Research On Demand Response Baseline Load Estimation Method Under High Penetration Of Distributed Photovoltaics

Compared with price-based demand response(DR),incentive-based DR has faster response ability and strong capacity expansion potential,thus plays an important role in improving operation flexibility of power grid and facilitate the absorption of renewable energy.As the calculation basis of IBDR participant compensation.Customer baseline load(CBL)is essential and crucial for the implementation of IBDR,its estimation accuracy has a direct impact on the benefit of IBDR provider and participants.Research on CBL estimation methods is of great significance to the large-scale implementation of IBDR.In this thesis,the individual CBL estimation method,the CBL estimation method for individual customers with distributed photovoltaics(PV),multi-customer cluster CBL estimation method under high penetration of distributed PVs,quantitative analysis of CBL estimation error under different application scenarios are studied.The main contents are as follows:1.Individual CBL estimation method.Current individual CBL estimation methods rely heavily on historical data,which show good performance when the load pattern on the DR day is similar to those in historical days,but yield large errors when they are not similar enough.This drawback is more apparent for residential customers with complex and volatile load patterns.On the basis of revealing the error generation mechanism of existing CBL estimation methods,this thesis points out the defects of existing CBL estimation methods at the principle level,and then a synchronous pattern matching principle-based CBL estimation approach is proposed.First,all customers are divided into two groups according to whether participating in IBDR program:DR group and Control group.Second,K-means clustering algorithm is used to group the daily load profiles of control group on the DR event day.Third,the synchronous pattern matching between DR customers and control group is performed.Fourth,CBL estimation is finished through optimized weight combination.Simulation results show that the proposed approach can effectively improve the individual residential CBL estimation accuracy.2.CBL estimation method for individual customers with distributed PVs.The installation of distributed PV has a significant impact on customers’load pattern.The stochastic and fluctuant characteristic of distributed PV introduce additional uncertainties to CBL estimation of customers with distributed PVs.It is very difficult to accurately estimate the CBL only using customers’ net load data(i.e.actual customer load minus distributed PV output power)when the output power data of distributed PV is unavailable.To this end,a "PV-load" decoupling-based CBL estimation method is proposed to realize accurate CBL estimation only using net load data and the output power data of a small amount of observable PVs.First,the classification of weather type is performed by using the output power data of observable PVs.On this basis,a net load curve optimal pairing method under different weather types is proposed.Afterwards,features reflecting PV capacity information are extracted.A support vector machine(SVM)-based ensemble estimation model of PV capacity and output power is established.Simulation results show the proposed approach can accurately estimate the capacity and output power of individual distributed PVs and can significantly improve the CBL estimation accuracy of customers with distributed PVs.3.Multi-customer cluster CBL estimation method under high penetration of distributed PVs.Multi-customer cluster CBL is not only the important basis of finial settlement for IBDR between system operators and load aggregators,but also is the essential reference for load aggregators to predict the available capacity and evaluate the IBDR implementation effect.The proportion of customers with PVs in the customer cluster continually increases with the increasing penetration of distributed PVs in distributed networks.The difficulty of customer cluster CBL estimation correspondingly increase.It is very difficult to realize the accurate estimation of customer cluster CBL estimation by using an accumulated or direct method.Therefore,the "PV-load" decoupling-based CBL estimation method is extended from indivial customer to multi-cutomer cluster.An optimal decouling method is proposed to disaggregate the PV ouput power and customer load from the aggregate net load.On this basis,a classified PV power forecasting and a CBL estimation model are established respectively and the estimated customer cluster CBL is obtained by subtraction.Simulation results show that the proposed method can realize the accurate decoupling of PV output power and customer load,and can significantly improve CBL estimation accuracy compared with accumulated and directed estimation methods.4.Quantitative analysis of CBL estimation error under different application scenarios.Different CBL estimation methods show different performances under different application scenarios,and have different impacts on the economic benefit of DR providers and participants.How to select a suitable CBL estimation method under a certain application scenario so as to ensure the benefit fairness and balance between DR providers and participants is an important issue needed to be addressed.To this end,the cost and benefit of load aggregator and customer for implementing/participating the IBDR are analyzed.A customer response behavior model under IBDR is established.The impact of CBL estimation errors on the profilt/benefit of load aggregators and customers is quantitatively evaluated.Afterwards,selection principles of CBL estimation method are given,which provide scientific guide for the selection of CBL estimation methods under certain scenarios.