A Hybrid Probabilistic Algorithm for Computationally Efficient Estimation of Power Generation in AC Optimal Power Flow

• Type: Article in International Conference Proceedings Book
• Date: 2020

Authors

Mohamed Lotfi (Author) Other; Shaden Fikry (Author) Other; Gerardo J. Osório (Author) Other; Mohammad Javadi (Author) Other; Sérgio F. Santos (Author) Other; João P. S. Catalão (Author) FEUP

Conference proceedings International

Title: 2020 IEEE 14th International Conference on Compatibility, Power Electronics and Power Engineering (CPE-POWERENG)

Pages: 169-174

Indexing

ISI Web of Knowledge - 0 Citations

Scopus - 0 Citations

INSPEC

Other information

Authenticus ID: P-00S-QXK

DOI: 10.1109/cpe-powereng48600.2020.9161685

Abstract (EN): Decentralization of power systems is creating a need for tools which can provide fast and accurate optimal power flow (OPF) solutions, without being dependent on the availability of all system information and/or uncertain variables. In this study, a hybrid probabilistic algorithm is proposed to accurately and efficiently predict ideal generation levels of individual generators to minimize the total system cost (as per AC-OPF), while having no information on the grid structure and with limited information on system variables. The proposed hybrid algorithm combines the use of correlation analysis, k-means clusters, and kernel density estimation (KDE), to predict ideal generation levels of each generator based only on historical datasets of local information (i.e. adjacent load centers). By simulating the AC-OPF problem on the IEEE 9-bus test system, a historical dataset of 1000 samples is synthetically generated and randomized local information is given as input for each agent. Quasi-deterministic Monte-Carlo simulations with 100000 samples were used for validation. In the most uncertain operating conditions, the proposed algorithm was capable of predicting the ideal generation level of the most expensive generator with a 1.65% error, while being three times faster than a Neural Network (NN), taking only 0.39 seconds to run on a standard laptop computer. © 2020 IEEE.

Language: English

Type (Professor's evaluation): Scientific

No. of pages: 6