A multi-constraint and non-linear target-oriented energy storage system optimization configuration method

Abstract

The application is a kind of energy storage system optimization configuration method for multi-constraint and nonlinear target, which is used for optimizing energy storage system configuration scheme in different application scenarios.The method of the application comprises: obtaining the power generation history data of the wind farm in the scene and the type of the energy storage system, establishing the nonlinear objective function and the constraint condition of the energy storage system optimization configuration, and solving the energy storage system optimization configuration problem through an optimization module; the optimization module first generates a unified optimization problem model, pre-selects a swarm intelligence optimization algorithm and a traditional optimization algorithm, in the iteration process of the swarm intelligence algorithm, forms a linear or quadratic term approximation near the individual by means of Taylor series expansion, calculates the optimal solution near the individual to represent the behavior of the individual by using the traditional optimization algorithm, carries out global search optimization by using the swarm intelligence algorithm, and finally outputs the optimal energy storage system configuration scheme.The application can meet the configuration optimization demand of the energy storage system in different scenes, and can quickly obtain a better energy storage system configuration scheme.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of energy storage system planning configuration, in particular to a kind of energy storage system optimization configuration method for multiple constraints and nonlinear target. BACKGROUND

[0002] As a key link of new power system construction, energy storage system is widely used to solve the problem of new energy uncertainty under high penetration rate. The application cost of energy storage system changes with its planning capacity. How to scientifically and effectively optimize the capacity of energy storage system to meet the application demand of new energy has become the primary problem of further development of energy storage system.

[0003] The existing research on energy storage system optimization configuration finds that the optimization problem of energy storage system often shows the following two characteristics: first, the objective function of the optimization problem often shows nonlinear characteristics, and the nonlinear performance is not limited to quadratic polynomial, which causes difficulty in solving optimization calculation; second, the constraint conditions of the optimization problem are numerous and show the form of a combination of a large number of equalities and inequalities, which makes it difficult to simply describe the feasible region of variables. At present, the common optimization methods for solving such energy storage system optimization configuration problems with multiple constraints and nonlinear objective functions mainly include two types: traditional optimization algorithms and swarm intelligence optimization algorithms. Traditional optimization algorithms such as interior point method, gradient descent method, and Lagrange multiplier method often use Lagrange multiplier method to transform the original problem into an unconstrained problem, and then use gradient to optimize the calculation. However, in the optimization of nonlinear objective functions, the initial configuration scheme of the solution has a great influence on the optimization result, i.e. strong local search ability but weak global search ability, which easily forms a suboptimal configuration scheme of the energy storage system. Swarm intelligence optimization algorithms such as genetic swarm intelligence optimization algorithm, particle swarm algorithm, and artificial bee colony algorithm do not need to calculate the gradient of the objective function, so they have great advantages in solving nonlinear objective functions without constraints, and have strong global search ability. However, in the optimization problem of energy storage system with multiple constraints, the complex feasible region easily causes the energy storage system configuration scheme represented by the population individuals after iteration to be an infeasible solution of the optimization problem, which further affects the iterative evolution of the swarm intelligence algorithm. Even if a penalty function is used, the penalty coefficient will also have a great influence on the optimization configuration result of the energy storage system. Therefore, for the above-mentioned energy storage system optimization configuration problem with multiple constraints and nonlinear objective functions, how to establish an effective method to solve the energy storage system optimization configuration problem with constraints and nonlinear objective functions is a difficult problem. SUMMARY

[0004] For the energy storage system planning configuration problem in different application scenarios, the present application provides a kind of energy storage system optimization configuration method for multiple constraints and nonlinear target, solves the solution of the optimization configuration problem with multiple constraints and complex nonlinear objective functions in the application of energy storage system, so as to obtain a more optimal configuration scheme of energy storage system.

[0005] The method comprises the following steps:

[0006] Step 1, obtaining the historical data of the power generation of the wind farm in the scene, determining the type of the energy storage system serving the wind farm, designing a nonlinear objective function of the energy storage system optimization configuration, and establishing constraint conditions to obtain a mathematical model of the energy storage system optimization configuration;

[0007] According to the annual power generation data of a wind farm, the typical day wind power data of the wind farm is obtained by using a scene generation method such as a clustering algorithm, the type of the energy storage system serving the wind farm is determined, the nonlinear objective function of the energy storage system optimization configuration is designed according to the grid-connected power benefit of the wind farm combined with the energy storage system, the cost of the energy storage system and the power generation cost of the wind farm, and meanwhile, the equality and inequality relations of the wind farm combined with the energy storage system in the application process are established according to the requirements of power balance, energy storage energy state balance and grid-connected power fluctuation, a series of constraint conditions are formed, and then the mathematical model of the energy storage system optimization configuration scene for improving the economic benefit of the wind farm is established;

[0008] Step 2, setting an optimization module in the computer, which is used to solve the above mathematical model. First, the optimization module generates a unified optimization problem model according to the mathematical model of the energy storage system optimization configuration established in step 1, as follows:

[0009]

[0010] Wherein, X represents the optimization variable vector, the objective function J is the nonlinear relation function f established in step 1 about the optimization variable vector, and corresponds to the benefit model of the energy storage system; A eq and B eq respectively represent the coefficient matrix and the constant matrix of the equality constraint; A in and B in respectively represent the coefficient matrix and the constant matrix of the inequality constraint; C represents the constant coefficient matrix of the optimization variable range;

[0011] Step 3, pre-selecting a swarm intelligence optimization algorithm in the optimization module, randomly initializing the population individuals under the constraint condition of only considering the optimization variable range of the optimization problem, and each individual is represented as a vector of binary coding splicing of all optimization variables;

[0012] Step 4, the optimization module executes the iteration of the swarm intelligence optimization algorithm, and obtains a linear objective function or a quadratic objective function in the neighborhood of each individual in the gth generation population by using Taylor series expansion, to form N independent linear or quadratic optimization problems with constraints; wherein, the initial value of g is 1, and N represents the number of population individuals.

[0013] Optimization variable of the i-th individual in the g-th generation population In its neighborhood ε i The linear objective function using Taylor series expansion is as follows:

[0014]

[0015] The quadratic objective function is as follows:

[0016]

[0017] Where, and respectively represent the first derivative and the second derivative of the nonlinear function f at , O m represents the m-th order infinitesimal under Taylor series expansion; the upper index T represents transposition;

[0018] According to the requirements of the application scenario, the linear objective function or the quadratic objective function is selected to continue the following steps;

[0019] Step 5: In the optimization module, a traditional optimization algorithm is pre-selected, the linear objective function or the quadratic objective function with constraints in the neighborhood of each individual in the g-th generation population obtained in step 4 is solved by using the selected traditional optimization algorithm, and the optimal solution in the neighborhood is obtained as the fitness function value of the individual; the current optimal individual is selected from the N individuals, and compared with the historical optimal individual X global , and it is judged whether the historical optimal individual is updated or not; then the population iteration number g is increased by 1, and it is judged whether g is greater than the set maximum iteration number or not, if not, step 6 is executed, and if yes, step 7 is executed;

[0020] Wherein, if there is no solution when the traditional optimization algorithm is used to solve the objective function, it means that the individual does not meet the constraint condition, and then the individual is randomly initialized again, and the individual is returned to step 4 for execution;

[0021] Step 6: The optimization module continues to execute the genetic optimization algorithm, updates the N individuals of the new generation, and then returns to step 4 for execution;

[0022] Step 7: The solving process of steps 4 and 5 is used to calculate the optimal solution in the neighborhood of the historical optimal individual X global According to the optimization variable vector, an optimized configuration scheme of the energy storage system is obtained.

[0023] ​The embodiment of the present application provides an optimization configuration scene of an energy storage system in the promotion of economic benefits of a wind farm, and represents swarm intelligence optimization algorithm as an optimization solving process of a traditional optimization algorithm, and the Lagrange multiplier method is a traditional optimization algorithm, but the present application does not limit the types of swarm intelligence optimization and traditional optimization algorithm in the optimization solving process. The optimization variable vector in the embodiment of the present application contains the rated power, rated capacity, output power at each moment and wind farm output power control variable of the energy storage system.

[0024] In step 5, the selected traditional optimization algorithm is the Lagrange multiplier method, and the Lagrange function is constructed based on the linear objective function or the quadratic objective function in step 4 converted into an unconstrained objective function as follows:

[0025] Or

[0026]

[0027] Wherein, λ k and respectively represent the Lagrange coefficient of the kth equality constraint and the st inequality constraint, and constitute the Lagrange coefficient vector λ and h k (X) and y s (X) respectively represent the kth equality constraint and the st inequality constraint in the mathematical model of the linear objective function or the quadratic objective function in step 4, K is the number of equality constraints, and S is the number of inequality constraints.

[0028] The Lagrange function is used to optimize the variables X and the Lagrange coefficient vectors λ and The partial derivative is solved, and the derivative function is zero to form a system of equations, and the optimization variables of the ith individual in the gth generation population are solved by simultaneously solving the system of equations The optimal solution of the neighborhood and the corresponding function value The function value is the fitness function value of the individual in the genetic optimization algorithm, if the system of equations has no solution, it indicates that the individual cannot meet the constraint condition, and the individual is randomly initialized again, and then step 4 is executed again.

[0029] The fitness function values of the N individuals in the gth generation population are calculated, and the optimal individual is selected by sorting, if the fitness function value of the selected optimal individual is better than that of the historical optimal individual, the historical optimal individual is updated by using the optimal individual selected in the gth generation, otherwise, no updating is performed.

[0030] Compared with the prior art, the application has the advantages and positive effects that:

[0031] (1) The method of the application combines the characteristics of swarm intelligence optimization algorithm and traditional optimization algorithm when solving the optimization problem of multi-constrained and nonlinear target of energy storage system optimization configuration. The linearization or quadratic process of the nonlinear objective function is realized locally by means of Taylor series for swarm intelligence individuals, and the local search advantage of the traditional optimization algorithm is exerted under the premise of considering neglecting the error of high-order terms. At the same time, the global solution of the optimization problem is realized by means of the iterative optimization characteristics of the swarm intelligence algorithm, and the optimal energy storage system optimization configuration can be obtained more efficiently.

[0032] (2) The technical means for solving the optimization problem of energy storage system optimization configuration adopted by the method of the application is different from the existing swarm intelligence algorithm. In the optimization iteration process of the energy storage system configuration scheme, that is, the individual of the swarm intelligence algorithm, the process of handling the energy storage system optimization configuration constraint by using the penalty function is cancelled, and the multi-constraint condition is transferred to the traditional optimization algorithm after the linearization or quadratic of the nonlinear objective function, which solves the technical bottleneck that the swarm intelligence algorithm is difficult to handle the optimization problem with multiple constraints, and only the global search capability of the swarm intelligence algorithm is exerted. Different from the traditional gradient optimization calculation method, after forming multiple initial configuration schemes of the energy storage system by using a plurality of individuals of the swarm intelligence algorithm population, the linearization or quadratic optimization problem of the nonlinear optimization problem is solved by using Taylor series expansion, and multiple optimization configuration schemes are used to describe the fitness of the individual in the swarm intelligence algorithm, and then the global optimization iteration effect of the swarm intelligence algorithm is exerted, and the problem that the single initial energy storage system configuration scheme is easily trapped in the local optimal energy storage system configuration result in the traditional optimization algorithm and the difficulty in selecting the initial configuration scheme of the energy storage system is reduced. Therefore, the method of the application inherits the excellent global search capability of the swarm intelligence algorithm and the strong local search capability of the traditional optimization algorithm, and breaks through the technical bottlenecks that the swarm intelligence algorithm is difficult to handle the optimization with multiple constraints and the traditional optimization algorithm cannot handle the complex nonlinear optimization problem and is sensitive to the initial configuration scheme. Therefore, compared with the prior art, the method of the application can obtain a more optimal energy storage system optimization configuration result more quickly.

[0033] (3) When solving the optimization problem of multi-constrained and nonlinear target of energy storage system optimization configuration, the method of the application does not limit specific swarm intelligence algorithm and traditional optimization algorithm, and can be combined and applied according to the advantages of each algorithm to maximize the characteristics of each algorithm, so that the configuration demand of the energy storage system in multiple application scenarios can be met. BRIEF DESCRIPTION OF DRAWINGS

[0034] Figure 1 is the flowchart of the energy storage system optimization configuration method of the application for multi-constrained and nonlinear target;

[0035] Figure 2 is a local optimal solution flowchart based on the Lagrange multiplier method in the neighborhood of a certain individual of a population at a certain iteration number in the embodiment of the present application;

[0036] Figure 3 is a global optimal solution flowchart of the genetic optimization algorithm used in the embodiment of the present application. DETAILED DESCRIPTION

[0037] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below by taking the optimization configuration scenario of the energy storage system for improving the economic benefits of a wind farm as an example, taking the genetic swarm intelligence optimization algorithm as a representative swarm intelligence optimization algorithm and the Lagrange multiplier method as a traditional optimization algorithm, and taking the combination of the two as an example, in combination with the accompanying drawings.

[0038] The energy storage system optimization configuration method for multiple constraints and nonlinear objectives provided by the present application is a new method combining the two to solve the energy storage system optimization configuration problem containing multiple constraints and nonlinear objectives, which is proposed from the characteristics of swarm intelligence algorithms and traditional optimization algorithms. The method can reduce the restriction of multiple constraint conditions on the global search ability of swarm intelligence algorithms in the energy storage system optimization configuration problem, and solve the dependence of the optimization configuration result on the initial configuration scheme of the energy storage system when the traditional optimization algorithm solves the nonlinear optimization, thereby realizing the optimal configuration of the energy storage system. It is worth noting that the present application is not a simple combination of the two existing optimization algorithms, and its characteristics lie in the advantages of swarm intelligence algorithm iteration optimization and traditional algorithm in linear or quadratic optimization, the formation of linear or quadratic term approximation near the individual of the swarm intelligence algorithm by means of Taylor series expansion, the calculation of the optimal solution near the individual by the traditional algorithm to represent the behavior of the individual, and the global search optimization by the swarm intelligence algorithm to solve the energy storage system optimization configuration problem containing multiple constraints and complex nonlinear objective functions. In addition, the present application does not limit the form of the swarm intelligence algorithm and the traditional optimization algorithm, and can be combined according to the advantages of each algorithm to maximize the characteristics of each algorithm and meet the configuration optimization requirements of the energy storage system in different scenarios.

[0039] The embodiment of the present application is in the optimization configuration scenario of the energy storage system for improving the economic benefits of a wind farm, and takes the classical genetic swarm intelligence optimization algorithm and the Lagrange multiplier method as an example to realize the energy storage system optimization configuration method for multiple constraints and nonlinear objectives, the overall process of which is shown in Figure 1 The following seven steps are described.

[0040] Step 1: According to the historical data of the wind farm and the type of the energy storage system, a mathematical model for optimizing the capacity configuration of the energy storage system for improving the economic benefits of the wind farm is established.

[0041] Step 1.1, according to the annual power generation data of a certain wind farm, using a scene generation method such as clustering algorithm, the wind power generation data P of a typical day of a certain wind farm is obtained w , the wind power sampling time interval is T s , unit: min, then the data amount Len of a day is 1440 / T s . Let the wind power generation cost be C w , and the wind power generation income of the wind farm combined with the energy storage system be C g , define the maximum wind power at time t as P w (t), the ratio of the actual wind power to the maximum wind power as λ(t), the energy storage output power as P ESS (t), the wind storage system grid-connected power as P g (t) = λ(t) × P w (t) + P ESS (t), the wind farm combined with the energy storage system grid-connected power generation income C g (t), t = 1, 2, …, Len.

[0042] Step 1.2, establish the objective function of the capacity optimization configuration of the energy storage system for improving the economic benefit of the wind farm, let the rated power and rated capacity of the energy storage system be P ESS,e and S ESS,e , the unit power and capacity construction cost of the energy storage system be α and β, the operation and maintenance cost of the energy storage system be proportional to the rated capacity cost of the energy storage system, the proportional coefficient be M, the service life of the energy storage system be T, the discount rate be γ, the upper and lower limits of the energy storage system capacity state be SOE u and SOE d , the efficiency be η, and the annual average degradation rate be θ, then the cost O ESS of the energy storage system under the typical day is:

[0043]

[0044] Among them, the rated power P ESS,e , the rated capacity S ESS,e , the energy storage system output P ESS (t) and the ratio of the actual wind power to the maximum wind power λ(t) are optimization variables, and the other related parameters can be obtained by consulting the information after selecting the type of energy storage system.

[0045] The grid-connected power generation income of the wind farm combined with the energy storage system, the energy storage system cost and the wind power generation cost under the typical day can be calculated, and the income model of the wind farm combined with the energy storage system under the typical day can be represented, and a nonlinear objective function J to be optimized is formed, which is shown as follows, and it is obvious that the objective function shows nonlinear characteristics;

[0046]

[0047] Step 1.3, considering the power balance and energy state equation constraints of the wind farm joint energy storage system during operation, grid-connected power fluctuation range, energy storage system energy state range, energy storage system output power range and other inequality constraints, a mathematical expression of the constraint condition of the energy storage system capacity optimization configuration for improving the economic benefit of the wind farm is established, as shown below:

[0048] P g (t)=P ESS (t-1)+λ(t)×P w (t) (3)

[0049]

[0050] SOE ESS (1)=SOE ESS (Len) (5)

[0051] |ΔP g (t)|=|P g (t)-P g (t-1)|≤δ (6)

[0052] 0≤λ(t)≤1 (7)

[0053] |P ESS (t)|≤P ESS,e (8)

[0054] SOE d ≤SOE ESS (t)≤SOE u (9)

[0055] Wherein, SOE ESS (t) represents the energy state of the energy storage system at time t, formula (5) represents the energy state of the energy storage system at the first time and the last time of each day; ΔP g (t) represents the difference between the current time t and the previous time wind storage system grid-connected power, and δ is the maximum difference threshold of the wind storage system grid-connected power between the previous and the current time.

[0056] Step 1.4, combining the nonlinear objective function established in step 1.2 and the multiple equality and inequality constraints shown in step 1.3, a mathematical model of the energy storage system optimization configuration scenario for improving the economic benefit of the wind farm is established.

[0057] Step 2: The method of the present invention sets up an optimization module in the computer. In this module, the constraints and objective function in the scenario of optimizing the configuration of energy storage system to improve the economic benefits of wind farms established in Step 1 are first transformed into a unified optimization problem mathematical model containing multiple constraints and a nonlinear objective function J. Its mathematical expression can be expressed as follows:

[0058]

[0059] Where X = [x1, x2, ..., x n ] T It represents n optimization variables x i The vector of optimization variables is composed of parameters that the energy storage system needs to optimize. Here, the superscript T indicates transpose. In this embodiment of the invention, the optimization variables include the rated power P of the energy storage system. ESS,e Rated capacity S ESS,e Energy storage system output P ESS (t) and the ratio of actual wind power generation to maximum wind power generation λ(t); f(·) represents the objective function, which is expressed as a nonlinear relationship of the optimization variable vector, as shown in formulas (1) and (2) of this invention; A eq and B eq Let A represent the coefficient matrix and constant matrix of the equality constraint, respectively. The equality constraint in this embodiment of the invention is shown in formulas (3) to (5); in and B in Let C represent the coefficient matrix and constant matrix of the inequality constraints, respectively. The inequality constraints in this embodiment of the invention are shown in formula (6). C represents the constant coefficient matrix of the optimization variable range, as shown in formulas (7) to (9) in this embodiment of the invention. Existing data shows that any optimization problem can be abstracted into the above mathematical model.

[0060] Step 3: The optimization module selects a swarm intelligence optimization algorithm, initializes the parameters of the selected algorithm, and initializes the population individuals under the constraint of only considering the range of optimization variables in the optimization problem. Multiple swarm intelligence optimization algorithms and multiple traditional optimization algorithms can be pre-set in the optimization module for use in different application scenarios.

[0061] The method of this invention addresses the optimization problem in step 2. Based on the characteristics of different swarm intelligence algorithms, a swarm intelligence optimization algorithm is selected, and the relevant parameters of the selected algorithm are set. Only the range constraint of the independent variable in the original optimization problem needs to be considered, i.e., X≤C. The initial population is randomly initialized. In this embodiment, a genetic optimization algorithm is selected, and initialization is performed as follows: steps 3.1 and 3.2.

[0062] Step 3.1: Set the relevant parameters in the genetic optimization algorithm, such as the number of individuals N in the population, the individual encoding length L, the maximum number of chromosome iterations G, and the crossover rate P.c , mutation rate P n and genetic operation mode, etc., let the iteration number g = 1;

[0063] Step 3.2, randomly initialize the gth generation population individuals without considering the constraint condition limit, according to the individual coding length L defined by the solution accuracy, randomly generate N binary coding strings with length n x L, wherein each continuous L length binary coding bit represents an optimization variable, such as the jth optimization variable of the ith individual in the gth generation population characterized by the coding bits between (j-1) x L + 1 and j x L of the binary coding string.

[0064] Step 4, linearize or quadraticize the objective function J in the neighborhood of each individual in the gth generation population by Taylor series expansion, to form N independent linear or quadratic optimization problems with constraints. Take the optimization variables of the ith individual in the gth generation population as an example, in the neighborhood ε i , the linear objective function and the quadratic objective function by Taylor series expansion are as follows:

[0065]

[0066]

[0067] wherein, and respectively represent the first and second derivatives of the nonlinear function f at , the calculation formula is shown in formula (13), O m represents the mth infinitesimal under Taylor series expansion; and the newly added optimization variable range constraint in the constraint condition is the intersection of the original optimization problem and the neighborhood range constraint. In addition, the selection of the linear objective function and the quadratic objective function depends on the requirements of the application scenario for the solution speed and accuracy. If the application scenario has higher requirements for the calculation speed, the equivalent linear objective function is selected; if the calculation accuracy is required to be higher, the equivalent quadratic objective function is selected.

[0068]

[0069] Step 5, solve the mathematical model with linear or quadratic objective function and constraint condition in the equivalent neighborhood of each individual in the gth generation population equivalent in step 4 by using a traditional optimization algorithm to obtain the fitness function value of each individual. In the embodiment of the present application, the Lagrange multiplier method is taken as an example for description, and the specific process is shown in Figure 2 .

[0070] Step 5.1: Based on the mathematical model with linear or quadratic objective functions and constraints obtained in Step 4, construct the Lagrange function using the Lagrange multiplier method and KKT (Karush-Kuhn-Tucker) conditions. The mathematical model is transformed into an unconstrained objective function, in the following form:

[0071]

[0072]

[0073] Where, λ k and Let represent the Lagrange coefficients of the k-th equality constraint and the s-th inequality constraint, respectively, and form the Lagrange coefficient vector λ and λ'. h k (X) and y s (X) represent the k-th equality constraint and the s-th inequality constraint in the mathematical model obtained in step 4, respectively. In formula (15), K and S represent the number of equality constraints and the number of inequality constraints, respectively.

[0074] Using the above Lagrange function For the optimization variable X and the Lagrange coefficient vector λ and Find the partial derivatives and set the derivatives to zero to form the following system of multivariable equations:

[0075]

[0076] Solving the system of simultaneous equations yields the optimization variables for the i-th individual in the g-th generation of the population. optimal solution in the neighborhood and the corresponding function values Define this function value as the value of an individual in the genetic optimization algorithm. The fitness function value. If the system of equations has no solution, proceed to step 5.2; if the system of equations has a solution, proceed to step 5.3.

[0077] Step 5.2, the equation system has no solution, which means that in the original optimization problem in step 2, the current individual cannot meet the constraints. It is necessary to re-randomly initialize the individual and return to step 4 to re-linearize or quadratize the nonlinear objective function and solve the problem in step 5.1 until a feasible solution exists or the cutoff condition is met. The cutoff condition is set such as the number of random initializations of the individual reaching the set maximum value.

[0078] Step 5.3: For each of the N individuals in the g-th generation, the fitness function value is solved using Step 5.1. The fitness function values ​​of the N individuals in the g-th generation are sorted to select the optimal individual.

[0079] Determine the best individual X in this generation (g)* fitness function value Is it superior to the historical best individual? fitness function If yes, then replace the historical best individual and its corresponding fitness function value with the current generation's best individual and fitness function value; if no, then do nothing and proceed to step 5.4.

[0080] Step 5.4: Increment the iteration count by 1, i.e., g = g + 1, and determine whether the iteration count g is less than the maximum iteration count. If yes, proceed to step 6; otherwise, proceed to step 7.

[0081] Step 6: Utilize evolutionary operations such as selection, crossover, mutation, and re-insertion using a genetic algorithm. The implementation flow of the genetic optimization algorithm is as follows: Figure 3 As shown, after establishing a new generation of N individuals, return to step 4 to perform a new round of linearization or quadraticization and Lagrange optimization solution process.

[0082] The updating of specific individuals in the population is performed according to the selected genetic optimization algorithm, which will not be elaborated further in this invention. In embodiments of this invention, such as... Figure 3 As described above, when updating the new generation of individuals, based on the fitness function values ​​of each individual in the current generation, several optimal individuals are selected. Then, from the selected individuals, they are randomly paired and crossover is performed to generate an equal number of new individual codes. Mutation is then performed on the selected individuals. The new individual code strings obtained through crossover and mutation, together with the remaining uncrossed and unmutated code strings of several minimum function values, form a new generation of individual codes equal in number to the parent generation individuals.

[0083] Step 7, take the historical best individual X obtained in step 5.4 global Input the data into steps 4 and 5 of the solution process to calculate individual X. global optimal solution in the neighborhood Then, substitute the values ​​into the original objective function from step 2 to obtain the optimal value. The process of optimizing the nonlinear objective function with constraints is completed, thereby achieving the optimized configuration of the energy storage system in step 1 to improve the economic benefits of the wind farm.

[0084] like Figure 2 and 3 As shown, this invention utilizes swarm intelligence optimization algorithms and traditional optimization algorithms to continuously search for the globally optimal individual, ultimately obtaining the historically optimal individual X. global Then, based on steps 4 and 5, the optimal solution within the domain is calculated, obtaining the desired optimal combination of optimization variables. In this embodiment of the invention, based on the iterative solution of steps 3 to 6 above, the optimal combination of optimization variables is obtained, namely, the rated power P of the optimal energy storage system. ESS,e Rated capacity SESS,e , the output of the energy storage system P ESS (t) and the ratio of the actual wind power generation power to the maximum power generation power λ(t), the energy storage system in the wind farm is adjusted according to the result to achieve the purpose of improving the optimal configuration of the energy storage system.

[0085] Finally, it should be noted that: obviously, the described embodiments are only part of the embodiments of the present application, not all. The technical solutions of the present application are not limited to the types of swarm intelligence optimization and traditional optimization algorithms, and are not limited to the classical genetic swarm intelligence optimization algorithm and Lagrange multiplier method in the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor shall fall within the scope of protection of the present application.

Claims

1. A method for optimal configuration of energy storage system oriented to multiple constraints and nonlinear objectives, characterized in that, Comprising the following steps: Step 1, obtaining the historical data of power generation of the wind farm in the application scene, determining the type of energy storage system serving the wind farm, designing a nonlinear objective function of the optimal configuration of the energy storage system according to the grid-connected power benefit of the wind farm combined with the energy storage system, the cost of the energy storage system and the power generation cost of the wind farm, and establishing the equality and inequality relations of the operation process of the wind farm combined with the energy storage system according to the requirements of power balance, energy storage energy state balance and grid-connected power fluctuation in the application process of the wind farm combined with the energy storage system, forming the constraint conditions, and then establishing a mathematical model of the optimal configuration of the energy storage system; Step 2, setting an optimization module in the computer; the optimization module generates a unified optimization problem model according to the mathematical model of the optimal configuration of the energy storage system established in step 1, as follows: Wherein, X represents an optimization variable vector, including parameters that the energy storage system needs to optimize configuration, the objective function J is the nonlinear relationship function f established in step 1, which is represented as about the optimization variable vector, and corresponds to the income model of the energy storage system; A eq and B eq respectively represent the coefficient matrix and the constant matrix of the equality constraint; A in and B in respectively represent the coefficient matrix and the constant matrix of the inequality constraint; C represents the constant coefficient matrix of the optimization variable range; Step 3, pre-selecting a swarm intelligence optimization algorithm in the optimization module, randomly initializing the population individuals under the constraint condition of only considering the optimization variable range of the optimization problem, and each individual is represented as a vector spliced by binary coding of all optimization variables; Step 4, the optimization module executes the iteration of the swarm intelligence optimization algorithm, and in the neighborhood of each individual in the gth generation population, the Taylor series expansion is used to obtain a linear objective function or a quadratic objective function, to form N independent linear or quadratic optimization problems with constraints; wherein, the initial value of g is 1, and N represents the number of population individuals; Optimization variable for the i-th individual in the g-th generation Within its neighborhood ε i The linear objective function using Taylor series expansion is as follows: The quadratic objective function is as follows: wherein and denote the first and second derivative of the non-linear function f at respectively, O m denotes an m-th order infinitesimal under Taylor series expansion; the upper index T denotes transposition; According to the requirements of the application scene, the linear objective function or the quadratic objective function is selected to continue the following steps; Step 5, preselect a traditional optimization algorithm in the optimization module, solve the linear objective function or quadratic objective function with constraints in the neighborhood of each individual in the gth generation population obtained in step 4 using the selected traditional optimization algorithm, obtain the optimal solution in the neighborhood as the fitness function value of the individual; select the current optimal individual from the N individuals, and compare it with the historical optimal individual X global Compare and judge whether to update the historical optimal individual; then the population iteration number g is increased by 1, and it is judged whether g is greater than the set maximum iteration number, if not, step 6 is executed, and if yes, step 7 is executed; Wherein, if there is no solution to the objective function by using the traditional optimization algorithm, it means that the individual does not meet the constraint condition, then the individual is randomly initialized again, and the individual is returned to step 4 for execution; Step 6, the optimization module continues to execute the genetic optimization algorithm, updates the N individuals of the new generation, and then goes to step 4; Step 7, the historical optimal individual X is calculated by using the solving process of step 4 and 5 global Optimal solution in the neighborhood The optimal configuration scheme of the energy storage system is obtained according to the optimization variable vector.

2. The method of claim 1, wherein, In step 4, the first derivative and the second derivative of the non-linear function f at are calculated as follows: where x i represents the i-th optimization variable, i = 1, 2, … n, and n is the number of optimization variables.

3. The method of claim 1, wherein, In the step 5, when the selected traditional optimization algorithm is the Lagrange multiplier method, based on the linear objective function or the quadratic objective function described in the step 4, a Lagrange function is constructed is converted into an unconstrained objective function as follows: or where λk k and respectively represent the Lagrange coefficients of the kth equality constraint and the st inequality constraint, and constitute the Lagrange coefficient vector λ and h k (X) and y s (X) respectively represent the kth equality constraint and the st inequality constraint in the mathematical model of the linear objective function or the quadratic objective function of step 4, K is the number of equality constraints, and S is the number of inequality constraints; The Lagrange function is The optimization variable X and the Lagrange coefficient vector λ and Take the partial derivative and set the derivative to zero to form a system of equations, and solve the system of equations to obtain the optimization variable X of the i-th individual in the g-th generation population The optimal solution of the neighborhood And the corresponding function value The function value Is the fitness function value of the individual in the genetic optimization algorithm If the system of equations has no solution, it means that the individual cannot satisfy the constraint condition, and the individual is randomly initialized again, and then step 4 is executed again; The fitness function value of the N individuals in the gth generation population is calculated, sorted and selected, and the optimal individual is selected, if the fitness function value of the selected optimal individual is better than that of the historical optimal individual, the historical optimal individual is updated by using the optimal individual selected from the gth generation, otherwise no update is performed.

Images