| To address the problems of climate change and environmental pollution,large-scale wind,photovoltaic and other renewable energy sources are continuously connected to the transmission and distribution grids in centralized,decentralized and distributed manners,which leads to the transformation of the power system.In the process of renewable energy sources replacing traditional fossil fuel generations,the contradiction between the increase of uncertainty of loads and the reduction of flexible regulation capacity of traditional generators is becoming more and more prominent.The traditional centralized regulation is now difficult to maintain,and the transformation of the operation manner that combines decentralized autonomy and centralized regulation is imperative.With the maturity of technologies of measurement,communication and power electronic-based distributed energy resources,the active distribution networks are capable of controlling the external exchange powers and the internal power flows.The active distribution network will become an important way to reduce the difficulty of transmission system regulation and realize the local consumption of distributed renewable energy sources.The dispatch and control of active distribution networks will be a critical issue of the operation of future power system.To relieve the stress of transmission system dispatch and control,the operation of active distribution networks should limit the volatility and randomness of distributed renewable energy powers within distribution networks.In other words,the disorderly fluctuation of loads and passive powers can be eased by the active control of distributed energy resources,so as to realize the certainty of exchange powers.The autonomy operation makes the dispatch and control of active distribution network an independent sub-problem of future power system,which is important to reduce the difficulty of transmission system regulation and to achieve the optimal operation of both transmission and distribution systems.Current researches on active distribution network operations have proposed various optimization methods with regards to day-ahead dispatch,intra-day operation and real-time control.However,the current day-ahead dispatch researches have not paid enough attention to the allocation of reserved powers,which cannot ensure the feasibility of power exchange plan.For the existing intra-day operation researches,there are contradictions between the real-time control and the power flow optimization.To track the power exchange plan while ensuring the operation optimality and voltage safety is the key to the dispatch and control of active distribution networks.This thesis is a theoretical research on the optimal operation of active distribution networks,which mainly focuses on the alleviation of power system regulation difficulty,the autonomy operation of active distribution networks,and the distributed optimization algorithms.On the premise of developed measurement and communication technologies,this thesis studies the data-driven power flow linearization method,and realizes the operation and control of active distribution networks without the physical information of the grid.The real-time state estimation methods of active distribution networks are proposed as the premise of time-varying optimization research in this thesis,which achieves the fast mapping from measurements to the optimal control signals of distributed energy resources.Based on these researches,the dayahead and time-varying optimal operation methods of active distribution networks are proposed.The former includes the unified optimization of basic and reserved powers of distributed energy resources,which ensures the feasibility of power exchange plan.The latter is the realization of the power exchange plan,which corrects the exchange power deviation in an optimal power flow manner.This study provides a solution and corresponding theoretical supports for the operation of future power systems where the active distribution networks are fully developed,which is of great significance for reducing the difficulty of centralized operation of power system and realizing the local consumption of distributed renewable energy sources.The main works of this thesis are as follows:1)A data-driven linear power flow model for active distribution networks are proposed.Compared with the existing methods,the proposed model focuses on its application in active distribution network operation and control,without the requirement of PMU measurements.The model for multi-area systems are proposed to further ensure its feasibility.A principal component regression method considering measurement errors is proposed to process multicollinearity historical data,and the regression relationships between true values of independent and dependent variables can be obtained using noisy measurements.The calculation method of determination coefficients considering measurement errors is also proposed,and the fitting accuracy of the established linear equations can be analyzed using raw measurements.2)A static state estimation model,a dynamic state estimation model and an unified distributed algorithm for active distribution networks are proposed.The unified computing framework of the two estimation models realizes the flexible switching of estimation methods to adapt to different scenarios.The proposed static state estimation model improves the computational efficiency of the traditional current-based state estimation,while the dynamic state estimation can realize the real-time estimation of the states without the physical information of the grid.A unified distributed method for static and dynamic state estimation is proposed.According to this method,each area performs local estimations in parallel,and the coordination of estimation results can be achieved by transmitting limited information between areas.For static state estimation,the proposed distributed algorithm can minimize the number of coordination and communication between areas.Dynamic state estimation can achieve distributed computation with only once inter-area communication,which ensures the real-time estimation of states.3)The day-ahead optimal scheduling model of active distribution networks and its distributed algorithm are proposed.The day-ahead optimal scheduling problem includes the optimal decision of power exchange plan and the reserve plan for the corrections of exchange power deviations.The multi-period active and reactive power stochastic optimization problem is established,which realizes the coordinated optimization of distributed energy resource power and reserve.The proposed model ensures the satisfaction of operation constraints before and after utilizing the reserves.A data-driven method is used to transform the chance constraints into deterministic constraints,and the data-driven linear power flow equations are utilized to reduce the computation difficulty of the problem.A distributed method for solving mixed integer quadratic programming problem is proposed.Based on the alternating direction method of multipliers,the problem is decomposed into weakly coupled linear systems between areas and small-sized decoupled mixed integer quadratic programming problems.The former can be calculated in a distributed manner with a forward and backward inter-area communication,while the latter can be calculated in parallel by each area.The proposed method greatly reduces the difficulty and solving times of the problem.4)The time-varying optimal operation model of active distribution networks and its distributed algorithm are proposed.The optimal power flow model of active distribution networks considering power exchange constraints is established.The real-time state information is used as the input of the optimal power flow problem.After obtaining the state values,the solution of the optimal power flow problem and the control of distributed energy resources are completed immediately.As a result,the real-time and optimal response to power flow fluctuations can be achieved,and the power exchange plan can be tracked in an optimal manner.A real-time algorithm for solving the optimal power flow problem is proposed.The optimization problem is transformed into unconstrained optimization problem,which is further approximated with the second order Taylor expansion.At each optimization period,the optimization problem can be solved by calculating a linear system of equations only once,without requirement of iterative calculations.A distributed algorithm for solving the optimal power flow problem is further proposed.Each area calculates and stores a part of the linear system,and the linear system can be solved requiring only one forward and backward communication between adjacent areas.The correction values of distributed energy resources can be obtained immediately after the solution.The time-varying optimization method does not rely on the physical information of the grid.It realizes the real-time mapping from measurements to optimal control signals. |
