (AI). While cutting-edge deep learning algorithms show promising results, they frequently need computing resources that are many orders of magnitude more than the available power and memory budgets for these devices. During the thesis, two unique learning algorithms (backpropagation and forward-forward) were developed and compared using the Teensy 4.1, a low-cost microcontroller board. This work seeks to bridge the gap between the necessary computing efficiency and the hardware’s restricted resources.

By creating and analyzing these algorithms, with the Fashion MNIST dataset as a validation set, this thesis creates a baseline for AI efficiency on microcontrollers, with performance targets set at a mini- mum of 80% test accuracy. The microcontroller software, implemented in C++, is limited to using less than 512 kB RAM for all online training methods. In addition, the potential of transfer learning was also explored.

Key performance parameters, including memory utilization, training and inference times, and accu- racy, were analyzed in a comparative study of the backpropagation and forward-forward algorithms. For each learning algorithm, several configurations were explored (such as topologies, and optimizers) to determine the most effective and efficient way for AI implementation on low-cost hardware. The key conclusions of this study reveal that backpropagation demonstrates superior performance in terms of both accuracy and computational efficiency. However, it requires more memory for storing variables, which may be a constraint in on-edge environments. Conversely, the forward-forward algorithm, while achieving lower accuracy, is more memory-efficient, making it a potential choice for less complex tasks or systems with severe RAM limitations.

The application of transfer learning showed potential to accelerate the learning process and to improve the final accuracy, hinting at an effective strategy for deploying advanced AI models on resource-limited edge devices.