Autonomous parameter estimation for electric machines

点击翻译按钮获取中文摘要

Electric machines are becoming more prevalent in the automotive industry as they become the main propulsion system in road vehicles with the industry's shift towards emissions free mobility. With over 15% of new car sales being electric, being able to accurately characterise electric machines virtually is imperative for maximising their performance and efficiency. A key predictor of a model's ability to replicate transient behaviour is the accuracy of the parameters used to characterise the motor. Relying solely on the information and specifications provided by the manufacturer to create a robust model is impractical as they often only include information required for the machine's operation. The overarching aim of this work is to develop a procedure to automate the parameterisation of electric motor models for later use in the vehicle development process.There are mant potential use cases for motor models, and many motor architectures of interest. In each combination of use case and motor architecture, the appropriate motor model structure is expected to differ. Typically, the level of spatial and temporal resolution will increase when more insight into detailed motor performance is needed. Once a model structure is defined, the data required to parameterise and validate this model can be defined. Then, the experiments necessary to generate this data, along with the instrumentation required can be defined.Focusing on the model development of electric machines, this project aims to create an end to end workflow between model and data to increase model accuracy and adaptability to new units under test.The work will explore the potential for a general motor model and parameterisation procedure that is compatible with all the likely motor topologies of interest: flux switching, induction, and synchronous motor architectures. It will focus on implementing an autonomous parameter characterisation process, and on streamlining the experimental procedure behind the collection of data required for the parameterisation of an electric machine. Electric machine architectures vary enough to require bespoke motor models to simulate it's behaviour. Over the course of the first 6 months of the project, a model of a synchronous motor will be created with the aim of reaching an acceptable level of accuracy for the model's given application. The model at present is an ideal vector control model built in Simulink with a discrepancy between 3% and 5% from real world data. The next step is adding losses which can be categorised into 3 main categories: those which occur in the electrical circuit, magnetic circuit, and mechanical and ventilation losses. Furthermore, below 400rpm, the accuracy of the model decreases with speed and is near tangential with the trend line. The cause of this behiaviour could be due to an inaccurate data sheet used to parameterise the model, or limitations in the measurement instrumentation hindering its ability to accurately capture low speed data. The electric machine will then be tested through a predetermined, AVL parameterisation cycle to ascertain whether there is a variation between the true parameters of the machine and those supplied by the manufacturer. If the lack of accuracy at low speeds is due to limitations of the instrumentation, identical tests on a different machine could potentially be conducted at the IAAPS facility.I initially looked into vector control during the summer project, along with the operating principles of synchronous machines and how they were linked. With a background in ICEs and batteries, learning the fundamental characteristics of the type of electric machine I am working with is a necessity. Beginning this project with a laymans knowledge of an electric machine's operation has been a challenge and could lead to delays in the project, however, speaking to the AVL stakeholders to assertain what their objectives for the project will allow me to plan my workflowmore ef