The modern power system has witnessed an increasing penetration of distributed energy resources and modern loads with variable frequency drives. The increasing complexity brings great challenges to modern power system operations. Significant efforts have been made to develop more accurate power system operation and optimization methods. The tradeoff between computation and the model structure makes the problem nontrivial to solve and analyze. Enabled by the wide deployment of PMUs and advanced machine learning algorithms, we improve the conventional power system operation and optimization techniques by improving the accuracy of power system measurements, implementing new power system modeling structures, and performing parameter reduction. In this thesis, the improved autoencoder models are firstly proposed to do data cleaning including outlier detection and the reconstruction of the true values of missing values and outliers. Then, an updated component-based load model is developed by adding a new load component with a variable frequency drive. The Fokker-Plank operator and tensor structure are utilized to do the sensitivity analysis and parameter reduction. To cope with the difficulty in solving parameters of the complex component-based load model, a free-form dynamic load model is proposed by synthesizing a large number of basic physics-driven mathematical functions, which is proved to have excellent accuracy and generalization performance. Experiments based on the simulated data verify the effectiveness and advantages of the proposed methods compared with state-of-the-art methods.

Degree Date

Summer 8-3-2022

Document Type


Degree Name



Electrical and Computer Engineering


Jianhui Wang

Number of Pages




Creative Commons License

Creative Commons Attribution-Noncommercial 4.0 License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License