Situation Prediction Technology Of Active Distribution Network

With the gradually urgent energy shortage and environmental pollution problem,renewable energy technology is in the midst of a vigorous development period.Especially,the small-capacity,decentralized distribution generations(DGs)which are principally used locally have become the important parts of intelligent power grid.Large scale of DG integration brings great challenges to reliability and safety of the distribution network operation,as well as the power grid control capabilities.Meanwhile,the demand of power consumers is increasing from "stable" to "high quality" and distribution network is the user-oriented terminal part of power system which directly decides power supply quality.In this context,the concept of Active Distribution Network(ADN)comes into being and has received extensive attention.Its vision of active control of power grid operation and active management of controllable resources is based on situation awareness technology: At a higher level of observation ability,ADN could predict potential risk with the help of situation prediction technology,and finally achieve active control through regulation of flexible resources.In order to enable ADN to perform regulatory operation and management prior to the actual operation status,situation awareness technology of ADN,especially the field of situational forecasting,is urgently needed for research and development.Based on the existing situation awareness framework of ADN,this paper focuses on the construction of the situation prediction framework of ADN and the specific technologies which contribute to the realization of the situation prediction.The main work done in this paper is as follows:(1)A situation prediction framework of ADN is proposed.This paper puts forward a framework including data source module,situation forecasting module and situational control module.The concrete implementation modalities for each module are also introduced.(2)A strategy for generating a set of typical operation scenarios for active distribution networks is proposed.This strategy integrates a static scenario generation strategy based on data-driven empirical probability distribution,a dynamic scenario generation strategy based on Markov Chain-Monte Carlo method,and a scenario dynamic reduction method.This “Data-to-data” method can be universally applicable to the scenario generation of various types of DGs and load in the ADN.The case of specific photovoltaic scenario generation verifies that the strategy is effective and practical.(3)An ultra-short-term prediction method for active distribution network is proposed.Elman neural network model is applied to ultra-short-term prediction work of ADN elements including DGs and loads.Similar scene identification work is added to the forecasting progress.The training samples entering the Elman neural network model are screened by similar weather recognition based on correlation coefficient and similar trend identification based on dynamic time warping.It is proved that the prediction accuracy can meet the actual demand of the ADN situation prediction.Finally,based on the ultra-short-term prediction results of the ADN elements,the trend of the operational status of the entire network is presented through the power flow analysis and visualized through the hotspot map.(4)A situation control scheme of ADN is proposed.In this paper,the operational status evaluation of the active distribution network is carried out from the risk perspective,and the evaluation objects include the typical scenarios and real-time scenarios.An ADN synthetic optimization model based on the improved particle swarm optimization algorithm is applied to the implement situation management to the typical risk scenarios and a typical scene optimization strategy set is formed.Gray correlation analysis is performed between real-time risk scenarios and typical risk scenarios,and the real-time risk scenario can be quickly optimized by using the optimization strategy of similar typical scenarios.The scheme realizes the vision of active management and control through the offline optimization of typical scenes and the online matching of real-time scenes.