Research On Source And Load Probability Model Based On Bayesian Network And Its Application

As a high proportion of renewable energy is integrated into the power grid,the uncertainty and correlation between wind power output and photovoltaic output and load is one of the important factors affecting the safe and stable operation of the power grid and the consumption of renewable energy.How to establish the joint probability distribution of the source and load random variables and describe the uncertainty and correlation of the source and load random variables is the basic research of power grid planning and operation control.A new source and load probability model based on Bayesian Networks(BN)is proposed in this paper.In the new model,the source and load correlation samples are used to calculate the probabilistic load flow of the distribution network,and the distributed generation(DG)with correlation between source and load is further studied.The key line identification and line overload risk assessment methods taking into account the correlation between source and load are proposed,and the electricity price interval prediction model considering high proportion of renewable energy participating in market competition is explored.The main work and innovation points include:(1)To solve the problem of difficultly establishing the joint probability distribution of multi-dimensional wind speed,multi-dimensional solar irradiation and load which follow arbitrary distribution,an innovative source and load probability model based on BN is proposed.To deal with information loss when the continuous data is converted into discrete data,the data discretization method with the maximum information coefficient is adopted to reduce the information loss and expand the application scope of the model.Then,taking the discrete data of the random variable’s probability values as samples,the source and load probability model is established through BN structure learning and parameter learning.And the model is sampled and the discrete data is transformed into continuous data.After the inverse transformation of cumulative probability distribution,the source and load correlation samples are obtained.The model was verified by comparing the probability density,cumulative probability distribution,numerical characteristics,three-dimensional scatter plot of the source and load correlation samples with the original samples and the C-vine Copula samples.The source and load correlation samples were used to calculate the probabilistic load flow of the IEEE 33 node and IEEE 69 node distribution networks,and the calculation accuracy of the source and load correlation samples was validated.(2)In the existing research on DG siting and sizing,the influence of the correlation between multi-dimensional wind speed,multi-dimensional solar irradiation and load on the planning scheme is not taken into account,therefore,the planning method of DG siting and sizing considering the correlation between source and load is proposed in the paper.Using annual comprehensive cost minimum as objective function,the node voltage amplitude and the branch of the apparent power as the chance constraints,and distribution network with DG installation capacity and superior power grid transmission and the direction of the trend as deterministic constraints,DG siting and sizing planning chance constrained model is set up,use elite retention strategy of genetic algorithm to solve the model.In the IEEE 33 node distribution network,the effects on the planning scheme of the source and load correlation are compared.The comparative results show that the planning model considering the correlation of source and load can accurately calculate the generation of DG and the power consumption of load,and the simulation of power flow of distribution network is in line with the actual situation.The annual comprehensive cost of the planning scheme obtained is high,and the probability of node voltage amplitude overlimit is low.In the IEEE 69 node distribution network,the planning schemes under the two conditions are compared,including the correlation between multidimensional wind speed,multidimensional solar irradiation and load,and the correlation between one-dimensional wind speed,onedimensional solar irradiation and load.The comparison results show that: compared with the latter,the total number of DG allowed to access by the former is reduced by11,and the annual comprehensive cost is reduced by 17.53%.The planning scheme is more in line with the engineering reality,and the configuration of renewable energy is optimized.(3)The existing static identification method for key lines cannot meet the new requirements for key line identification after high integration of renewable energy,and it is difficult to find potential key lines.In view of the lack of assessment methods taking correlation between source and load into account in the existing research on line overload risk assessment,a new method of key line identification and line overload risk assessment considering correlation between source and load is proposed.This novel proposed method combines the probabilistic optimal load flow with the downstream load flow tracing method,and takes the probability load flow interval and overload risk coefficient of the line as indexes to realize the identification of key lines and the overload risk assessment of the line.The identification and evaluation results of IEEE39 and IEEE 118 nodes show that the misjudgment rates of critical lines are 40% and50%.The misjudgment rates of critical lines are 20% and 40% when the uncertainty of source and load is considered but the correlation of source and load is not considered.The misjudgment rates of line overload risk assessment are both 40%.There is no misjudgment in the identification of key lines and the risk assessment of overload,and the accuracy of the identification and assessment results is high,which verifies the effectiveness of the proposed method.(4)In view of the low cost of renewable energy generation,strong output uncertainty,and easy to cause drastic fluctuations in electricity price,the existing direct interval prediction method based on neural network(NN)is difficult to select the weight coefficient of the sub-objective function in the comprehensive objective function.A dynamic Bayesian network(DBN)based electricity price interval prediction model was innovatively proposed.After analyzing the background in high proportion of raw energy participating in market competition,the characteristics and the main factors influencing the electricity price,using wind power generation,total capacity and total power consumption as random variables,DBN electricity price forecasting model is established,and then to wind power generation,the total capacity and total electricity consumption forecast as evidence reasoning,the electricity price interval forecast is realized by joint tree reasoning.The forecast results of electricity price range show that:DBN electricity price prediction model can give the electricity price forecast interval and the corresponding probability as well as the mean of the electricity price point forecast,and optimize kernel Extreme learning Machine(KELM)with Particle swarm Optimization(PSO),DBN has higher coverage probability,narrower average band width,smaller cumulative bandwidth deviation,and higher quality of electricity price forecast range.