This is Qixing's github, which include the the experiment code and result of my research paper, with both my own method and the method used for comparision.
If you find some errors in my statement, or if you are interest in my research, please contact with me. My email is 10031@ahu.edu.cn.
In My first paper: "Rotor resistance and excitation inductance estimation of induction motor using deep-q-learning algorithm", I proposed an artificial intelligence method(Deep-Q Learning) to estimate induction motor's parameters. As Deep-Q Learning method is a data-based method, it has some advantages over some other traditional model-based strategies, such as Model Reference Adaption System(MRAS), Extenal Kalman Filter(EKF), Least Square Method(LSM) and so on.
In order to justify it, four comparision experiments were set up : 1, IEEE standard 211, the most common method, which is very familiar to the engineer; 2, Multiobjective Particle Swarm Optimization(MOPSO), which is easy to implement and very effective; 3, A latest model-based method: Least Square Method with start-up Transient Measurement(LSMTM); 4, Deep-Q Learning method(DQL). Because IEEE standard 211 is quite common, I will not open source it. I just open source the other three method: MPSO, LSMSM and DQL. If the author of MPSO or LSMSM think it is inappropriate to open source it, or find some erros in the source code, please contack with me, so I can delete it or revise it at your request.
The experiment is implemented in my test bench, you can read anoter document "Experiment Setup.md" to find more details. However, as it is difficult to release a product-level source code in github, I only released some demo code based on my MATLAB/SIMULINK test bench. Also you can find some information of SIMULINK test bench in "Experiment Setup.md".
Here are instructions about MOPSO:
- Download PSO.m and Sphere.m
- Open MATLAB, run PSO.m.
- When running finised, open workspace, find Gbest, open it.
- you can get the estimated rotor resistance value.
And here are instructions about LSMTM:
- Download qxforLsmooth.mdl, lsmforpaper-1.py, and lsmdatacollect.m
- Run qxforLsmooth.mdl in Simulink. Noticed that my MATLAB version is MATLAB2016.b, Unbuntu. If you find some error when running, maybe there are two reasons: 1), your MATLAB verson must be 2016b or higher; 2) you have to run the model in Unbuntu system, as I haven't test it in Windows yet.
- If step 2 finised successfully, run lsmdatacollect.m, the data will be stored in CSV format.
- run lsmforpaper-1.py, finially you can get the right answer. (K1 and K2 calculated only, and you can transform it to Rr or Lm)