Flexible resource distributed optimal scheduling method and system considering wind power uncertainty

Abstract

The invention discloses a flexible resource distributed optimization scheduling method and system considering wind power uncertainty, and the method comprises the steps: constructing a distribution robust opportunity constraint according to a wind power prediction error fuzzy set, and carrying out the convex approximation of the distribution robust opportunity constraint through introducing a conditional value-at-risk, carrying out linearization processing on the distribution robust opportunity constraint after convex approximation; based on the distribution robust opportunity constraint subjected to linear processing, multi-element flexible resource differentiation operation characteristics of an energy storage system, an electric vehicle and an adjustable load are considered, and a flexible resource collaborative scheduling model considering the wind power uncertainty distribution robust opportunity constraint is constructed; adopting a consensus alternating direction multiplier algorithm to reconstruct the flexible resource collaborative scheduling model considering the wind power uncertainty distribution robust opportunity constraint into a distributed optimal scheduling model; and improving a consensus alternating direction multiplier method through a branch and bound method, processing an adjustable load discrete variable, and solving the distributed optimization scheduling model to obtain a flexible resource distributed optimization scheduling strategy.

Description

Technical Field

[0001] The present invention belongs to the technical field of power systems, and in particular relates to a flexible resource distributed optimization scheduling method and system taking into account wind power uncertainty. Background Art

[0002] Driven by the global shift toward a low-carbon energy structure, the power system is undergoing a profound transformation from a centralized, rigid architecture to a diversified, intelligent one. The evolution of this new power system exhibits three notable characteristics: First, the large-scale penetration of distributed resources, represented by renewable energy, energy storage systems, and electric vehicles, has reshaped the dynamic characteristics of the power grid. Energy storage devices achieve spatiotemporal decoupling of source and load through bidirectional power regulation, balancing load fluctuations on multiple timescales while providing rapid frequency response and voltage support for the system. Flexible resources, such as electric vehicles and adjustable loads, leverage demand-side response mechanisms to achieve dynamic regulation capabilities that match the fluctuating characteristics of renewable energy. Second, the rapid development of microgrids and regional autonomous power grids has given rise to new grid architectures. While their cellular structures enhance system flexibility and fault recovery capabilities, they also introduce the complexity of multi-level coordinated control. Microgrids must achieve dynamic balancing within seconds when switching between on-grid and off-grid modes, while the optimal allocation of distributed resources in regional power grids requires overcoming the response bottlenecks of traditional centralized dispatching models. Third, the uncertainty associated with the high proportion of renewable energy integrated into the grid has become a core challenge for system operation. The intermittent and meteorological-dependent nature of wind and solar resources means that power forecast deviations may trigger multiple chain reactions, including frequency exceeding the limit, line flow overload, and power quality degradation.

[0003] Therefore, effectively allocating various resources is crucial for solving the problem of collaborative energy management on both the supply and demand sides. Compared to traditional centralized optimization, which suffers from high communication pressure and the vulnerability of privacy information to leakage, distributed optimization scheduling demonstrates significant advantages through its two-tier architecture of "global collaboration and local autonomy," offering excellent scalability and privacy protection. Furthermore, extending traditional deterministic optimization scheduling to include the stochastic nature of renewable energy output is crucial for mitigating the uncertainty of renewable energy output at high penetration rates. Summary of the Invention

[0004] The purpose of the present invention is to provide a flexible resource distributed optimization scheduling method and system taking into account wind power uncertainty, which is used to improve the algorithm convergence problem of the distributed scheduling method under discrete resource access and the scheduling reliability problem under the scenario of uncertainty of new energy output.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is:

[0006] The present invention provides a method for distributed optimization scheduling of flexible resources taking into account wind power uncertainty, the method comprising:

[0007] The fuzzy set of wind power forecast error is defined based on Wasserstein distance to quantify the uncertainty of the probability distribution of wind power forecast error.

[0008] Based on the fuzzy set of wind power forecast error, a distributed robustness chance constraint is constructed. The distributed robustness chance constraint is convexly approximated by introducing the conditional value at risk, and the distributed robustness chance constraint after convex approximation is linearized.

[0009] Based on the distributed robustness opportunity constraints of linear processing, and considering the differentiated operating characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads, a flexible resource collaborative scheduling model is constructed that takes into account the distributed robustness opportunity constraints of wind power uncertainty.

[0010] The consensus alternating direction multiplier algorithm is used to reconstruct the flexible resource collaborative scheduling model with distributed robust opportunity constraints taking into account wind power uncertainty into a distributed optimization scheduling model.

[0011] The consensus alternating direction multiplier method is improved by the branch and bound method to process the discrete variables of adjustable loads, and the distributed optimization scheduling model is solved to obtain a flexible resource distributed optimization scheduling strategy taking into account the uncertainty of wind power.

[0012] In one embodiment, the steps of defining a prediction error fuzzy set based on the Wasserstein distance and quantifying the uncertainty of the probability distribution of the prediction error include:

[0013] The predicted power of wind farm i in period t Considered as a random variable ξ, a collection of N samples Defining the empirical probability distribution for:

[0014]

[0015] in, represents the lth sample of the random variable The Dirac measure of

[0016] Define two probability distributions P and The Wasserstein distance d between w for:

[0017]

[0018] where inf(·) represents the lower bound; Ξ∈R k represents the support set of the random variable ξ; ||·|| p Represents R k The p-norm on , takes the 1-norm; Φ represents ξ and On marginal probability distribution P and empirical probability distribution The joint probability distribution of

[0019] Based on two probability distributions P and The Wasserstein distance d between w , define the wind power prediction error fuzzy set Y as:

[0020]

[0021] Where P(Ξ) represents the set of all probability distributions in the support set; ε represents the radius of the Wasserstein sphere.

[0022] In one embodiment, the steps of constructing a distributed robustness chance constraint based on a fuzzy set of wind power prediction errors, convexly approximating the distributed robustness chance constraint by introducing a conditional value at risk, and linearizing the distributed robustness chance constraint after the convex approximation include:

[0023] According to the actual output and predicted output of each wind farm i in period t, the wind power output constraint is obtained;

[0024] Based on the fuzzy set Y of wind power forecast error, the wind power output constraint is rewritten as a distributed robust chance constraint under the confidence level 1-η;

[0025] The conditional value at risk is introduced to make a convex approximation of the distributed robustness chance constraint, and the distributed robustness chance constraint after the conditional value at risk is obtained by combining the saddle point theorem.

[0026] Take the polyhedron set Ξ={ξ∈R N :Hξ≤d}, combined with the maximum combination of two linear functions, the distributed robust chance constraint after convex approximation is linearized.

[0027] In one embodiment, assuming H = -I, d = 0, the linearized distributional robustness constraint is expressed as:

[0028]

[0029] Among them, λ t , π t,l , γ t,lj represents the dual variable introduced in the transformation process, Represents γ t,lj The transpose of Represents the candidate value of the risk threshold, and we need to find the optimal β t To calculate the conditional value at risk; γ t,lj is a column vector of the same dimension as d; express The lth sample of tN represents the tth component of 1 t dimensional standard basis vectors; P t W represents the actual output of the wind farm in period t; η represents the tolerance for constraint violation; T represents the total number of periods; Indicates a quantifier.

[0030] In one embodiment, the distributed robust opportunity constraint based on linearization processing considers the differentiated operating characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads, and constructs a flexible resource collaborative scheduling model that takes into account the distributed robust opportunity constraint of wind power uncertainty. The specific steps include:

[0031] The objective function is constructed based on the system's comprehensive operating cost and the minimum conditional value-at-risk cost; combined with Monte Carlo sampling simulation, the charging load of electric vehicles in autonomous charging mode is calculated;

[0032] Establish constraints for a multi-resource collaborative scheduling model; the constraints of the multi-resource collaborative scheduling model include: linearized distributed blue-robust opportunity constraints, node active power balance constraints, line transmission capacity constraints, thermal power unit operation constraints, energy storage system operation constraints, interruptible load aggregate constraints, and transferable load aggregate constraints.

[0033] In one embodiment, the objective function is:

[0034] minf total =minC G +C W +C ESS +C DR +f CVaR

[0035]

[0036] Among them, C G 、C W 、C ESS 、C IL 、C RL and f CVaR They represent the fuel cost of thermal power units, the cost of wind farm wind curtailment, the operating cost of energy storage systems, the compensation cost for interruptible loads, the compensation cost for transferable loads and the value-at-risk cost under the worst-case conditions; N G Indicates the total number of thermal power units; represents the output of thermal power unit i during period t; and Represent the fuel cost coefficient of thermal power unit i; N W represents the total number of wind farms; represents the unit wind curtailment penalty cost of wind farm i in period t; represents the actual output of wind farm i during period t; N ESS represents the total number of energy storage systems; represents the unit operating cost of energy storage i; represents the capacity of energy storage i in period t; N IL and N RL Represents the total number of interruptible load and transferable load aggregates respectively; c IL and c RL They represent the unit regulation power compensation cost of interruptible load and transferable load aggregate respectively; and They represent the regulation power of interruptible load and transferable load aggregate i in period t respectively.

[0037] In one embodiment, the charging load model of an electric vehicle in autonomous charging mode is:

[0038]

[0039] Among them, f EV (d) represents the lognormal distribution probability density function of the EV daily mileage d; μ1 and σ1 represent the expected value and standard deviation of the daily mileage function, respectively; f EV (t0) represents the normal distribution probability density function of the EV daily driving end time t0; μ2 and σ2 represent the expected value and standard deviation of the daily driving end time function, respectively; T i EV,ch Indicates the charging duration of EVi; Indicates the charging start time of the EVi, that is, the end time of driving for the day; d i Indicates the daily mileage of EVi; Indicates the power consumption of EVi per 100 kilometers; P i EV,ch Indicates the charging power of EVi; Indicates the charging efficiency of EVi; Indicates the charging end time of EVi; P t EV,chload represents the total charging power at time t; N EV Indicates the total number of EVs; represents the charging power of EVi at time t.

[0040] In one embodiment, the consensus alternating direction multiplier algorithm is used to reconstruct the flexible resource collaborative scheduling model that takes into account the distributed robust opportunity constraints of wind power uncertainty into a distributed optimization scheduling model. The specific steps include:

[0041] The variables of the flexible resource coordinated scheduling model considering the uncertainty distribution and robust opportunity constraints of wind power are sorted out, and the node active power balance constraints and line transmission power constraints are rewritten as follows:

[0042]

[0043] in, Indicates adjustable power; Indicates non-adjustable power; in N B Indicates the total number of nodes;

[0044] A i =[χ i ,-B i ,0], where Where E represents the system's correlation matrix, column vector and Each element represents P ij,max / X ij and P ij,min / X ij ;

[0045] According to the shared variable x in the scheduling period T i =[x i,1 ,...,x i,T ], divide the system into regions:

[0046]

[0047] Among them, f total,n represents the total cost of the nth sub-region; N sub Indicates the total number of sub-regions; D n Represents the set of nodes in the nth sub-region;

[0048] The Lagrangian dual function of subregion n is:

[0049]

[0050] Among them, λ y and μ y They represent the dual variables of the rewritten equations of the node active power balance constraint and the line transmission power constraint; x i ∈φ i Represents x i The constraint set φ that satisfies the distributed blue stick opportunity constraint, thermal power unit operation constraint, energy storage system operation constraint, interruptible load aggregation constraint, and transferable load aggregation constraint i ;

[0051] Set local variable z n To replace the consensus variable y = [λ y ,μ y], the Lagrangian dual problem corresponding to the Lagrangian dual function of subregion n is:

[0052]

[0053] The consensus alternating direction multiplier method (C-ADMM) is used to iteratively solve the Lagrangian dual problem, and the variables at the k+1th iteration are expressed as:

[0054]

[0055] Among them, := is the assignment symbol; δ n represents the Lagrange multiplier corresponding to the equality constraint in the subregion n of the Lagrange dual problem, represents the value solved at the kth iteration; ρ represents the given penalty factor;

[0056] The global optimization scheduling model is transformed into a regional sub-problem that can be solved in parallel. The sub-region scheduling model is expressed as:

[0057]

[0058] Among them, τ is an auxiliary variable; and They are The subvectors corresponding to λ and μ in ; and y k+1 The subvectors in that correspond to λ and μ.

[0059] In one embodiment, the method improves the consensus alternating direction multiplier method by using the branch and bound method, processes the adjustable load discrete variables, solves the distributed optimization scheduling model, and obtains a flexible resource distributed optimization scheduling strategy that takes into account wind power uncertainty, specifically including the following steps:

[0060] (1) For each sub-area after division, the relaxation includes the discrete variables of the interruptible load and the transferable load operation state. and The node power conservation constraint is used to transform the sub-area scheduling model into a linear programming model:

[0061]

[0062] The linear programming model is convex, if the optimal solution and If both are 0 or 1, then output; otherwise, go to step (2);

[0063] (2) Optional and A non-zero element in γ i,t Branch out and reconstruct the two LP problems:

[0064] LP1:

[0065]

[0066] LP2:

[0067]

[0068] (3) Solve the LP1 and LP2 problems with the newly added constraints respectively and obtain the optimal values L1 and L2, and the corresponding solutions are and According to different cases of the solutions, it is subdivided into:

[0069] If both γ1 and γ2 are 0 or 1, then min(L1, L2) is the value of the objective function to be found;

[0070] If both γ1 and γ2 are decimals, then select the variable corresponding to min(L1, L2) and continue to branch downward, and execute step (2);

[0071] If γ1 is a decimal and γ2 is an integer, if L1 > L2, then γ2 is the optimal solution of the sub-region scheduling model; if L1 < L2, then continue to branch downward for γ2 and execute step (2);

[0072] The above branching process stops calculating when a feasible solution is found or there is no feasible solution, otherwise enter step (4);

[0073] (4) Substitute the and values into the sub-problem of the region that can be solved in parallel, and the predefined tolerance parameter If the convergence criterion is not satisfied, then return to step (1); if it is satisfied, output the optimal solution.

[0074] The present invention also provides a flexible resource distributed optimal scheduling system considering the uncertainty of wind power, and the system includes:

[0075] A fuzzy set module for defining a fuzzy set of wind power prediction errors based on the Wasserstein distance and quantifying the probability distribution uncertainty of wind power prediction errors;

[0076] A distributionally robust chance-constrained module for constructing a distributionally robust chance constraint according to the fuzzy set of wind power prediction errors and introducing conditional value at risk for convex approximation of the distributionally robust chance constraint;

[0077] A scheduling model module for constructing a flexible resource collaborative scheduling model considering the uncertainty of wind power and the distributionally robust chance constraint based on the linearized reconstructed distributionally robust chance constraint, considering the differentiated operation characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads;

[0078] A distributed optimization scheduling model module is used to reconstruct the flexible resource collaborative scheduling model that takes into account the uncertainty distribution and robust opportunity constraints of wind power into a distributed optimization scheduling model using the consensus alternating direction multiplier algorithm;

[0079] The solution module is used to improve the consensus alternating direction multiplier method through the branch and bound method, process the discrete variables of adjustable load, solve the distributed optimization scheduling model, and obtain a flexible resource distributed optimization scheduling strategy taking into account the uncertainty of wind power.

[0080] The beneficial effects of the present invention are:

[0081] In this invention, a flexible resource distributed scheduling method that takes into account the distributed robust opportunity constraints of wind power uncertainty is proposed. By utilizing flexible resources such as energy storage and adjustable loads, the efficiency of local consumption and utilization of renewable energy is significantly improved, the uncertainty risk brought about by large-scale grid connection of distributed renewable energy is reduced, and the stable operation of the power grid is improved. Distributed scheduling effectively protects the privacy of sensitive users, thereby enhancing the operational reliability and adaptability of the system in uncertain environments. In terms of algorithm, an improved consensus alternating direction multiplier method based on the branch and bound method is used to improve the convergence under discrete variables of adjustable loads, significantly improving the iterative efficiency and robustness of distributed optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0082] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0083] Figure 1 A schematic flow chart of a method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty provided by an embodiment of the present invention;

[0084] Figure 2 This is a flow chart for calculating the total charging power of an EV in autonomous charging mode according to one embodiment of the present invention;

[0085] Figure 3 A schematic diagram of an improved IEEE 30-node system provided in one embodiment of the present invention;

[0086] Figure 4 The wind farm power at different confidence levels provided in an embodiment of the present invention;

[0087] Figure 5The convergence comparison of the C-ADMM before and after improvement provided in one embodiment of the present invention is shown in FIG. Figure 5 (a) is the tolerance parameter σ in traditional C-ADMM y As the number of iterations changes, Figure 5 (b) Tolerance parameter σ in improved C-ADMM y Changes with the number of iterations.

[0088] It should be noted that these drawings and textual descriptions are not intended to limit the conceptual scope of the present invention in any way, but rather to illustrate the concept of the present invention for those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0089] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments.

[0090] like Figure 1 As shown, this embodiment provides a method for distributed optimization scheduling of flexible resources taking into account wind power uncertainty, which specifically includes the following steps:

[0091] Step S100: defining a fuzzy set of wind power prediction error based on the Wasserstein distance, and quantifying the uncertainty of the probability distribution of the wind power prediction error.

[0092] In the embodiment of the present application, a prediction error fuzzy set is defined based on the Wasserstein distance to quantify the uncertainty of the probability distribution of the prediction error. The specific steps include:

[0093] (1) The predicted power of wind farm i in period t Considered as a random variable ξ, a collection of N samples Defining the empirical distribution for:

[0094]

[0095] in, represents the lth sample of the random variable The Dirac measure of .

[0096] (2) Define two probability distributions P and The Wasserstein distance d between w for:

[0097]

[0098] where inf(·) represents the lower bound; Ξ∈R krepresents the support set of the random variable ξ; ||·|| p Represents R k The p-norm on , takes the 1-norm; Φ represents ξ and On marginal probability distribution P and empirical probability distribution The joint probability distribution of .

[0099] (3) Based on two probability distributions P and The Wasserstein distance d between w , define the wind power prediction error fuzzy set Y as:

[0100]

[0101] Where P(Ξ) represents the set of all probability distributions in the support set; ε represents the radius of the Wasserstein sphere.

[0102] Step S200: constructing a distributional robustness chance constraint based on the wind power prediction error fuzzy set, convexly approximating the distributional robustness chance constraint by introducing the conditional value at risk, and linearizing the distributional robustness chance constraint after the convex approximation.

[0103] In the embodiment of the present application, the specific steps of implementing step S200 include:

[0104] (1) Based on the actual output and predicted output of each wind farm i in period t, the wind power output constraint is obtained.

[0105] Furthermore, the wind power output constraint is expressed as:

[0106]

[0107] in, represents the actual output of wind farm i during period t, denoted as x; N W represents the total number of wind farms; T represents the total number of time periods.

[0108] (2) Based on the fuzzy set Y of wind power forecast error, the wind power output constraint is rewritten as a distributed robust chance constraint under the confidence level 1-η.

[0109] Specifically, based on the fuzzy set Y, Equation (4) is rewritten as a distributionally robust chance constraint (DRCC) under the confidence level 1-η:

[0110]

[0111] (3) The conditional value at risk is introduced to make a convex approximation of the distributed robustness chance constraint, and the distributed robustness chance constraint after the conditional value at risk is obtained by combining the saddle point theorem.

[0112] The worst-case conditional value at risk (CVaR) is introduced to transform Equation (5), and combined with the saddle point theorem, we can obtain:

[0113]

[0114] Where sup(·) represents the supremum; represents the candidate value of the risk threshold; E P [·] represents the expectation under the random distribution P; [·] + =max{·,0} represents the positive part function, which only retains the positive part.

[0115] (4) Take the polyhedron set Ξ={ξ∈R N :Hξ≤d}, combined with the maximum combination of two linear functions, the distributed robust chance constraint after convex approximation is linearized.

[0116] Specifically, take the polyhedron set Ξ={ξ∈R N :Hξ≤d}, H and d are the coefficient matrix and vector of Ξ respectively. Combined with the maximum combination of two linear functions, equation (6) is linearized as follows:

[0117]

[0118] ||H T γ t,lj -y t,j (x)|| ∞ ≤λ t (9)

[0119] γ t,lj ≥0(10)

[0120]

[0121] Among them, λ t , π t,l , γ t,lj represents the dual variable introduced in the transformation process, γ t,lj is a column vector of the same dimension as d; express The lth sample of .

[0122] Furthermore, taking H = -I, d = 0, and linearizing equations (8) and (9), we obtain the distributional robustness constraint as follows:

[0123]

[0124] Among them, λ t , πt,l , γ t,lj represents the dual variable introduced in the transformation process, Represents γ t,lj The transpose of Represents the candidate value of the risk threshold, and we need to find the optimal β t To calculate the conditional value at risk; γ t,lj is a column vector of the same dimension as d; express The lth sample of t N represents the tth component of 1 t dimensional standard basis vectors; P t W represents the actual output of the wind farm in period t; η represents the tolerance for constraint violation; T represents the total number of periods; Indicates a quantifier.

[0125] Step S300: Based on the distributed robustness opportunity constraints processed by linearization, considering the differentiated operating characteristics of multiple resources such as energy storage systems, electric vehicles and adjustable loads, a flexible resource collaborative scheduling model is constructed that takes into account the distributed robustness opportunity constraints of wind power uncertainty.

[0126] In the embodiment of the present application, the specific steps of implementing step S300 include:

[0127] (1) Construct the objective function based on the system's comprehensive operating cost and the minimum conditional risk value cost.

[0128] Furthermore, the objective function is:

[0129]

[0130] Among them, C G 、C W 、C ESS 、C IL 、C RL and f CVaR They represent the fuel cost of thermal power units, the cost of wind farm wind curtailment, the operating cost of energy storage systems, the compensation cost for interruptible loads, the compensation cost for transferable loads and the value-at-risk cost under the worst-case conditions; N G Indicates the total number of thermal power units; represents the output of thermal power unit i during period t; and Represent the fuel cost coefficient of thermal power unit i; N W represents the total number of wind farms; represents the unit wind curtailment penalty cost of wind farm i in period t; represents the actual output of wind farm i during period t; N ESS represents the total number of energy storage systems; represents the unit operating cost of energy storage i; represents the capacity of energy storage i in period t; N IL and N RL Represents the total number of interruptible load and transferable load aggregates respectively; c IL and c RL They represent the unit regulation power compensation cost of interruptible load and transferable load aggregate respectively; and They represent the regulation power of interruptible load and transferable load aggregate i in period t respectively.

[0131] (2) Combined with Monte Carlo sampling simulation, the charging load of electric vehicles in autonomous charging mode is calculated as:

[0132]

[0133] Among them, f EV (d) represents the lognormal distribution probability density function of the EV daily mileage d; μ1 and σ1 represent the expected value and standard deviation of the daily mileage function, respectively; f EV (t0) represents the normal distribution probability density function of the EV daily driving end time t0; μ2 and σ2 represent the expected value and standard deviation of the daily driving end time function, respectively; T i EV,ch Indicates the charging duration of EVi; Indicates the charging start time of the EVi, that is, the end time of driving for the day; d i Indicates the daily mileage of EVi; Indicates the power consumption of EVi per 100 kilometers; P i EV,ch represents the charging power of EVi; η i EV,cha Indicates the charging efficiency of EVi; Indicates the charging end time of EVi; P t EV,chload represents the total charging power at time t; N EV Indicates the total number of EVs; represents the charging power of EVi at time t.

[0134] (3) Establish the constraints of the multi-resource collaborative scheduling model, including: linearized distributed blue-robust opportunity constraints, node active power balance constraints, line transmission capacity constraints, thermal power unit operation constraints, energy storage system operation constraints, interruptible load aggregate constraints, and transferable load aggregate constraints.

[0135] Furthermore, the node active power balance constraint is:

[0136]

[0137] in, represents the set of nodes where power flows into node i; represents the node set from which power flows out of node i; P ji,t represents the line power flowing from node j to node i during period t; P ik,t represents the line power flowing from node i to node k during period t; B ji and B ik Respectively represent the admittance of line ji and line ik; θ i,t represents the phase angle of node i during period t; i ∈{0,1} represents a discrete variable indicating whether node i has access to resources, χ i =1 means access; and They represent the output of thermal power unit, energy storage discharge power, energy storage charging power, rigid load, interruptible load, transferable load and EV charging load at node i during period t respectively; P i,t represents the net injected power of node i during period t.

[0138] Furthermore, the line transmission capacity constraint is:

[0139]

[0140] Among them, P ij,max and P ij,min Respectively represent the maximum and minimum transmission power of line ij; X ij Represents the reactance of line ij.

[0141] Furthermore, the operation constraints of thermal power units are:

[0142]

[0143] in, and Respectively represent the maximum and minimum output of thermal power unit i; P i G,u and P i G,d They represent the ramp-up and ramp-down rate limits of thermal power unit i respectively.

[0144] Furthermore, the energy storage system operation constraints are:

[0145]

[0146] in, and They represent the maximum charge and discharge power of energy storage system i respectively; and They represent the charging and discharging efficiency of energy storage system i respectively; Δt represents the scheduling time interval; and represent the maximum and minimum capacities of energy storage system i respectively.

[0147] Furthermore, the interruptible load constraint is:

[0148]

[0149] in, Indicates the operating status of the interruptible load aggregate i during period t, Indicates running; represents the rated power of interruptible load aggregate i; represents the minimum number of operating periods of interruptible load aggregate i in a scheduling cycle; T i IL and They represent the actual and minimum continuous running time of interruptible load aggregate i respectively.

[0150] Furthermore, the transferable load constraint is:

[0151]

[0152] in, represents the operating status of the transferable load aggregate i during period t, Indicates running; represents the rated power of the transferable load aggregate i; T i,start and T i,end are the start and end time of use of transferable load aggregate i; T i,min and T i,max They represent the earliest start time and the latest end time of the transferable load aggregate i.

[0153] To verify the feasibility of the above-mentioned embodiment method, Figure 2 This is a flow chart for calculating the total charging power of EVs in autonomous charging mode. Through Monte Carlo sampling, the total charging power of EVs within the scheduling period T is simulated and incorporated into the distributed power grid as charging load. Figure 3 Schematic diagram of the improved IEEE 30-node system. The improved IEEE 30-node system is divided into three sub-areas. The node sets in each area are D1 = [1-8, 28], D2 = [9, 11, 12-13, 16-17, 25-27, 29-30] and D3 = [10, 14-15, 18-24]; Figure 4 For wind farm power at different confidence levels, 30 groups of random samples are generated by random simulation using the Monte Carlo method as the wind power prediction error sample set, and the support set Ξ: = {ξ∈R k:-Iξ≤d}, where I is the unit matrix, and the radius of the Wasserstein sphere ε=0.03, from Figure 4 It can be seen that as 1-η increases, the output of the two wind farms gradually decreases and does not exceed the actual wind power prediction value, indicating that the proposed model is highly reliable.

[0154] Step S400: using a consensus alternating direction multiplier algorithm, the flexible resource collaborative scheduling model taking into account the distributed robust opportunity constraints of wind power uncertainty is reconstructed into a distributed optimization scheduling model.

[0155] In the embodiment of the present application, the consensus alternating direction multiplier algorithm is used to reconstruct the flexible resource collaborative scheduling model that takes into account the distributed robust opportunity constraints of wind power uncertainty into a distributed optimization scheduling model. The specific steps include:

[0156] (1) The variables of the flexible resource collaborative scheduling model taking into account the uncertainty distribution and opportunity constraints of wind power are sorted out, and the node active power balance constraints and line transmission power constraints are rewritten.

[0157] Specifically, the model variables are rearranged. The local private model includes equations (12) and (17)-(20), and the shared coupling model is rewritten from equations (15) and (16) as follows:

[0158]

[0159] in, Indicates adjustable power; Indicates non-adjustable power; in N B Indicates the total number of nodes; A i =[χ i ,-B i ,0], where Where E represents the system's correlation matrix, column vector and Each element represents P ij,max / X ij and P ij,min / X ij .

[0160] (2) According to the shared variable x within the scheduling period T i =[x i,1 ,...,x i,T ], divide the system into regions:

[0161]

[0162] Among them, f total,nrepresents the total cost of the nth sub-region; N sub Indicates the total number of sub-regions; D n Represents the set of nodes in the nth sub-region.

[0163] The Lagrangian dual function of subregion n is:

[0164]

[0165] Among them, λ y and μ y They represent the dual variables of the rewritten equations of the node active power balance constraint and the line transmission power constraint; x i ∈φ i It represents the constraints that satisfy the distributed robust opportunity constraints, thermal power unit operation constraints, energy storage system operation constraints, interruptible load aggregation constraints, and transferable load aggregation constraints.

[0166] (3) Set local variable z n To replace the consensus variable y = [λ y ,μ y ], the Lagrangian dual function (24) of subregion n corresponds to the Lagrangian dual problem:

[0167]

[0168] (4) The consensus alternating direction multiplier method (C-ADMM) is used to iteratively solve the Lagrangian dual problem (25), and the variables at the k+1th iteration are expressed as:

[0169]

[0170] Among them, := is the assignment symbol; δ n represents the Lagrange multiplier corresponding to the equality constraint in the subregion n of the Lagrange dual problem, represents the value solved at the kth iteration; ρ represents the given penalty factor.

[0171] (5) The global optimization scheduling model is transformed into regional sub-problems that can be solved in parallel. The sub-region scheduling model is expressed as:

[0172]

[0173] Among them, τ is an auxiliary variable; and They are The subvectors corresponding to λ and μ in ; and y k+1 The subvectors in that correspond to λ and μ.

[0174] Step S500: Improve the consensus alternating direction multiplier method through the branch and bound method, handle the discrete variables of adjustable loads, solve the distributed optimization scheduling model, and obtain a flexible resource distributed optimization scheduling strategy considering the uncertainty of wind power.

[0175] In the embodiment of the present application, the consensus alternating direction multiplier method is improved through the branch and bound method, the discrete variables of adjustable loads are processed, the distributed optimization scheduling model is solved, and a flexible resource distributed optimization scheduling strategy considering the uncertainty of wind power is obtained. The specific steps include:

[0176] (1) For each sub-region after partitioning, relax the integer constraints (21) including the operating state discrete variables of interruptible loads and shiftable loads and to convert the original mixed integer linear programming model into a linear programming model:

[0177]

[0178] The linear programming model is convex. If the optimal solutions and are both 0 or 1, output; otherwise, execute step (2);

[0179] (2) Arbitrarily select and a non-zero element γ i,t to perform branching and reconstruct two LP problems:

[0180] LP1:

[0181]

[0182] LP2:

[0183]

[0184] (3) Solve the LP1 and LP2 problems with the added new constraints respectively and obtain the optimal values L1 and L2. The corresponding solutions are and According to different cases of the solutions, it is subdivided into:

[0185] If both γ1 and γ2 are 0 or 1, then min(L1, L2) is the value of the objective function to be found;

[0186] If both γ1 and γ2 are decimal numbers, select the variable corresponding to min(L1, L2) to continue branching downward and execute step (2);

[0187] If γ1 is a decimal number and γ2 is an integer, if L1 > L2, then γ2 is the optimal solution of the sub-region scheduling model; if L1 < L2, then continue to branch downward for γ2 and execute step (2);

[0188] The above branch process stops calculating when a feasible solution is found or no feasible solution is found, otherwise it goes to step (4);

[0189] (4) and The value of is substituted into the regional subproblem that can be solved in parallel, and the predefined tolerance parameter If the convergence criterion is not satisfied Then return to step (1); if satisfied, output the optimal solution.

[0190] The present invention defines the prediction error fuzzy set based on the Wasserstein distance, and approximately quantifies and evaluates the complex uncertainty of wind power prediction error; based on the distributed robust opportunity constraint, the conditional risk value is introduced to linearize it, and the impact of wind power output uncertainty on scheduling is characterized; considering the differentiated operating characteristics of multiple flexible resources such as energy storage systems, electric vehicles and adjustable loads, a flexible resource collaborative scheduling model taking into account the distributed robust opportunity constraint of wind power uncertainty is constructed; based on the consensus alternating direction multiplier algorithm, the proposed model is reconstructed into a distributed optimization scheduling model; based on the improved consensus alternating direction multiplier method of the branch and bound method, the discrete variables of the adjustable load are processed, the proposed model is solved and a flexible resource distributed optimization scheduling strategy taking into account the uncertainty of wind power is obtained.

[0191] To verify the feasibility of the above-mentioned embodiment method, Table 1 shows the total cost comparison of wind power prediction error samples under four different distributions. The results show that the proposed model has good applicability when processing wind power prediction error data samples under different distributions. It does not require the precise probability distribution of wind power prediction errors to be obtained in advance, and can fully mine the distribution information of random variables based on limited data samples.

[0192] Table 1 Comparison of total costs under different distributions of wind power forecast error samples

[0193]

[0194] To verify the feasibility of the above-mentioned embodiment method, Figure 5 Convergence comparison of C-ADMM before and after improvement. Figure 5 (a) It can be seen that when solving non-convex problems, the traditional C-ADMM has y It decreases with the increase of the number of iterations, but there is volatility in the process of iteration and the convergence stability is poor. The local enlarged image shows that in the 750th to 800th iteration interval, σ y It fluctuates around 1.14, and the optimal solution cannot be stabilized near a certain precise value. Figure 5 (b) It can be seen that the improved C-ADMM convexifies the constraints containing discrete variables through the branch and bound method, and then gradually approaches the optimal solution. yIt shows a smooth change trend during the entire iteration process and gradually tends to 0 after the 700th iteration. The convergence accuracy can be achieved after 1000 iterations. The local enlarged image shows that σ y The order of magnitude is 10 -2 , which is 1 / 100 of the traditional C-ADMM. This shows that the improved C-ADMM can improve the convergence stability when solving distributed optimization scheduling problems with discrete adjustable loads.

[0195] The following is an embodiment of a system for distributed optimization and scheduling of flexible resources that takes into account wind power uncertainty, which can be used to implement an embodiment of a method for distributed optimization and scheduling of flexible resources that takes into account wind power uncertainty, according to the present invention. For details not disclosed in the embodiment of the system for distributed optimization and scheduling of flexible resources that takes into account wind power uncertainty, please refer to the embodiment of the method for distributed optimization and scheduling of flexible resources that takes into account wind power uncertainty, according to the present invention.

[0196] In one embodiment, a flexible resource distributed optimization scheduling system taking into account wind power uncertainty is provided, the system comprising:

[0197] The fuzzy set module is used to define the fuzzy set of wind power forecast error based on Wasserstein distance and quantify the uncertainty of the probability distribution of wind power forecast error;

[0198] The distributed robustness chance constraint module is used to construct distributed robustness chance constraints based on the fuzzy set of wind power forecast errors and introduce the conditional value at risk to provide a convex approximation for distributed robustness chance constraints;

[0199] The dispatch model module is used to construct a flexible resource collaborative dispatch model based on distributed robust opportunity constraints based on linearized reconstruction, taking into account the differentiated operating characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads, and taking into account the distributed robust opportunity constraints of wind power uncertainty;

[0200] A distributed optimization scheduling model module is used to reconstruct the flexible resource collaborative scheduling model that takes into account the uncertainty distribution and robust opportunity constraints of wind power into a distributed optimization scheduling model using the consensus alternating direction multiplier algorithm;

[0201] The solution module is used to improve the consensus alternating direction multiplier method through the branch and bound method, process the discrete variables of adjustable load, solve the distributed optimization scheduling model, and obtain a flexible resource distributed optimization scheduling strategy taking into account the uncertainty of wind power.

[0202] The various functional modules in the embodiments of the present invention may be integrated into a processing module, or may be physically separate units, or two or more units may be integrated into a module. The above-mentioned integrated modules may be implemented in the form of hardware or software functional modules.

[0203] The above description is merely a preferred embodiment of the present invention and does not constitute any form of limitation to the present invention. Although the present invention has been disclosed as above in terms of a preferred embodiment, it is not intended to limit the present invention. Any technician familiar with this patent can make slight changes or modifications to equivalent embodiments using the above-mentioned technical contents without departing from the scope of the technical solution of the present invention. However, any simple modifications, equivalent changes and modifications made to the above embodiments based on the technical essence of the invention without departing from the content of the technical solution of the present invention are still within the scope of the present invention.

Claims

1. A flexible resource distributed optimization scheduling method taking into account wind power uncertainty, characterized by: The method comprises: The fuzzy set of wind power forecast error is defined based on Wasserstein distance to quantify the uncertainty of the probability distribution of wind power forecast error. Based on the fuzzy set of wind power forecast error, a distributed robustness chance constraint is constructed. The distributed robustness chance constraint is convexly approximated by introducing the conditional value at risk, and the distributed robustness chance constraint after convex approximation is linearized. Based on the distributed robustness opportunity constraints of linear processing, and considering the differentiated operating characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads, a flexible resource collaborative scheduling model is constructed that takes into account the distributed robustness opportunity constraints of wind power uncertainty. The consensus alternating direction multiplier algorithm is used to reconstruct the flexible resource collaborative scheduling model with distributed robust opportunity constraints taking into account wind power uncertainty into a distributed optimization scheduling model. The consensus alternating direction multiplier method is improved by the branch and bound method to process the discrete variables of adjustable loads, and the distributed optimization scheduling model is solved to obtain a flexible resource distributed optimization scheduling strategy taking into account the uncertainty of wind power.

2. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 1 is characterized in that: The method of defining a prediction error fuzzy set based on the Wasserstein distance and quantifying the uncertainty of the probability distribution of the prediction error includes the following specific steps: The predicted power of wind farm i in period t Considered as a random variable ξ, a collection of N samples Defining the empirical probability distribution for: in, represents the lth sample of the random variable The Dirac measure of Define two probability distributions P and The Wasserstein distance d between w for: where inf(·) represents the lower bound; Ξ∈R k represents the support set of the random variable ξ; ‖·‖ p Represents R k The p-norm on , takes the 1-norm; Φ represents ξ and On marginal probability distribution P and empirical probability distribution The joint probability distribution of Based on two probability distributions P and The Wasserstein distance d between w , define the wind power prediction error fuzzy set Y as: Where P(Ξ) represents the set of all probability distributions in the support set; ε represents the radius of the Wasserstein sphere.

3. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 1 or 2, characterized in that: The method of constructing a distributed robustness chance constraint based on a fuzzy set of wind power prediction errors, introducing a conditional value at risk to convexly approximate the distributed robustness chance constraint, and linearizing the distributed robustness chance constraint after the convex approximation, specifically includes the following steps: According to the actual output and predicted output of each wind farm i in period t, the wind power output constraint is obtained; Based on the fuzzy set Y of wind power forecast error, the wind power output constraint is rewritten as a distributed robust chance constraint under the confidence level 1-η; The conditional value at risk is introduced to make a convex approximation of the distributed robustness chance constraint, and the distributed robustness chance constraint after the conditional value at risk is obtained by combining the saddle point theorem. Take the polyhedron set Ξ={ξ∈R N :Hξ≤d}, combined with the maximum combination of two linear functions, the distributed robust chance constraint after convex approximation is linearized.

4. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 3 is characterized in that: Assuming H = -I, d = 0, the linearized distribution of the robust chance constraint is expressed as: Among them, λ t , π t,l , γ t,lj represents the dual variable introduced in the transformation process, Represents γ t,lj The transpose of Represents the candidate value of the risk threshold, and we need to find the optimal β t To calculate the conditional value at risk; γ t,lj is a column vector of the same dimension as d; express The lth sample of t N represents the tth component of 1 t dimensional standard basis vectors; P t W represents the actual output of the wind farm in period t; η represents the tolerance for constraint violation; T represents the total number of periods; Indicates a quantifier.

5. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 4 is characterized in that: The distributed robust opportunity constraint based on linear processing considers the differentiated operating characteristics of multiple flexible resources including energy storage systems, electric vehicles, and adjustable loads, and constructs a flexible resource collaborative scheduling model that takes into account the distributed robust opportunity constraint of wind power uncertainty. The specific steps include: Construct the objective function based on the system comprehensive operating cost and the minimum conditional risk value cost; Combined with Monte Carlo sampling simulation, the charging load of electric vehicles in autonomous charging mode is calculated; Establish constraints for a multi-resource collaborative scheduling model; the constraints of the multi-resource collaborative scheduling model include: linearized distributed blue-robust opportunity constraints, node active power balance constraints, line transmission capacity constraints, thermal power unit operation constraints, energy storage system operation constraints, interruptible load aggregate constraints, and transferable load aggregate constraints.

6. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 5, characterized in that: The objective function is: minf total =minC G +C W +C ESS +C DR +f CVaR Among them, C G 、C W 、C ESS 、C IL 、C RL and f CVaR They represent the fuel cost of thermal power units, the cost of wind farm wind curtailment, the operating cost of energy storage systems, the compensation cost for interruptible loads, the compensation cost for transferable loads and the value-at-risk cost under the worst-case conditions; N G Indicates the total number of thermal power units; represents the output of thermal power unit i during period t; and Represent the fuel cost coefficient of thermal power unit i; N W represents the total number of wind farms; represents the unit wind curtailment penalty cost of wind farm i in period t; represents the actual output of wind farm i during period t; N ESS represents the total number of energy storage systems; represents the unit operating cost of energy storage i; represents the capacity of energy storage i in period t; N IL and N RL Represents the total number of interruptible load and transferable load aggregates respectively; c IL and c RL They represent the unit regulation power compensation cost of interruptible load and transferable load aggregate respectively; and They represent the regulation power of interruptible load and transferable load aggregate i in period t respectively.

7. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 5, characterized in that: The charging load model of electric vehicles in autonomous charging mode is: Among them, f EV (d) represents the lognormal distribution probability density function of the EV daily mileage d; μ1 and σ1 represent the expected value and standard deviation of the daily mileage function, respectively; f EV (t0) represents the normal distribution probability density function of the EV daily driving end time t0; μ2 and σ2 represent the expected value and standard deviation of the daily driving end time function, respectively; T i EV,ch Indicates the charging duration of EVi; Indicates the charging start time of the EVi, that is, the end time of driving for the day; d i Indicates the daily mileage of EVi; Indicates the power consumption of EVi per 100 kilometers; P i EV,ch Indicates the charging power of EVi; Indicates the charging efficiency of EVi; Indicates the charging end time of EVi; P t EV,chload represents the total charging power at time t; N EV Indicates the total number of EVs; represents the charging power of EVi at time t.

8. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 5, characterized in that: The consensus alternating direction multiplier algorithm is used to reconstruct the flexible resource collaborative scheduling model with distributed robust opportunity constraints taking into account wind power uncertainty into a distributed optimization scheduling model. The specific steps include: Sort out the variables of the flexible resource collaborative scheduling model considering the distribution robust chance constraint of wind power uncertainty, and rewrite the nodal active power balance constraint and line transmission power constraint as follows: in, Indicates adjustable power; Indicates non-adjustable power; in N B Indicates the total number of nodes; A i =[χ i ,-B i ,0], where Where E represents the system's correlation matrix, column vector and Each element represents P ij,max / X ij and P ij,min / X ij ; According to the shared variable x in the scheduling period T i =[x i,1 ,...,x i,T ], divide the system into regions: Among them, f total,n represents the total cost of the nth sub-region; N sub Indicates the total number of sub-regions; D n Represents the set of nodes in the nth sub-region; The Lagrangian dual function of sub-region n is: Among them, λ y and μ y They represent the dual variables of the rewritten equations of the node active power balance constraint and the line transmission power constraint; x i ∈φ i Represents x i The constraint set φ that satisfies the distributed blue stick opportunity constraint, thermal power unit operation constraint, energy storage system operation constraint, interruptible load aggregation constraint, and transferable load aggregation constraint i ; Set local variable z n To replace the consensus variable y = [λ y ,μ y ], the Lagrangian dual problem corresponding to the Lagrangian dual function of subregion n is: Use the alternating direction method of multipliers with consensus to iteratively solve the Lagrangian dual problem, and express each variable at the (k + 1)-th iteration as: Among them, := is the assignment symbol; δ n represents the Lagrange multiplier corresponding to the equality constraint in the subregion n of the Lagrange dual problem, represents the value solved at the kth iteration; ρ represents the given penalty factor; The global optimal scheduling model is transformed into a region sub-problem that can be solved in parallel, and the sub-region scheduling model is expressed as: Among them, τ is an auxiliary variable; and They are The subvectors corresponding to λ and μ in ; and y k+1 The subvectors in that correspond to λ and μ.

9. The method for distributed optimization and scheduling of flexible resources taking into account wind power uncertainty according to claim 8, characterized in that: Improve the alternating direction method of multipliers with consensus by the branch and bound method to handle the discrete variables of adjustable loads, and solve the distributed optimization scheduling model to obtain the flexible resource distributed optimization scheduling strategy considering wind power uncertainty, which specifically includes the following steps: (1) For each sub-area after division, the relaxation includes the discrete variables of the interruptible load and the transferable load operation state. and The node power conservation constraint is used to transform the sub-area scheduling model into a linear programming model: The linear programming model is convex, if the optimal solution and If both are 0 or 1, then output; otherwise, go to step (2); (2) Optional and A non-zero element in γ i,t Branch out and reconstruct the two LP problems: LP1: LP2: (3) Solve the LP1 and LP2 problems with the new constraints and obtain the optimal values ​​L1 and L2. The corresponding solutions are and According to different solutions, it is divided into: If both γ1 and γ2 are 0 or 1, then min(L1, L2) is the value of the objective function sought; If both γ1 and γ2 are decimals, then select the variable corresponding to min(L1, L2) and continue to branch downward, and execute step (2); If γ1 is a decimal and γ2 is an integer, if L1 > L2, then γ2 is the optimal solution of the sub-region scheduling model; if L1 < L2, then continue to branch downward for γ2 and execute step (2); The above branching process stops calculating when a feasible solution is found or there is no feasible solution, otherwise enter step (4); (4) and The value of is substituted into the regional subproblem that can be solved in parallel, and the predefined tolerance parameter If the convergence criterion is not satisfied Then return to step (1); if satisfied, output the optimal solution.

10. A flexible resource distributed optimization scheduling system taking into account wind power uncertainty, characterized in that: The system includes: A fuzzy set module for defining a fuzzy set of wind power prediction errors based on the Wasserstein distance to quantify the probability distribution uncertainty of wind power prediction errors; A distribution robust chance constraint module for constructing a distribution robust chance constraint according to the fuzzy set of wind power prediction errors and introducing conditional value at risk for convex approximation of the distribution robust chance constraint; A scheduling model module for constructing a flexible resource collaborative scheduling model considering the distribution robust chance constraint of wind power uncertainty based on the linearly reconstructed distribution robust chance constraint and considering the differentiated operation characteristics of multiple flexible resources such as energy storage systems, electric vehicles, and adjustable loads; A distributed optimization scheduling model module for reconstructing the flexible resource collaborative scheduling model considering the distribution robust chance constraint of wind power uncertainty into a distributed optimization scheduling model using the alternating direction method of multipliers with consensus; A solving module for improving the alternating direction method of multipliers with consensus by the branch and bound method to handle the discrete variables of adjustable loads and solving the distributed optimization scheduling model to obtain the flexible resource distributed optimization scheduling strategy considering wind power uncertainty.

Images