Learning-Based Probabilistic Power Flow Calculation Considering the Correlation Among Multiple Wind Farms

Xi Zhu,Chenbo Su,Jiaomin Liu,Chunming Liu

doi:10.1109/access.2020.3011511

Xi Zhu, Chenbo Su + Show 2 more

Open Access

https://doi.org/10.1109/access.2020.3011511

Copy DOI

Abstract

Wind power generation provides a new route for the sustainable development of energy. However, with the large scale integration of wind farms, the volatile wind energy and the correlation among various wind speeds bring a considerable quantity of uncertainty for the power system operation. In this paper, a probabilistic power flow (PPF) analysis model for the power system incorporating the correlation among various wind speeds is proposed. As distinct from existing studies, in this paper, we introduce the relevance vector machine (RVM) into correlation modeling of wind speeds via historical learning samples to construct bivariate joint distribution. Compared with the conventional parameter estimation methods, the proposed method has higher flexibility and computational efficiency. On this basis, the regular vine copula approach is adopted further to build the multivariate joint distribution model of wind speeds. To calculate the PPF of power system with the integration of wind power, we employ the three-point estimation method (3PEM) while the Rosenblatt Transformation technique is proposed to transform the input variables into independent variables. The effectiveness of the proposed calculation framework is examined through simulation studies, and the obtained results illustrate the advantages of the proposed method.

Highlights

With increasing concerns about sustainable utilization of energy resources, wind energy becomes one of the most popular substitutes for power generation
Compare with the existing studies, the major contributions of this work is that we introduce a novel learning-based distribution estimation approach to determine the multiple dimensional joint distribution functions to model the correlated wind speeds, which does not require parameter estimation
Compared with traditional estimation parameter methods, the proposed distribution estimation method more accurately construct the joint distribution functions for modeling the correlation among various wind farms

Summary

INTRODUCTION

With increasing concerns about sustainable utilization of energy resources, wind energy becomes one of the most popular substitutes for power generation. The results of PPF without considering the correlation will be too rough to provide reliable information for further planning and operational evaluation of the power system Facing these problems, the Copula function is proposed to model the correlation among wind speeds via constructing joint distribution function [20]–[28]. To accurately model the wind speeds correlation of multiple wind farms, a novel bivariate joint distribution estimate approach based on the relevance vector machine (RVM) has been adopted. Comparing with existing works in the relevant research field, the major contributions of this paper can be summarized as follows: 1) We propose a calculation framework to determine the PPF of the power system incorporating the correlation among multiple wind farms.

CONSTRUCTION OF CORRELATION AMONG WIND FARMS

R-VINE MODEL AND MULTIVARIATE DISTRIBUTIONS

ROSENBLATT TRANSFORMATION BASED 3PEM

METHOD

Findings

CONCLUSION