Energy storage system control parameter identification method based on TS-NSGA-II algorithm

Abstract

An energy storage system control parameter identification method based on a TS-NSGA-II algorithm comprises the steps of building an energy storage system actual measurement model based on a hardware-in-the-loop test platform, and collecting energy storage system output response data under various fault ride-through working conditions; building an energy storage system identification model, and determining the to-be-identified parameters as energy storage inverter PI control parameters according to the energy storage system identification model; a parameter identification problem is converted into a multi-objective optimization problem, and average absolute errors of different stages of active currents and reactive currents output by the energy storage system identification model and the energy storage system actual measurement model and the maximum absolute deviation of a steady-state stage are used as objective functions respectively; pI control parameters of the energy storage system are used as population individuals, a TS-NSGA-II algorithm is adopted to solve the objective function, and the obtained optimal population is the identification result. According to the method, the control parameters of the energy storage system model are identified through the TS-NSGA-II algorithm, so that the parameter identification precision can be effectively improved; compared with the traditional method, the method has the advantage that the parameter identification result has higher consistency.

Description

Technical Field

[0001] The present invention relates to the technical field of energy storage system parameter identification, and particularly relates to a method for identifying control parameters of an energy storage system based on the TS-NSGA-II algorithm. Background Art

[0002] The energy storage system is an important and indispensable part of the modern energy system, and it plays a key role in building a clean energy system, improving energy utilization efficiency, and ensuring the safe and stable operation of the power system. With the large-scale access of renewable energy (such as solar energy and wind energy) and the popularization of electric vehicles, the volatility and uncertainty of the power system have increased significantly. The energy storage system can store excess energy and release it when needed, playing the role of peak shaving and valley filling and improving power quality.

[0003] Establishing an accurate energy storage system simulation model is the key to studying the operating characteristics and regulation strategies of large-scale new energy access to the power grid, and parameter identification is a necessary means to solve the problem of missing model parameters. By obtaining the key parameters of the energy storage system and new energy units through measurement and identification methods, the deficiencies of insufficient manufacturer data can be made up, providing important support for power system dispatching optimization, new energy operating characteristic analysis, and power quality research. This can not only improve the reliability of simulation analysis but also lay a solid foundation for the stable operation of high-proportion new energy access to the power grid.

[0004] When using the existing NSGA-II algorithm to identify the control parameters of the energy storage system, there are problems of low efficiency and inaccurate results. The main manifestations are as follows: when the traditional NSGA-II algorithm is applied to the identification problem of the energy storage system model, the algorithm performs well in multi-objective optimization problems, but in the case of high-dimensional objectives (the number of objectives exceeds 3 or more), the performance will significantly decline, and the identification of the control parameters of the energy storage system involves dynamic performance and steady-state performance, which is usually a high-dimensional problem. In addition, the NSGA-II algorithm has a slow convergence speed, high computational complexity, and is prone to falling into local optima, making it difficult to optimize convergence and diversity simultaneously, greatly reducing the efficiency of identifying the control parameters of the energy storage system. In summary, the existing NSGA-II algorithm is not suitable for identifying the control parameters of the energy storage system. Summary of the Invention

[0005] To solve the above technical problems, the present invention provides a method for identifying control parameters of an energy storage system based on the TS-NSGA-II algorithm. By using the TS-NSGA-II algorithm to identify the control parameters of the energy storage system model, the accuracy of parameter identification can be effectively improved; compared with traditional methods, the parameter identification results of this method have higher consistency.

[0006] The technical solution adopted by the present invention is as follows:

[0007] A method for identifying control parameters of an energy storage system based on the TS-NSGA-II algorithm includes the following steps:

[0008] Step 1: Based on a hardware-in-the-loop test platform, build an actual measurement model of the energy storage system, and collect the output response data of the energy storage system under various fault ride-through conditions;

[0009] Step 2: Build an identification model of the energy storage system, and determine the parameters to be identified as the PI control parameters of the grid-side inverter of the energy storage system according to the identification model of the energy storage system;

[0010] Step 3: Convert the parameter identification problem into a multi-objective optimization problem, and use the average absolute error in different stages and the maximum absolute deviation in the steady state of the active current and reactive current output by the energy storage system identification model and the actual measurement model of the energy storage system as the objective functions respectively;

[0011] Step 4: Use the PI control parameters of the energy storage system as population individuals, and use the TS-NSGA-II algorithm to solve the objective function. The obtained optimal population is the identification result.

[0012] It also includes Step 5: Verify the identification result of the energy storage system fault ride-through. Substitute the identification result into the energy storage system identification model to obtain the output active and reactive current response curves, compare them with the active and reactive current response curves output by the actual measurement model of the energy storage system, calculate the root mean square error, and evaluate the accuracy of the identification result.

[0013] In the above Step 1, as Figure 1 shown, build the main circuit model of the energy storage system in the semi-physical simulation platform, and complete the digital and analog signal interaction between the simulation platform and the energy storage controller through wiring to realize the grid connection of the energy storage system, and conduct tests under various fault ride-through conditions, and collect the response waveform data of the voltage U, active current I d , reactive current I q output by the energy storage system under different conditions.

[0014] In the above Step 2, build an identification model of the energy storage system. Among them, the main circuit of the identification model of the energy storage system is the same as the main circuit model of the energy storage system built by the semi-physical simulation platform. The control part adopts closed-loop control, which is divided into a power outer-loop controller model and a current inner-loop controller model. The power outer-loop control model is:

[0015]

[0016] In formula (1): i d_ref , i q_ref are the output reference values of the DC voltage outer-loop controller and the grid-connected voltage outer-loop controller respectively; P ref , Q refare the active power reference value and the reactive power reference value respectively; u g is the grid voltage amplitude.

[0017] The current inner-loop controller model is:

[0018]

[0019] In Equation (2): K pd , K id are the proportional coefficient and the integral coefficient of the active current inner-loop controller respectively; K pq , K iq are the proportional coefficient and the integral coefficient of the reactive current inner-loop controller respectively; u d , u q are the d-axis and q-axis voltage control signals respectively; e d , e q are the d-axis and q-axis components of the grid-connected voltage respectively; L is the filter inductor; ω is the synchronous angular velocity; i d , i q are the d-axis and q-axis current signals respectively.

[0020] In the energy storage system identification model, parameters such as P ref , Q ref , e d , e q , L, ω, etc. can be found in the technical manual or calculated, so the control parameters to be identified in the model are determined as K pd , K id , K pq and K iq .

[0021] Step 3 includes the following steps:

[0022] Step 3.1: As shown in Figure 3 , the whole process of the energy storage system fault ride-through response curve is divided into 5 stages: pre-fault steady state, transient during fault, quasi-steady state during fault, transient after fault clearance, and steady state after fault clearance, which are respectively labeled as: Stage I, Stage II, Stage III, Stage IV, Stage V. Among them, Stage I, Stage III, and Stage V can be regarded as steady-state stages, and Stage II and Stage IV can be regarded as transient stages.

[0023] Step 3.2: In order to improve the accuracy of parameter identification, the sum of the mean absolute errors between the measured values and the identified values of the active current and the reactive current in the steady-state stage is used as the objective functions f 1 and f 2 respectively, the maximum absolute errors in the steady-state stage are used as the objective functions f 3 and f 4 respectively, and the mean absolute errors in the transient stage are used as the objective functions f 5 and f6 , in addition, considering that the active current and reactive current vary differently in different fault conditions, different weight ratios need to be set for the objective function. The expression of the objective function is as follows:

[0024]

[0025] In formula (3): I d_mea_QS_i and I d_sim_QS_i are the measured value and the identified value of the active current in the steady state stage respectively; I d_mea_TR_i and I d_sim_TR_i are the measured value and the identified value of the active current in the transient stage respectively; I q_mea_QS_i and I q_sim_QS_i are the measured value and the identified value of the reactive current in the steady state stage respectively; I q_mea_TR_i and I q_sim_TR_i are the measured value and the identified value of the reactive current in the transient stage respectively; ω 1 , ω 2 , ω 3 , ω 4 , ω 5 and ω 6 are the weight values of different objective functions respectively; n is the number of sampling data.

[0026] In step 4, the TS-NSGA-II algorithm adopted is a two-stage evolutionary algorithm, aiming to solve the problem that it is difficult to balance the convergence and the diversity of the solution set in high-dimensional multi-objective optimization problems. By designing a two-stage dynamic framework, this algorithm emphasizes different objectives in different stages. The first stage: enhancing the convergence of the population through the subregion-based PBI dominance relationship (SubregionPBI-Dominance, SPD), dividing the objective space into a series of subregions, and applying the Penalty Boundary Intersection (PBI) within each subregion to make the solution quickly approach the Pareto front; The second stage: enhancing the diversity of the solution set through the level-based Pareto dominance relationship (Level-based Pareto Dominance, LPD), covering a wider Pareto front.

[0027] The TS-NSGA-II algorithm identifies the control parameters of the energy storage system, including the following steps:

[0028] Step B1: Initialize the algorithm, set the random initial population P 0 , the population size is μ, the number of objective functions is m, the dimension is D, the maximum number of iterations is G max , and the current number of iterations is g.

[0029] Step B2: Calculate the objective function values of the population individuals according to the objective function formula (3).

[0030] Step B3: Determine the current optimization stage according to the generation number g, and the formula is as follows:

[0031]

[0032] In formula (4): M stage is the stage parameter; η is the stage switching parameter.

[0033] Among them, in the first stage, based on the Subregion PBI-Dominance (SPD), the convergence of the population is enhanced, and the Pareto front is approximated preferentially; in the second stage, based on the Level-based Pareto Dominance (LPD), the uniform distribution of the solutions on the Pareto front is ensured.

[0034] Step B4: According to the value of M stage , judge whether to use the SPD dominance sorting strategy in the first stage or the LPD dominance sorting strategy in the second stage to sort the current population and generate the parent population P 1 .

[0035] Step B5: Perform crossover and mutation on P 1 to generate the offspring population O, and merge it with P 1 to form the combined population C.

[0036] Step B6: According to the value of M stage , judge whether to use the SPD rule in the first stage (select the individual with a smaller PBI value according to the PBI sorting of individuals) or the LPD rule in the second stage (calculate the level of the individual according to the hierarchical mechanism and select the solution with a lower level), and select a population P 2 with a fixed size from C.

[0037] Step B7: If the iteration number G max is reached, stop the iteration, and output the optimal solution set as the parameter identification result; otherwise, repeat steps B2 - B6 until the iteration number is reached.

[0038] In step 4, the SPD optimization process of the TS-NSGA-II algorithm includes the following steps:

[0039] Step C1: Normalize the objective function values to eliminate the influence of different objective magnitudes. The calculation formula is as follows:

[0040]

[0041] In Equation (5): x is an individual in the population, is the normalization result of the j-th objective function f j (x); is the minimum point of this objective function; is the maximum point of this objective function.

[0042] Step C2: Generate a reference vector set W by the uniform distribution method, decompose the objective space, and use the reference vectors to divide the objective space into multiple sub-regions, with each sub-region corresponding to a reference vector. The position of individual x in the objective space is determined by its objective values. Calculate the angle between the normalized objective function value of x and each reference vector w j , and x can be assigned to the sub-region corresponding to the reference vector w j with the closest direction, and the calculation formula is as follows:

[0043]

[0044] In Equation (6): is the normalized objective function vector and the reference vector w i The angle between them, k(x) is the sub-region index, and i is the number of reference vectors.

[0045] Step C3: For individual x, within the sub-region it belongs to, calculate the projection distance d 1 (x) and the perpendicular distance d 2 (x), and the calculation formula is:

[0046]

[0047] In the formula: is the normalized objective function vector of x, and w is the reference vector of this sub-region.

[0048] Step C4: Calculate the PBI value of x, and the expression is as follows:

[0049] PBI(x) = d 1 (x) + θ · d 2 (x) (9);

[0050] In Equation (9): PBI(x) is the PBI value of x; θ represents the penalty factor.

[0051] Step C5: Judge the SPD dominance relationship of individuals. If u and v are both random individuals in the population, the condition that individual u SPD dominates individual v is: the two solutions need to be in the same sub-region, and the PBI value of u is less than that of v, and the calculation formula is as follows:

[0052] k(u) = k(v) & PBI(u) < PBI(v) (10);

[0053] In formula (10): k(u) and k(v) are the sub-region indices of u and v respectively; PBI(u) and PBI(v) are the PBI values of u and v respectively.

[0054] In step 4, the second-stage LPD optimization process of the TS-NSGA-II algorithm includes the following steps:

[0055] Step D1: Divide the range of the j-th objective value into multiple equal intervals, and calculate the level number l of the individual x j , and the calculation formula is as follows:

[0056]

[0057] In the formula: and are the minimum and maximum objective function values of the j-th objective of the individual x respectively; △f j (x) is the objective function interval value; a is the level and a ∈ {1, n p} and n p represents the number of equal intervals; f j (x) is the objective function value of the j-th objective.

[0058] Step D2: Combine the levels of all objectives, and take the maximum value as the overall level of the individual. The calculation formula is:

[0059]

[0060] In formula (13): l(x) represents the overall level of the individual; j represents the number of objectives.

[0061] Step D3: Judge the LPD dominance relationship of the individual. If u and v are both random individuals in the population, the condition that individual u LPD dominates individual v is: individual u Pareto dominates individual v, or l(u) < l(v), where l(u) and l(v) represent the overall levels of u and v respectively.

[0062] In step 5, substitute the identification result into the energy storage system identification model, simulate to obtain the active and reactive current response curves, and calculate the error between the simulation result and the measured result. The formula is as follows:

[0063]

[0064] In the formula: σ d and σ q are the relative errors of the active and reactive currents respectively; i d_mea and i q_mea are the measured active and reactive current values respectively; id_sim and i q_sim are the active and reactive current values identified by the simulation of the identification model respectively; n is the number of sampled data. A method for identifying control parameters of an energy storage system based on the TS-NSGA-II algorithm of the present invention has the following beneficial effects:

[0065] 1) The present invention divides the output response curve of the energy storage system into five different stages, and uses the mean absolute error and maximum error between the measured values and the identified values in different stages as the objective function, and adds different weight ratios to the objective function according to needs, so that the parameter identification can consider both the steady-state error and the transient error at the same time, and the optimization result is more comprehensive.

[0066] 2) The present invention identifies the control parameters of the energy storage system model in the time-domain simulation, which can effectively reflect the actual operating state of the energy storage system model and the influence of each control parameter on the stable operation of the actual system.

[0067] 3) The present invention uses the TS-NSGA-II algorithm for multi-objective optimization. Through a staged optimization strategy, it effectively combines the advantages of global search and local search. By enhancing the convergence of the population in the SPD stage and improving the diversity of the solution set in the LPD stage, the parameter identification achieves a balance between convergence and diversity in high-dimensional objective optimization.

[0068] 4) The present invention identifies the control parameters of the energy storage system model through the TS-NSGA-II algorithm, which can effectively improve the accuracy of parameter identification.

[0069] 5) In the present invention, the TS-NSGA-II (Two-Stage Non-dominated Sorting Genetic Algorithm II) algorithm solves the problem of balance between convergence and diversity in high-dimensional optimization problems by introducing a reference vector, a two-stage optimization framework and a dynamic switching mechanism, and significantly improves the identification efficiency and accuracy in the energy storage system. Description of the Drawings

[0070] The following further illustrates the present invention in conjunction with the drawings and examples;

[0071] Figure 1 is the semi-physical simulation test block diagram of the energy storage system of the present invention.

[0072] Figure 2 is the flow chart of the method for identifying the control parameters of the energy storage system of the present invention.

[0073] Figure 3 are the voltage U and the reactive current I q and the active current I d The different stage division diagram of the fault ride-through response curve.

[0074] Figure 4 This is the current inner loop control block diagram of the energy storage system of the present invention.

[0075] Figure 5 This is the flow chart of parameter identification of the TS-NSGA-II algorithm of the present invention.

[0076] Figure 6 This is the comparison diagram of the active and reactive current response curves of the two identification algorithms of the present invention and the measured curves under Condition 1.

[0077] Figure 7 This is the comparison diagram of the active and reactive current response curves of the two identification algorithms of the present invention and the measured curves under Condition 2. Specific implementation mode

[0078] Based on the hardware-in-the-loop test platform, the present invention builds an actual measurement model of the energy storage system and collects the output response data of the energy storage system under various fault ride-through conditions. Build an energy storage system identification model, and determine the parameters to be identified as the PI control parameters of the energy storage inverter according to the actual measurement model of the energy storage system. Transform the parameter identification problem into a multi-objective optimization problem, and use the average absolute error of different stages of the active current and reactive current output by the energy storage identification model and the actual measurement model of the energy storage system and the maximum absolute deviation in the steady state stage as the objective functions respectively. And take the PI control parameters of the energy storage system as the population individuals, and use the TS-NSGA-II algorithm to solve the objective functions. The obtained optimal population is the identification result. Finally, verify the identification result of the energy storage system fault ride-through. Substitute the identification result into the energy storage system identification model to obtain the output active and reactive current response curves, and compare them with the active and reactive current response curves output by the energy storage actual measurement model, and calculate the root mean square error to evaluate the accuracy of the identification result

[0079] As Figure 2 shown, the method for identifying the control parameters of the energy storage system based on the TS-NSGA-II algorithm includes the following steps:

[0080] Step 1: Based on the hardware-in-the-loop test platform, build an actual measurement model of the energy storage system and collect the output response data of the energy storage system under various fault ride-through conditions.

[0081] Step 2: Build an energy storage system identification model, and determine the parameters to be identified as the PI control parameters of the energy storage inverter according to the actual measurement model of the energy storage system.

[0082] Step 3: Transform the parameter identification problem into a multi-objective optimization problem, and use the average absolute error of different stages of the active current and reactive current output by the energy storage identification model and the actual measurement model of the energy storage system and the maximum absolute deviation in the steady state stage as the objective functions respectively.

[0083] Step 4: Using the PI control parameters of the energy storage system as population individuals, the TS-NSGA-II algorithm is used to solve the objective function, and the optimal population obtained is the identification result.

[0084] Step 5: Verify the fault ride-through identification result of the energy storage system. Substitute the identification result into the energy storage system identification model to obtain the output active and reactive current response curves, compare them with the active and reactive current response curves output by the actual measurement model of the energy storage system, calculate the root mean square error, and evaluate the accuracy of the identification result.

[0085] As Figure 1 shown, the output response data of the energy storage system refers to the response data set during the entire fault period of the voltage U, active current I d , and reactive current I q output by the energy storage system based on the hardware-in-the-loop simulation platform, building the main circuit model of the energy storage system, and interacting with the actual energy storage system controller through digital-to-analog signals.

[0086] Specifically, establish a closed-loop control model for the energy storage system, where the power outer-loop control model is:

[0087]

[0088] In the formula: i d_ref , i q_ref are the output reference values of the DC voltage outer-loop controller and the grid voltage outer-loop controller respectively; P ref , Q ref are the active power reference value and the reactive power reference value respectively; u g is the amplitude of the grid voltage.

[0089] As Figure 4 shown, the current inner-loop controller model is:

[0090]

[0091] In the formula: K pd , K id are the proportional coefficient and integral coefficient of the active current inner-loop controller respectively; K pq , K iq are the proportional coefficient and integral coefficient of the reactive current inner-loop controller respectively; u d , u q are the d-axis and q-axis voltage control signals respectively; e d , e q are the d-axis and q-axis components of the grid voltage respectively; L is the filter inductor; ω is the synchronous angular velocity.

[0092] According to the energy storage system control model, determine that the control parameters to be identified in the model include Kpd , K id , K pq and K iq .

[0093] Specifically, the establishment of the objective function is divided into the following steps:

[0094] Step A1, as shown in Figure 3 , the whole process of the energy storage system fault ride-through response curve is divided into 5 stages: pre-fault steady state, transient during fault, quasi-steady state during fault, transient after fault clearance, and steady state after fault clearance, which are respectively labeled as stage Ⅰ, stage Ⅱ, stage Ⅲ, stage Ⅳ, and stage Ⅴ. Among them, stage Ⅰ, stage Ⅲ, and stage Ⅴ can be regarded as steady state stages, and stage Ⅱ and stage Ⅳ can be regarded as transient stages.

[0095] Step A2, in order to improve the accuracy of parameter identification, the sum of the mean absolute errors between the measured values and the identified values of the active current and reactive current in the steady state stage is used as the objective function f 1 and f 2 , and the maximum absolute error in the steady state stage is used as the objective function f 3 and f 4 , and the mean absolute error in the transient stage is used as the objective function f 5 and f 6 . In addition, considering the different changes in the active current and reactive current in each stage under different fault conditions, different weight ratios need to be set for the objective function. The expression of the objective function is as follows:

[0096]

[0097] In the formula: I d_mea_QS_i , I d_sim_QS_i are respectively the measured value and the identified value of the active current in the steady state stage; I d_mea_TR_i , I d_sim_TR_i are respectively the measured value and the identified value of the active current in the transient stage; I q_mea_QS_i , I q_sim_QS_i are respectively the measured value and the identified value of the reactive current in the steady state stage; I q_mea_TR_i , I q_sim_TR_i are respectively the measured value and the identified value of the reactive current in the transient stage; ω 1 , ω 2 , ω 3 , ω 4 , ω 5 and ω 6 are respectively the weight values of different objective functions; n is the number of sampled data.

[0098] As shown in Figure 5As shown, the adopted TS-NSGA-II algorithm is a two-stage evolutionary algorithm aimed at solving the problem of difficult balance between convergence and diversity of the solution set in high-dimensional multi-objective optimization problems. By designing a two-stage dynamic framework, this algorithm emphasizes different objectives in different stages. The first stage: enhance the convergence of the population through the Subregion PBI-Dominance (SPD), divide the objective space into a series of subregions, and apply the Penalty Boundary Intersection (PBI) in each subregion to make the solution quickly approach the Pareto front. The second stage: improve the diversity of the solution set through the Level-based Pareto Dominance (LPD) to cover a wider Pareto front. The steps for the TS-NSGA-II algorithm to identify the control parameters of the energy storage system are as follows:

[0099] Step B1, initialize the algorithm, generate a reference vector set W, and set a random initial population P 0 , the population size is μ, the number of objective functions is m, the dimension is D, and the maximum number of iterations is G max , the current iteration number is g;

[0100] Step B2, calculate the objective function values of the population individuals according to the objective function formula (3);

[0101] Step B3, determine the current optimization stage according to the generation number g, and the formula is as follows:

[0102]

[0103] In the formula: M stage is the stage parameter; η is the stage switching parameter;

[0104] Among them, in the first stage, SPD mainly strengthens the convergence of the algorithm and gives priority to approaching the Pareto front. In the second stage, LPD ensures the uniform distribution of the solutions on the Pareto front.

[0105] Step B4, use different dominance sorting strategies for sorting according to the sorting rules of the current stage to obtain the parent population P 1 ;

[0106] Step B5, perform crossover and mutation on P 1 to generate an offspring population O, and merge it with P 1 to form a combined population C;

[0107] Step B6, according to the current stage of environmental selection rules, the first stage adopts the SPD rule, and prioritizes individuals with small PBI values ​​according to the PBI sorting. The second stage adopts the LPD rule, uses a hierarchical mechanism to calculate the level of individuals, and selects low-level solutions. According to the rules, a fixed-size population P is selected from C. 2 ;

[0108] Step B7: If the number of iterations G is reached max , then the iteration stops and the optimal solution set is output as the parameter identification result. Otherwise, repeat steps B2-B6 until the number of iterations is reached;

[0109] The first stage SPD optimization process of the TS-NSGA-II algorithm includes the following steps:

[0110] Step C1, normalize the objective function value to eliminate the influence of different objective magnitudes. The calculation formula is as follows:

[0111]

[0112] Where: x is an individual in the population, is the jth objective function f j Normalized result of (x); is the minimum point of the objective function; is the maximum point of the objective function.

[0113] Step C2, decompose the target space and divide the target space into multiple sub-regions using reference vectors. Individual x is assigned to the sub-region corresponding to the reference vector closest to its direction and is counted in the sub-region index k(x).

[0114] Step C3: for individual x, calculate the projection distance d of x in the sub-region to which it belongs 1 (x) and the vertical distance are both d 2 (x), the calculation formula is:

[0115]

[0116] Where: is the normalized objective function vector of x, and w is the reference vector of the sub-region.

[0117] Step C4, calculate the PBI value of x, the expression is as follows:

[0118] PBI(x)=d 1 (x)+θ·d 2 (x) (9);

[0119] Where: θ represents the penalty factor.

[0120] Step C5: Determine the SPD domination relationship of individuals. If u and v are both random individuals in the population, the condition for u to SPD-dominate v is that the two solutions need to be in the same sub-region and the PBI value of u is less than that of v. The calculation formula is as follows:

[0121] k(u) = k(v) & PBI(u) < PBI(v) (10);

[0122] The second-stage LPD optimization process of the TS-NSGA-II algorithm includes the following steps:

[0123] Step D1: Divide the range of the j-th objective value into n p equal intervals, and calculate the level number l j of individual x. The calculation formula is as follows:

[0124]

[0125] In the formula: and are the minimum and maximum objective function values of the j-th objective of individual x respectively; △f j (x) is the objective function interval value; a is the level and a ∈ {1, n p}}.

[0126] Step D2: Combine the levels of all objectives and take the maximum value as the overall level of the individual. The calculation formula is:

[0127]

[0128] Step D3: Determine the LPD domination relationship of individuals. If u and v are both random individuals in the population, the condition for u to LPD-dominate v is that u Pareto-dominates v, or l(u) < l(v).

[0129] For the energy storage system control parameter identification method based on the TS-NSGA-II algorithm, taking the mean absolute error as the evaluation index, substituting the identification results into the energy storage system identification model, simulating to obtain the active and reactive current response curves, and calculating the error between the simulation results and the measured results. The formula is as follows:

[0130]

[0131] In the formula: σ d and σ q are the relative errors of the active and reactive currents respectively, i d_mea and i q_mea are the measured active and reactive current values respectively, i d_sim and i q_sim are the active and reactive current values simulated by the identification model respectively.

[0132] The present invention provides a method for identifying control parameters of an energy storage system based on the TS-NSGA-II algorithm. Under time-domain simulation, the mean absolute percentage errors between the identified curves and the true curves under two different working conditions are calculated as shown in Table 1.

[0133] Table 1 Identification errors of the response curves of the energy storage system model under different working conditions

[0134]

[0135]

[0136] It can be seen from Table 1 that the identification errors of the response curves of the energy storage system model under different working conditions are small, and the method proposed by the present invention can effectively improve the accuracy of identifying the control parameters of the energy storage system.

[0137] As Figure 6 shown, Condition 1: A three-phase short-circuit fault is applied at t = 0.5 s, causing the grid-connected point voltage to drop to 20% of the rated voltage, and the fault disappears at t = 1.125 s.

[0138] As Figure 7 shown, Condition 2: A three-phase short-circuit fault is applied at t = 0.5 s, causing the grid-connected point voltage to rise to 125% of the rated voltage, and the fault disappears at t = 1.5 s.

Claims

1. The energy storage system control parameter identification method based on TS-NSGA-II algorithm is characterized by The following steps are involved: Step 1: Build a measurement model of the energy storage system and collect output response data of the energy storage system under various fault ride-through conditions; Step 2: Build an energy storage system identification model, and determine the parameters to be identified as PI control parameters of the grid-side inverter of the energy storage system according to the energy storage system identification model; Step 3: Convert the parameter identification problem into a multi-objective optimization problem, and use the average absolute error of the active current and reactive current output by the energy storage system identification model and the energy storage system measured model in different stages and the maximum absolute deviation in the steady state stage as the objective functions; Step 4: Take the PI control parameters of the energy storage system as population individuals, use the TS-NSGA-II algorithm to solve the objective function, and the optimal population obtained is the identification result.

2. The energy storage system control parameter identification method based on the TS-NSGA-II algorithm according to claim 1 is characterized in that: It also includes step 5: verifying the fault ride-through identification results of the energy storage system, bringing the identification results into the energy storage system identification model, obtaining the output active and reactive current response curves, comparing them with the active and reactive current response curves output by the measured model of the energy storage system, calculating the root mean square error, and evaluating the accuracy of the identification results.

3. The energy storage system control parameter identification method based on the TS-NSGA-II algorithm according to claim 1 is characterized in that: In step 1, the main circuit model of the energy storage system is built, and the digital and analog signal interaction between the simulation platform and the energy storage controller is completed to realize the grid connection of the energy storage system, and to conduct tests of various fault ride-through conditions, and to collect the output voltage U and active current I of the energy storage system under different conditions. d , reactive current I q Response waveform data.

4. The energy storage system control parameter identification method based on the TS-NSGA-II algorithm according to claim 1 is characterized in that: In step 2, an energy storage system identification model is constructed, wherein the control part of the energy storage system identification model adopts closed-loop control, which is divided into a power outer loop controller model and a current inner loop controller model. The power outer loop control model is: In formula (1): i d_ref 、i q_ref are the DC voltage outer loop controller output reference value and the grid voltage outer loop controller output reference value respectively; P ref , Q ref They are active power reference value and reactive power reference value respectively; u g is the grid voltage amplitude; The current inner loop controller model is: In formula (2): K pd , K id are the proportional coefficient and integral coefficient of the active current inner loop controller respectively; K pq , K iq are the proportional coefficient and integral coefficient of the reactive current inner loop controller respectively; u d 、u q are the d-axis and q-axis voltage control signals respectively; d 、e q are the d-axis and q-axis components of the grid-connected voltage respectively; L is the filter inductance; ω is the synchronous angular velocity; i d 、i q They are d-axis and q-axis current signals respectively; In the energy storage system identification model, the control parameter to be identified in the model is determined as K pd , K id , K pq and K iq .

5. The energy storage system control parameter identification method based on the TS-NSGA-II algorithm according to claim 1 is characterized by: The step 3 comprises the following steps: Step 3.1: Divide the entire process of the energy storage system fault ride-through response curve into five stages: steady state before fault, transient state during fault, transient steady state during fault, transient state after fault clearing, and steady state after fault clearing, respectively marked as: stage I, stage II, stage III, stage IV, stage V, among which stage I, stage III and stage V can be regarded as steady state stages, and stage II and stage IV as transient stages; Step 3.2: In order to improve the accuracy of parameter identification, the mean absolute error of the measured value and the identified value of the active current and reactive current in the steady state stage is used as the objective function f1 and f2, the maximum absolute error in the steady state stage is used as the objective function f3 and f4, and the mean absolute error in the transient stage is used as the objective function f5 and f6. In addition, considering that the changes of active current and reactive current in each stage are different under different fault conditions, different weight ratios need to be set for the objective function; the expression of the objective function is as follows: In formula (3): I d_mea_QS_i ,I d_sim_QS_i are respectively the measured value and the identified value of the active current in the steady state stage; I d_mea_TR_i ,I d_sim_TR_i are respectively the measured value and the identified value of active current in the transient stage; I q_mea_QS_i ,I q_sim_QS_i are the measured value and identified value of reactive current in steady state respectively; I q_mea_TR_i ,I q_sim_TR_i are the measured value and identified value of reactive current in the transient stage respectively; ω1, ω2, ω3, ω4, ω5 and ω6 are the weight values ​​of different objective functions respectively; n is the number of sampling data.

6. The energy storage system control parameter identification method based on the TS-NSGA-II algorithm according to claim 1 is characterized by: In step 4, the TS-NSGA-II algorithm is used to identify the control parameters of the energy storage system, including the following steps: Step B1: Initialize the algorithm, set the random initial population P0, the population size μ, the number of objective functions m, the dimension D, and the maximum number of iterations G max , the current number of iterations g; Step B2: Calculate the objective function value of the population individuals according to the objective function formula (3); Step B3: According to the algebra g, determine the current optimization stage, the formula is as follows: In formula (4): M stage is the stage parameter; η is the stage switching parameter; In the first stage, the convergence of the population is enhanced based on the subregion PBI-Dominance (SPD) relationship, giving priority to approaching the Pareto frontier; in the second stage, the level-based Pareto Dominance (LPD) relationship is used to ensure the uniform distribution of solutions on the Pareto frontier; Step B4: According to M stage The value of determines whether to use the SPD-dominated sorting strategy of the first stage or the LPD-dominated sorting strategy of the second stage to sort the current population and generate the parent population P1; Step B5: Perform crossover mutation on P1 to generate the offspring population O, and merge it with P1 to form the combined population C; Step B6: According to M stage The value of determines whether to use the SPD rule in the first stage or the LPD rule in the second stage to select a fixed-size population P2 from C; Step B7: If the number of iterations G is reached max , then the iteration stops and the optimal solution set is output as the parameter identification result. Otherwise, repeat steps B2-B6 until the number of iterations is reached.

7. The energy storage system control parameter identification method based on TS-NSGA-II algorithm according to claim 6 is characterized by: The first-stage SPD optimization process of the TS-NSGA-II algorithm includes the following steps: Step C1: Normalize the objective function values to eliminate the influence of different objective magnitudes; the calculation formula is as follows: In formula (5), x is an individual in the population, is the jth objective function f j Normalized result of (x); is the minimum point of the objective function; is the maximum point of the objective function; Step C2: Generate a reference vector set W through the uniform distribution method, decompose the objective space, and use the reference vectors to divide the objective space into multiple sub-regions, with each sub-region corresponding to a reference vector; the position of an individual x in the objective space is determined by its objective values; Calculate the normalized objective function value of x and each reference vector w j The angle between x and w can be used to assign x to the reference vector w that is closest to its direction. j The corresponding sub-area is calculated as follows: In formula (6): is the normalized objective function vector With the reference vector w i The angle between them, k(x) is the sub-region index, and i is the number of reference vectors; Step C3: For the individual x, within the sub-region to which it belongs, calculate the projection distance d1(x) and the perpendicular distance d2(x) of x, and the calculation formula is: Where: is the normalized objective function vector of x, and w is the reference vector of the sub-region; Step C4: Calculate the PBI value of x, and the expression is as follows: PBI(x) = d1(x) + θ · d2(x) (9); In Equation (9): PBI(x) is the PBI value of x; θ represents the penalty factor; Step C5: Judge the SPD dominance relationship of individuals. If u and v are both random individuals in the population, the condition that individual u SPD dominates individual v is: the two solutions need to be in the same sub-region, and the PBI value of u is less than that of v. The calculation formula is as follows: k(u) = k(v) & PBI(u) < PBI(v) (10); In Equation (10): k(u) and k(v) are the sub-region indices of u and v respectively; PBI(u) and PBI(v) are the PBI values of u and v respectively.

8. The energy storage system control parameter identification method based on TS-NSGA-II algorithm according to claim 6 is characterized by: The second-stage LPD optimization process of the TS-NSGA-II algorithm includes the following steps: Step D1: Divide the range of the jth target value into multiple equal intervals and calculate the number of levels l of individual x j , the calculation formula is as follows: Where: and are the minimum and maximum objective function values ​​of the jth objective of individual x; △f j (x) is the interval value of the objective function; a is the level and a∈{1,n p },n p represents the number of equal intervals; f j (x) is the objective function value of the jth target; Step D2: Synthesize the levels of all objectives and take the maximum value as the overall level of the individual. The calculation formula is: In Equation (13): l(x) represents the overall level of the individual; j represents the number of objectives; Step D3: Judge the LPD dominance relationship of individuals. If u and v are both random individuals in the population, the condition that individual u LPD dominates individual v is: individual u Pareto dominates individual v, or l(u) < l(v), where l(u) and l(v) represent the overall levels of u and v respectively.

9. The energy storage system control parameter identification method based on TS-NSGA-II algorithm according to claim 2 is characterized by: In Step 5, substitute the identification result into the energy storage system identification model, simulate to obtain the active and reactive current response curves, and calculate the error between the simulation result and the measured result. The formula is as follows: Where: d and σ q are the relative errors of active and reactive current respectively; i d_mea and i q_mea are the measured active and reactive current values ​​respectively; i d_sim and i q_sim are the active and reactive current values ​​of the identification model simulation respectively; n is the number of sampling data.

Images