Research On Multi-objective Optimal Capacity Configuration For Stand-alone Microgrid With WT-PV-DE-BES

Exploitation and utilization of renewable energy such as wind and solar energy in the form of the microgrid is an important method to achieve the sustainable development of energy and an effective way to achieve the energy structure transformation of"clean,low-carbon,safe and efficient".For remote areas where there is no power grid while there is abundant renewable energy,it is suitable to generate,transmit and use electric energy in the form of the microgrid.The planning and design for microgrid is the primary task of the constructing microgrid projects,and optimal capacity configuration is the core task of the planning and design for microgrid.Based on the needs of multi-objective optimization of planning and design for microgrid,this thesis takes the stand-alone microgrid with wind turbine(WT),photovoltaic(PV),diesel generator(DE),and battery energy storage(BES)as the research object.On the basis of establishing the multi-objective optimal capacity configuration model for microgrid,and taking the stochastic scenario technology to solve the optimal model as the mainline,this thesis focuses on the three key technical issues to carry out research,which are establishing a joint probability model for wind-solar-load data,formulating operation policy,and designing optimization algorithm.It has important theoretical significance and practical application value.The main research is listed as follow:(1)According to the topology of the microgrid with WT-PV-DE-BES,a multi-objective optimal capacity configuration model is established,in which objective function comprehensively considers economy,reliability and environmental protection,and constrain condition includes power balance,output power limit of distributed generation(DG),installed number limit of DG,operating range limit of State of Charge(S_OC),timing sequence change of BES and other aspects.The establishment of the optimization model lays the foundation for the follow-up research.(2)Considering the excellent performance of Wasserstein Generative adversarial networks(WGAN)in image generation and combing with the analysis of the characteristics of wind-solar-load data,a joint probability model of wind-solar-load data based on WGAN is proposed.This model does not need to know the probability distribution of wind-solar-load data in advance,but it uses two interrelated deep learning models,namely generators and discriminators,to directly fit the joint probability model of the wind-solar-load data.In addition,and this model is convenient for generating scene data.To improve the fitting accuracy and prevent over-fitting,the idea of transfer learning is used to construct the discriminator.The network structure of discriminator is constructed based on ResNet50V2which is a deep learning model with a strong classification ability.The network layer used to characterize the general characteristics of ResNet50V2 is retained,and the network layer used to characterize the special characteristics is removed.In addition,according to the characteristics of wind-load data,the convolution layer and the full connection layer are re-added at the bottom and top of the discriminant respectively.Corresponding to the structure of the discriminator,the generator is constructed by using convolution layer and full connection layer.Combined with the structure characteristics of the discriminator,generator and wind-solar-load data,the data structure of training samples is reconstructed by means of normalization twice and amplitude limited quadratic interpolation once,so that the data structure of training samples matches with the network model of WGAN,and the statistical characteristics of wind-solar-load data can be easily captured either.Standardization and regularization are adopted in training algorithm.Eventually,the monthly average,autocorrelation,cross-correlation and diversity of the wind-solar-load data generated by WGAN are analyzed.It shows that the the structure of WGAN model designed in this thesis is reasonable,training algorithm of WGAN is not overfitting,and the WGAN model can capture the statistical characteristics of multidimensional wind-solar-load data in temporal and spatial dimensions essentially.(3)Aiming at the operation policy problem,based on the analysis of the operation characteristics of the microgrid,the start-up and shut-down conditions of the DE are formulated,and the operation conditions of the microgrid are divided into seven states.Then the key problem of the operation policy for the stand-alone microgrid with WT-PV-DE-BES is found out,which is how to allocate the net load power when the BES and the DE follow the net load together.Three operation policies,namely BES priority tracking net load,DE priority tracking net load and DE priority operation at rated power,are designed.The implementation process of the operation policy is given.(4)In view of the complex characteristics of the optimal capacity configuration model for microgrid,such as strong nonlinearity,multi-objective and multi-constrained conditions,etc.,based on strength Pareto evolutionary algorithm 2(SPEA2),by taking advantage of the fast convergence speed in the early stage of the optimization process and overcoming the shortcoming of prematurity in the later stage of the optimization process,and inspired by the omnidirectional search of enumeration method and interpolation mathematical ideas,SARAP(a search algorithm referencing adjacent points based on SPEA2)is proposed.Firstly,the SARAP determines the contour of the optimal solution set quickly by the SPEA2to narrow the search space,and then omnidirectional search in a small search space constructed through adjacent points which are extracted in the optimal solution set solved by SPEA2,so that the dominance of the optimal solution set is further improved,and the integrity is close to the real Pareto optimal solution set.Eventually,the unity between the integrity of the solution result and the operation efficiency is realized.The analysis of both computational complexity and convergence for this algorithm shows that the operation speed of SARAP is much higher than that of the enumeration,and the obtained result is close to the complete real Pareto optimal solution set.Furthermore,the idea of the SARAP algorithm can be extended and applied to improve other intelligent algorithms,which has general methodological significance.(5)Multi-view point analysis of the optimization calculation results further verifies the effectiveness of the optimal capacity configuration for microgrid.Firstly,the optimization calculation results of deterministic method based on both historical data and SARAP algorithm are presented in a variety of spatial forms.Combined with several groups of typical capacity configuration scheme selected by TOPSIS,a multi-objective decision algorithm,the factors affecting the economy,reliability and environmental protection of optimal capacity configuration for microgrid are analyzed.Secondly,compared with the deterministic method,the solution method of random scene technology,which is based on historical data,scene data generated by WGAN,and SARAP algorithm,can effectively deal with the uncertainty of wind-solar-load data,thus improving the robustness of the optimal capacity configuration results of microgrid.Thirdly,the calculation results under the three operation policies and the production simulation results of the same optimal configuration scheme under the three operation policies are compared and analyzed.The results show that the operation policy of BES priority tracking net load has the best environmental protection,the operation policy of DE priority tracking net load has the best economy,and the operation policy of DE priority operation at rated power has the highest reliability.Finally,the effectiveness of the optimal capacity configuration of microgrid is verified in this thesis by the production simulation results of typical configuration schemes with one-year data and typical daily data.The research work of this thesis enriches the theory of multi-objective optimal capacity configuration for the stand-alone microgrid,expands the application of deep learning in scenario generation of wind-solar-load dada,improves the performance of algorithm for solving multi-objective optimization problem,and has certain reference value for improving the accuracy of microgrid planning and design.