Research Of Data-driven Two-stage Robust Optimization For Large-scale Variable Renewable Energy Integration

As an important measure to deal with fossil energy depletion and global warming,renewable energy power generation has attracted the attention of countries in the world.In recent years,renewable energy in China,represented by photovoltaics and wind power,has achieved rapid development,and its installed capacity has ranked first in the world.However,the output of renewable energy sources is subject to weather conditions,which makes it difficult to predict.With strong volatility and randomness,the renewables are quite different from traditional generator sets.With the large number of new energy sources connected to the grid,the traditional power generation dispatching strategy of power systems will face severe challenges.Therefore,how to achieve a new type of power generation dispatching that effectively responds to new energy forecasting errors and volatility will be the major problem for the operational department of the power system.The current power system scheduling methods for the renewables is summarized firstly.The robust optimization is then introduced to deal with the stochastic power for the renewables.In addition,the data-driven robust dispatch method is studied systematically.The main research results obtained are as follows:(1)The economic adjustment stratege of the controllable equipment in active distribution network with large-scale distributed reables is studied.Firstly,the economic dispatch problem is established based on the branch flow model.The second-order cone relaxation and the big-M approach are employed to convexify the original model to deal the optimization effectively.The historical data is then employed to construct the uncertainty set.Besides,the applicability of the proposed robust uncertainty set in the convex optimization model is proved theoretically.Furthermore,the two-stage solving strategy and algorithm for the robust model are introduced.The proposed model and solution technique are tested on a modified 33-bus and IEEE 123-bus electricity system.Numerical results show that the coordination of the regional and district multi-energy system can significantly improve the consumption of wind power.Case studies show that the proposed algorithm is superior to the current two-stage robust optimization with less conservativeness without the loss of security.(2)In the existing robust unit commitment,the decision is always too conservative.The data-driven robust unit commitment model considering the feasibility test is founded.The uncertainty set based on the actual data is produced in which the temporal and spatial correlations are considered simultaneously.Since the proposed problem is a large-scale programming which is difficult to solve directly,the decomposition architecture based on the column-and-constraint generation algorithm is introduced.The proposed model and solution technique are tested on a modified 14-bus and RTS 79 electricity system.Case studies show that the proposed algorithm improves the utilization rate of the online unit capability significantly and reduces the operation cost while guarantees the robustness of the system.This leads to the improved economy of the decision with uncompromised security.(3)The data-driven robust optimization in the context of multi-renewables and actual power system is a large-scale mixed interger double-layer programming problem,which is hard to sovle in the expected time.Firstly,the reasons why the exiting fast algorithm is difficuly to apply are analyzed.Then,the improved column-and-constraint generation algorithm is introduced combined with the characteristics of the data-driven robust uncertainty set.The sub-problem is decoupled and parallelized to speed up the process.In addition,the invalid constraint identification stratege is proposed to further reduce the complexity of the model.The proposed model and solution technique are tested on the modified IEEE 118-bus system and practical Polish system with 3120-bus.Case studies show that the proposed algorithm balances the robustness and economics of system operation well by taking the correlations of the renewables into account.In addition,the computational efficiency is significantly improved so that the proposed method is applicable to large systems.