Photovoltaic grid-connected system controller parameter identification method based on uncertainty evaluation

Abstract

The application discloses a photovoltaic grid-connected system controller parameter identification method based on uncertainty evaluation and relates to the technical field of power system simulation. The method comprises the following steps: an electromagnetic transient simulation model of a photovoltaic grid-connected system is constructed to determine controller parameters to be identified; the controller parameters are grouped to obtain multiple parameter sets; parameters are selected from the parameter sets based on local sensitivity analysis to form an initial dominant parameter set; parameters in the current dominant parameter set are taken as optimization variables to perform parameter optimization to obtain a current optimal parameter solution; the controller parameters are sampled around the current optimal parameter solution to calculate uncertainty indexes of the parameters; the current dominant parameter set is dynamically updated according to the uncertainty indexes, and iteration is repeated until a termination condition is met, and controller parameter identification results are output. The dynamic focusing mechanism driven by the uncertainty indexes makes the optimization focus adaptively concentrate on the current dominant parameters, thereby improving the parameter identification precision.

Description

Technical Field

[0001] This invention relates to the field of power system simulation technology, specifically to a method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment. Background Technology

[0002] With the continuous growth of installed capacity of new energy power generation, photovoltaic grid-connected converters have become important power electronic equipment affecting the transient response of power systems. To accurately analyze the operating characteristics and fault ride-through capability of photovoltaic grid-connected systems, it is essential to establish accurate electromagnetic transient simulation models. However, in actual engineering projects, the control parameters of grid-connected photovoltaic inverters are usually controlled by the equipment manufacturers, making it difficult to obtain complete and accurate controller parameters. Furthermore, after long-term operation, factors such as control strategy tuning, changes in the operating environment, and equipment aging can lead to deviations between actual parameters and the parameters set in the simulation model, making it difficult for the simulation model to accurately reproduce the true transient response of the photovoltaic grid-connected system. Therefore, it is necessary to identify the controller parameters of the electromagnetic transient simulation model for photovoltaic grid-connected systems.

[0003] Currently, commonly used parameter identification methods include analytical methods, data-driven methods, and heuristic optimization methods. Existing parameter identification methods typically first screen key parameters through empirical judgment or one-time sensitivity analysis, and then perform subsequent identification within a fixed set of key parameters. This static screening method is difficult to reflect the changes in the degree of parameter influence during the iterative optimization process. When the optimization process enters different local regions, the influence of some parameters on the system's transient response may gradually weaken, while some parameters that were not initially selected may become important factors affecting the identification results. Therefore, the identification strategy with a fixed set of key parameters is difficult to adapt to the dynamic changes in parameter dominance during the parameter identification process of photovoltaic grid-connected systems, resulting in insufficient parameter identification accuracy. Summary of the Invention

[0004] The technical problem to be solved by this invention is how to achieve dynamic focusing of the dominant parameters in the iterative identification process of photovoltaic grid-connected system controller parameters in order to improve the accuracy of parameter identification. The purpose is to provide a photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment, which solves the above-mentioned problem.

[0005] This invention is achieved through the following technical solution:

[0006] In a first aspect, the present invention provides a method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment, comprising:

[0007] S1. Construct an electromagnetic transient simulation model of a photovoltaic grid-connected system and determine multiple controller parameters and their normalized parameter ranges;

[0008] S2. Based on the triggering conditions and scope of action of different control links, the multiple controller parameters are grouped to obtain multiple parameter sets;

[0009] S3. Construct a comprehensive fitness function based on the reference response data and the simulation response data output by the electromagnetic transient simulation model;

[0010] S4. Calculate the local sensitivity of each controller parameter according to the comprehensive fitness function, and select parameters from the multiple parameter sets according to the local sensitivity to form the first round of dominant parameter set;

[0011] S5. Let the iteration round k=1;

[0012] S6. Using the parameters in the dominant parameter set of the kth round as optimization variables and the comprehensive fitness function as the optimization objective, perform parameter optimization within the normalized parameter range corresponding to each parameter to obtain the optimal parameter solution for the kth round.

[0013] S7. Sample the multiple controller parameters based on the optimal parameter solution of the kth round, and calculate the uncertainty index of each parameter;

[0014] S8. Update the dominant parameter set for the kth round according to the uncertainty index to obtain the dominant parameter set for the (k+1)th round.

[0015] S9. Let k = k + 1, and repeat S6 to S8 until the preset iteration termination condition is met, and output the current optimal parameter solution.

[0016] Optionally, the multiple controller parameters are grouped according to the triggering conditions and scope of action of different control links to obtain multiple parameter sets, including:

[0017] Parameters that participate in control during normal operation, low voltage ride-through, and high voltage ride-through are divided into a shared parameter set;

[0018] The parameters that only participate in the control during low voltage ride-through are divided into the low voltage ride-through parameter set;

[0019] The parameters that only play a controlling role during high-voltage ride-through are classified as the high-voltage ride-through parameter set.

[0020] Optionally, the comprehensive fitness function is obtained by weighting a dynamic trajectory matching term, a key transient feature matching term, and an inner-loop dynamic characteristic matching term; wherein, the dynamic trajectory matching term is used to evaluate the overall similarity between the simulated response curve and the reference response curve; the key transient feature matching term is used to evaluate the similarity between the reference response and the simulated response for key physical features during low-voltage ride-through and high-voltage ride-through; and the inner-loop dynamic characteristic matching term is used to evaluate the similarity between the reference response and the simulated response for the inner-loop controller response characteristics during the initial stage of the fault.

[0021] Optionally, the step of calculating the local sensitivity of each controller parameter based on the comprehensive fitness function, and selecting parameters from the plurality of parameter sets according to the local sensitivity to form the first round of dominant parameter set, includes:

[0022] A perturbation is applied to the midpoint of the normalized parameter range for each controller parameter to generate multiple sets of initial parameter samples;

[0023] Calculate the comprehensive fitness function value of each set of initial parameter samples, and take the set of initial parameter samples with the highest comprehensive fitness function value as the initial point;

[0024] The local sensitivity of each controller parameter at the initial point is calculated using the central difference method;

[0025] According to the local sensitivity, all parameters in the multiple parameter sets are sorted in descending order, and the top-ranked parameters from each parameter set are selected to form the first round of dominant parameter set.

[0026] Optionally, the formula for calculating the local sensitivity is:

[0027]

[0028] In the formula, Let j be the value of the j-th parameter in the initial point; To apply to parameters The disturbance The comprehensive fitness function is... This represents the fitness value after the j-th parameter has been positively perturbed; This represents the fitness value after the j-th parameter has been negatively perturbed; This represents the local sensitivity of the j-th parameter.

[0029] Optionally, the step of using the parameters in the dominant parameter set of the k-th round as optimization variables, and the comprehensive fitness function as the optimization objective, to perform parameter optimization within the normalized parameter range corresponding to each parameter, and obtaining the optimal parameter solution for the k-th round, includes:

[0030] The comprehensive fitness function values ​​under low voltage ride-through and high voltage ride-through conditions are averaged with equal weights to construct a multi-condition comprehensive fitness function.

[0031] Parameters that do not belong to the dominant parameter set of the kth round among the multiple controller parameters are regarded as non-dominant parameters, and soft regularization constraints are applied to the non-dominant parameters.

[0032] Based on the multi-condition integrated fitness function and the soft regularization constraint term, a total loss function is constructed;

[0033] Using the parameters in the dominant parameter set of the kth round as optimization variables and the total loss function as the optimization objective, the Nelder-Mead simplex optimization algorithm is used to optimize the parameters within the normalized parameter range corresponding to each parameter, thereby obtaining the optimal parameter solution for the kth round.

[0034] Optionally, the step of sampling the multiple controller parameters based on the optimal parameter solution in the kth round and calculating the uncertainty index of each parameter includes:

[0035] Each of the multiple controller parameters is subjected to M local random perturbations;

[0036] In the m-th local random perturbation, only the value of the j-th parameter is changed, while the other parameters remain at their corresponding values ​​from the optimal parameter solution in the k-th round, thus obtaining the m-th set of perturbation parameter samples; where, ;

[0037] Based on the multi-condition integrated fitness function and the m-th group of disturbance parameter samples, calculate the change in multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance.

[0038] The uncertainty index of the j-th parameter is calculated based on the change in the comprehensive fitness of the j-th parameter under M local random disturbances.

[0039] Optionally, the formula for calculating the change in the comprehensive fitness under multiple operating conditions is as follows:

[0040]

[0041] In the formula, This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; This represents the multi-condition integrated fitness function; This represents the optimal parameter solution in the k-th round; This represents the corresponding values ​​of the remaining parameters (excluding the j-th parameter) in the optimal parameter solution of the k-th round; This represents the value of the j-th parameter after the m-th local random perturbation.

[0042] Optionally, the formula for calculating the uncertainty index is:

[0043]

[0044] In the formula, The uncertainty index of the j-th parameter is represented by M; the number of local random perturbation samplings is represented by M; and α is the weighting coefficient between the average fitness change term and the fitness fluctuation term. This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; This represents the average value of the change in the comprehensive fitness of the j-th parameter under M local random disturbances.

[0045] Optionally, updating the dominant parameter set for the kth round based on the uncertainty index to obtain the dominant parameter set for the (k+1)th round includes:

[0046] If the uncertainty index of any parameter in the dominant parameter set of round k is greater than the preset retention threshold, then the parameter is retained in the dominant parameter set of round k+1.

[0047] If the uncertainty index of any parameter in the dominant parameter set of the kth round is less than or equal to the preset retention threshold, then the parameter is removed from the dominant parameter set of the kth round.

[0048] When a parameter is removed from the dominant parameter set of round k, the parameters that do not belong to the dominant parameter set of round k among the multiple controller parameters to be identified are used as a candidate parameter pool. The parameter with the largest uncertainty index is selected from the candidate parameter pool and added to the dominant parameter set of round k+1, so that the number of dominant parameters in round k+1 is the same as that in round k.

[0049] If the difference between the maximum uncertainty index in the candidate parameter pool and the minimum uncertainty index in the dominant parameter set of the kth round is greater than a preset difference threshold, then the parameter with the minimum uncertainty index in the dominant parameter set of the (k+1)th round will be replaced with the parameter with the maximum uncertainty index in the candidate parameter pool.

[0050] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0051] This application provides a parameter identification method for photovoltaic grid-connected system controllers based on uncertainty assessment. By calculating the uncertainty index of each parameter in each iteration and dynamically updating the dominant parameter set based on the uncertainty index, the optimization focus remains on the parameters that have the most significant impact on the system response at the current stage. Compared with the static screening strategy in the prior art, this application can adapt to the dynamic changes in parameter dominance during the iteration process, thereby effectively improving the parameter identification accuracy. Furthermore, this application groups parameters according to the triggering conditions and scope of action of different control links. When selecting the dominant parameter set, it selects the top-ranked parameters from each group, rather than simply selecting all parameters globally. This avoids a certain type of parameter from dominating the set for a long time due to its numerical or sensitivity advantage, ensuring that all types of parameters are represented and enter the optimization focus during the iteration process, achieving balanced optimization of parameters in various control links. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the exemplary embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of the present invention and should not be considered as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort. In the drawings:

[0053] Figure 1 A flowchart illustrating the photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment provided in this application embodiment;

[0054] Figure 2 This is a schematic diagram of the electrical topology of a photovoltaic grid-connected power generation system provided in an embodiment of this application;

[0055] Figure 3 This is a schematic diagram of the inner loop control structure of a grid-connected inverter provided in an embodiment of this application;

[0056] Figure 4 This is a schematic diagram of the outer loop control structure of a grid-connected inverter provided in an embodiment of this application;

[0057] Figure 5 This is a voltage ride-through control logic diagram provided in an embodiment of this application. Detailed Implementation

[0058] To make the objectives, technical solutions, and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments and accompanying drawings. The illustrative embodiments and descriptions of the present invention are only used to explain the present invention and are not intended to limit the present invention.

[0059] Please refer to Figure 1This is a flowchart illustrating a photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment provided in this application embodiment. The following is a further explanation. Figure 1 The method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment is introduced.

[0060] S1. Construct an electromagnetic transient simulation model of a photovoltaic grid-connected system and determine multiple controller parameters and their normalized parameter ranges;

[0061] S2. Based on the triggering conditions and scope of action of different control links, group multiple controller parameters to obtain multiple parameter sets;

[0062] S3. Construct a comprehensive fitness function based on the reference response data and the simulation response data output by the electromagnetic transient simulation model;

[0063] S4. Calculate the local sensitivity of each controller parameter based on the comprehensive fitness function, and select parameters from multiple parameter sets based on the local sensitivity to form the first round of dominant parameter set;

[0064] S5. Let the iteration round k=1;

[0065] S6. Using the parameters in the dominant parameter set of the kth round as optimization variables and the comprehensive fitness function as the optimization objective, perform parameter optimization within the normalized parameter range corresponding to each parameter to obtain the optimal parameter solution for the kth round.

[0066] S7. Sample multiple controller parameters based on the optimal parameter solution in the kth round, and calculate the uncertainty index of each parameter;

[0067] S8. Update the dominant parameter set for round k based on the uncertainty index to obtain the dominant parameter set for round k+1.

[0068] S9. Let k = k + 1, and repeat S6 to S8 until the preset iteration termination condition is met, and output the current optimal parameter solution.

[0069] In one possible embodiment, the electrical topology of the photovoltaic grid-connected system in S1 is as follows: Figure 2 As shown, the electromagnetic transient simulation model of the photovoltaic grid-connected system includes a photovoltaic DC-side model, DC bus capacitor, grid-connected inverter, filter inductor, grid-connected transformer, equivalent AC grid, and photovoltaic grid-connected controller.

[0070] The photovoltaic grid-connected controller includes a basic dual-loop control structure, a low-voltage ride-through control loop, and a high-voltage ride-through control loop. The basic dual-loop control structure consists of an outer voltage control loop and an inner current control loop.

[0071] The inner current loop is used to achieve rapid regulation of active and reactive current components and provides support for the outer loop power control. A schematic diagram of the grid-connected inverter inner loop control structure is shown below. Figure 3 As shown, using the d-axis and q-axis current deviations as inputs, a PI controller is used to generate the outputs of the two-axis current controller, as follows:

[0072]

[0073] in, , These are the reference values ​​for the grid-connected current on the d and q axes, respectively. , These are the proportional and integral coefficients of the d-axis current loop PI controller, respectively. , These are the proportional and integral coefficients of the q-axis current loop PI controller, respectively. , These are the grid-connected current values ​​for the d and q axes, respectively. , These are the grid-connected voltage reference values ​​for the d and q axes, respectively.

[0074] The outer voltage loop generates a grid-connected current reference value based on the DC-side operating status and reactive power regulation requirements, thus providing a reference value for the inner current loop. A schematic diagram of the outer loop control structure of the grid-connected inverter is shown below. Figure 4 As shown, it can be adjusted separately. , The relationship between active and reactive power control is as follows:

[0075]

[0076] in, This is the reference value for the DC bus voltage; This is a reference value for reactive power. , These are the proportional coefficient and time constant of the DC voltage outer-loop PI controller, respectively; , These are the proportional coefficient and time constant of the reactive power outer loop PI controller, respectively; Q represents the DC bus voltage; Q represents the reactive power.

[0077] When a voltage dip / surge fault occurs in the power grid, the current reference value under normal operating conditions is insufficient to simultaneously meet the requirements of continuous grid-connected operation, voltage support, and overcurrent limiting. Therefore, the grid-connected inverter control system also needs to be configured with High-Voltage Ride-Through (HVRT) and Low-Voltage Ride-Through (LVRT) to reconstruct the current reference value of the outer loop output under fault conditions. Its logic structure is as follows: Figure 5 As shown.

[0078] When the voltage amplitude at the grid connection point Below the set threshold When this occurs, LVRT control is triggered. At this time, the reactive current reference value is increased to enhance reactive power support during the fault period, while the active current reference value is suppressed. The reactive current reference value under LVRT conditions can be expressed as follows:

[0079]

[0080] in, It represents the reactive voltage deviation gain under low voltage ride-through conditions and determines the support strength; It represents the reactive current proportionality coefficient under low voltage ride-through conditions and participates in the amplitude adjustment and limiting of the final command. It is a saturation function; , These represent the dynamic upper and lower limits of reactive current, respectively. The voltage threshold at the grid connection point; This is the voltage at the grid connection point.

[0081] For the active current reference value, firstly, based on the active current proportionality coefficient... Determine its target value, and then restore the slope constraint. To achieve dynamic adjustment, the expression is:

[0082] in, This represents the target value of the active current along the d-axis. This indicates the active current command value before the fault; Indicates the adjustment time constant; , These represent the dynamic upper and lower limits of the active current, respectively; Restore the slope constraint value for LVRT.

[0083] When the voltage amplitude at the grid connection point Higher than the set threshold When this occurs, HVRT control is triggered. At this time, the reactive current reference value is also adjusted according to the voltage deviation, and its expression is as follows:

[0084]

[0085] in, The reactive voltage deviation gain under high voltage ride-through conditions; It is the reactive current proportional coefficient under high voltage ride-through conditions, used to adjust and limit the amplitude of reactive current commands. This indicates the reactive current command value before the fault.

[0086] The target value of active current in the initial stage of HVRT fault is determined by the initial active current coefficient. The rate of change during the recovery phase is determined by the current-limiting recovery slope. The constraints are as follows:

[0087]

[0088] in, These are the initial active value coefficients; To restore the slope constraint value for HVRT; , These represent the dynamic upper and lower limits of the active current, respectively.

[0089] The multiple controller parameters to be identified include voltage outer loop PI control parameters, current inner loop PI control parameters, active current constraint parameters, reactive current support parameters, and current recovery control parameters during low voltage ride-through, as well as active current regulation parameters, reactive current absorption parameters, and current recovery control parameters during high voltage ride-through.

[0090] The identification range of each controller parameter is determined based on the engineering recommended values, controller physical constraints, and parameter feasible region. Each parameter is then normalized to the [0,1] interval to obtain the normalized parameter range. The normalization formula is as follows:

[0091]

[0092] in, This represents the actual value of the j-th controller parameter. This represents the normalized value of the j-th controller parameter. This shows the lower limit value of the j-th controller parameter. This represents the upper limit value of the j-th controller parameter.

[0093] In one possible embodiment, step S2 includes:

[0094] Parameters that participate in control during normal operation, low voltage ride-through, and high voltage ride-through are classified into a shared parameter set; parameters that participate in control only during low voltage ride-through are classified into a low voltage ride-through parameter set; and parameters that participate in control only during high voltage ride-through are classified into a high voltage ride-through parameter set.

[0095] The shared parameter set includes voltage outer-loop PI control parameters and current inner-loop PI control parameters. The low-voltage ride-through parameter set includes active current constraint parameters, reactive current support parameters, and current recovery control parameters during low-voltage ride-through. The high-voltage ride-through parameter set includes active current regulation parameters, reactive current absorption parameters, and current recovery control parameters during high-voltage ride-through.

[0096] In the embodiments of this application, sensitivity analysis, uncertainty assessment and dominant parameter set update can be carried out according to the control range of different parameters, thereby avoiding confusion between low voltage ride-through parameters, high voltage ride-through parameters and shared parameters, and improving the physical interpretability and computational efficiency of the parameter identification process.

[0097] In one possible embodiment, the reference response data in S3 refers to the response output data of the photovoltaic grid-connected system obtained from the actual system under specific fault conditions (such as low-voltage ride-through or high-voltage ride-through), which serves as the identification benchmark. The simulation response data refers to the system response data output by the electromagnetic transient simulation model after substituting the current values ​​(or a set of parameter samples) of multiple controller parameters to be identified into the electromagnetic transient simulation model. Based on the reference response data and the simulation response data, a comprehensive fitness function is constructed to evaluate the degree of fit between the simulation model output and the actual system response under the current parameter values.

[0098] In one possible embodiment, the comprehensive fitness function is obtained by weighting the dynamic trajectory matching term, the key transient feature matching term, and the inner loop dynamic characteristic matching term, and its expression is:

[0099]

[0100] in, Indicates dynamic trajectory matching items. Indicates key transient feature matching terms. This represents the inner loop dynamic feature matching term. , , Let be the weight coefficients of each matching item, satisfying... .

[0101] To balance global trajectory matching and key transient feature capture, this embodiment employs a fixed weight strategy. Based on preliminary experimental analysis, the weights of each matching item are set as follows: , , 20. This configuration ensures that the optimization process is effectively guided by both macroscopic dynamic trends and key inflection points in fault response. The matching items are described below.

[0102] (1) Dynamic trajectory matching term: used to evaluate the overall similarity between the simulated response curve and the reference response curve.

[0103] The dynamic trajectory matching term is constructed based on the Pearson correlation coefficient and the normalized root mean square error similarity. The Pearson correlation coefficient characterizes the consistency between the simulated and reference response curves in terms of shape change trends, while the normalized root mean square error similarity characterizes the consistency between the simulated and reference response curves in terms of amplitude error. The calculation formula for the dynamic trajectory matching term is as follows:

[0104]

[0105] In the formula, Indicates the number of selected response metrics. This represents the Pearson correlation coefficient operator. The normalized root mean square error similarity is represented by t; time is represented by t. Indicates the values ​​of the controller parameters; This indicates the simulated response curve at time t and parameters. The value to be taken below; This indicates the value of the reference response curve at time t.

[0106] (2) Key transient feature matching term: used to evaluate the similarity between the reference response and the simulation response of key physical features during low voltage ride-through and high voltage ride-through.

[0107] For two typical fault scenarios, low-voltage ride-through (LVRT) and high-voltage ride-through (HVRT), key physical features characterizing controller performance are extracted, and their similarity is calculated. For the LVRT process, key physical features extracted include the voltage minimum point, voltage recovery amplitude, peak reactive power, reactive power rise time, voltage recovery slope, and steady-state reactive power. For the HVRT process, key physical features extracted include DC voltage overshoot, active power sag amplitude, DC voltage settling time, active power recovery rate, and reactive power response range.

[0108] The key transient feature matching term uses an improved normalized relative error form to calculate feature similarity, as shown in the following formula:

[0109]

[0110] In the formula, Represents the simulated value of the i-th feature; This represents the reference value of the i-th feature; N is the total number of key physical features.

[0111] (3) Inner loop dynamic characteristic matching term: used to evaluate the similarity between the reference response and the simulation response of the inner loop controller in the initial stage of the fault.

[0112] Regarding the inner-loop dynamic characteristics of the dual-loop controller, this embodiment extracts two key inner-loop dynamic features: the initial slope of the DC voltage change. and active power rise time These two characteristics reflect the rapid response of the inner current loop and are of great significance for evaluating the dynamic performance of the controller in the initial stage of a fault.

[0113] Initial slope of DC voltage change Defined as the average slope of the DC voltage within a preset short time interval after a fault occurs, it is used to reflect the initial suppression speed of the inner current loop against voltage disturbances. Its calculation formula is as follows:

[0114]

[0115] In the formula, The time when the fault occurred. The sampling interval is... This represents the number of sampling points; This indicates the value of the DC bus voltage at the corresponding moment.

[0116] active power rise time Defined as the time required for active power to recover from its lowest point to 90% of its steady-state value during a fault, it reflects the recovery speed of active power under inner-loop control. Its calculation formula is as follows:

[0117]

[0118] in, The value of active power at time t; This represents the steady-state value of active power.

[0119] The average similarity of each feature is taken as the inner loop dynamic characteristic matching term, and the calculation formula is as follows:

[0120]

[0121] in, The similarity of the initial slope of the DC voltage change; This represents the similarity of active power rise time.

[0122] Existing technologies typically evaluate parameters based solely on the overall error of the response curve, making it difficult to distinguish the impact of different control elements on the system response. In this embodiment, the fitness function is divided into three levels: a dynamic trajectory matching term evaluates curve similarity based on overall trend and amplitude accuracy; a key transient feature matching term focuses on key physical characteristics during fault traversal; and an inner-loop dynamic characteristic matching term focuses on the rapid response characteristics of the inner-loop controller in the initial stage of a fault. These three dimensions, from macroscopic to local, and from steady-state to transient, provide a comprehensive characterization of the system response characteristics, enabling the comprehensive fitness function to more fully reflect the impact of controller parameters on the system's dynamic behavior.

[0123] In one possible embodiment, step S4 includes:

[0124] A perturbation is applied to the midpoint of the normalized parameter range for each controller parameter to generate multiple sets of initial parameter samples. The comprehensive fitness function value of each set of initial parameter samples is calculated, and the set of initial parameter samples with the highest comprehensive fitness function value is taken as the initial point. The local sensitivity of each controller parameter in the initial point is calculated using the central difference method. All parameters in multiple parameter sets are sorted in descending order according to the magnitude of local sensitivity, and the top-ranked parameters from each parameter set are selected to form the first round of dominant parameter set.

[0125] In practical implementation, the formula for calculating local sensitivity is:

[0126]

[0127] In the formula, Let j be the value of the j-th parameter in the initial point; To apply to parameters The disturbance; This is the overall fitness function; This represents the fitness value after the j-th parameter has been positively perturbed; This represents the fitness value after the j-th parameter has been negatively perturbed; This represents the local sensitivity of the j-th parameter.

[0128] For the shared parameter, its local sensitivity under low voltage ride-through conditions and high voltage ride-through conditions is calculated separately, and the two are combined with equal weights to obtain the comprehensive local sensitivity of the shared parameter, which accurately reflects its dominance under all operating conditions.

[0129] Finally, based on local sensitivity, parameters ranked high in the shared parameter set, low-voltage ride-through parameter set, and high-voltage ride-through parameter set are selected to form corresponding dominant parameter subsets, which are then merged to obtain the first round of dominant parameter subsets. .

[0130]

[0131] in, Indicates the dominant subset of shared parameters. This represents the dominant subset of low-voltage ride-through parameters. This represents the dominant subset of high-voltage ride-through parameters.

[0132] In this embodiment, by sampling initial parameters and evaluating overall fitness, the parameter combination with the highest fitness is selected from multiple random samples as the initial optimization point, avoiding slow convergence or getting trapped in local optima due to poor initial point selection. The local sensitivity of each parameter near the initial point is calculated using the central difference method, quantitatively evaluating the influence of each parameter on the system response, and thus selecting key parameters, avoiding subjectivity based on experience.

[0133] In one possible embodiment, step S6 includes:

[0134] The comprehensive fitness function values ​​under low-voltage ride-through and high-voltage ride-through conditions are averaged with equal weights to construct a multi-condition comprehensive fitness function. Parameters that do not belong to the dominant parameter set of the k-th round are treated as non-dominant parameters, and soft regularization constraints are applied to these non-dominant parameters. Based on the multi-condition comprehensive fitness function and the soft regularization constraint, a total loss function is constructed. Using the parameters in the dominant parameter set of the k-th round as optimization variables and the total loss function as the optimization objective, the Nelder-Tead simplex optimization algorithm is used to optimize the parameters within the normalized parameter range corresponding to each parameter, thus obtaining the optimal parameter solution for the k-th round.

[0135] In the specific implementation process, to avoid the parameter identification results being biased towards a single fault type, a multi-scenario weighted mechanism is further introduced. The comprehensive fitness function values ​​for low-voltage ride-through and high-voltage ride-through conditions are calculated separately, and then equally weighted averaged to construct the multi-condition comprehensive fitness function as follows:

[0136]

[0137] in, This represents the overall fitness value across multiple operating conditions. This represents the overall fitness function value under low voltage ride-through conditions. This represents the overall fitness function value under high voltage ride-through conditions.

[0138] When k=1, the parameters in the first round of dominant parameter set are used as optimization variables, the multi-condition integrated fitness function is used as the optimization objective, and the Nelder-Tead simplex optimization algorithm is used to optimize the parameters in the normalized parameter range corresponding to each parameter to obtain the optimal parameter solution for the first round.

[0139] After the first iteration, parameters that do not belong to the dominant parameter set of the first iteration are treated as non-dominant parameters. Soft regularization constraints are applied to these non-dominant parameters to keep them near the engineering recommended values. The expressions for the non-dominant parameters are as follows:

[0140]

[0141] In the formula, This represents the i-th non-dominant parameter; It is represented as the soft regularization constraint term imposed by the i-th non-dominant parameter; Represents the regularization coefficient; This represents the recommended engineering value for the i-th non-dominant parameter. This represents the range of values ​​for the i-th non-dominant parameter; This represents the dominant parameter set for the k-th round.

[0142] Based on the multi-condition integrated fitness function and soft regularization constraints, the total loss function is constructed as follows:

[0143]

[0144] This is the total loss function; This is a multi-condition integrated fitness function; This is a soft regularization constraint.

[0145] When k>1, the parameters in the dominant parameter set of the kth round are used as optimization variables, and the total loss function is used as the optimization objective. The Nelder-Tead simplex optimization algorithm is employed to optimize the parameters within the normalized parameter range corresponding to each parameter, thus obtaining the optimal parameter solution for the kth round. .

[0146] In this embodiment, considering the balance between computational cost and accuracy of parameter identification, the Nelder-Mead simplex optimization algorithm is used to optimize the parameters in the current dominant parameter set. Let the number of parameters in the current dominant parameter set be n. Then, an initial simplex consisting of n+1 vertices is constructed in the normalized parameter space, and the vertices of the simplex are continuously updated through reflection, expansion, contraction, and compression operations to search for the optimal parameter combination corresponding to the current dominant parameter set.

[0147] Among them, the reflection operation is used to generate reflection points to explore new search directions; the expansion operation is used to further expand the search range when the reflection direction shows an improved effect; the contraction operation is used to contract the simplex when the effect of the reflection point is average; and the compression operation is used to contract the simplex as a whole towards the current optimum when the effect of the reflection direction is poor.

[0148] The main parameter configurations for the Nelder-Mead simplex optimization algorithm are shown in Table 1:

[0149] Table 1. Parameter configuration of Nelder-Mead simplex optimization algorithm

[0150] Reflectance coefficient 1.00 Expansion coefficient 2.00 Shrinkage coefficient 0.50 Compression coefficient 0.50 Maximum number of function evaluations max{180, 60(n+1)}

[0151] Each objective function evaluation requires performing electromagnetic transient simulations under both low-voltage ride-through and high-voltage ride-through conditions, and calculating the multi-condition comprehensive fitness function based on the simulation responses under both conditions. Therefore, to avoid excessive computational load during a single round of parameter optimization, this embodiment sets an early stopping condition in the Nelder-Mead simplex optimization algorithm. When any of the following conditions are met, the parameter optimization for the current round is terminated early, and the current optimal parameter solution for that iteration is output:

[0152] (1) If the size of the simplex is less than the preset threshold, it indicates that the vertex of the simplex has shrunk to the vicinity of a local optimum;

[0153] (2) The current round of function evaluation reaches the maximum function evaluation count max{180,60(n+1)}.

[0154] In this embodiment, a multi-condition comprehensive fitness function is constructed by weighted averaging of fitness values ​​under both low-voltage ride-through and high-voltage ride-through fault scenarios. This avoids bias in the identification results towards a single fault type and ensures the generalization performance of parameters under various power grid fault scenarios. By applying soft regularization constraints to non-dominant parameters that do not belong to the current dominant parameter set, the meaningless drift of non-dominant parameters under the nonlinear coupling effect of the model is effectively suppressed, reducing the compensatory error between dominant and non-dominant parameters and ensuring the credibility of the final identification results. By combining the multi-condition comprehensive fitness function with the soft regularization constraint term to form the total loss function, parameter optimization aims to improve simulation accuracy and maintain parameter rationality simultaneously, achieving synergistic optimization of fitting accuracy and parameter rationality.

[0155] In one possible embodiment, step S7 includes:

[0156] For each parameter among multiple controller parameters, apply M local random perturbations; in the m-th local random perturbation, only change the value of the j-th parameter, while the other parameters remain at their corresponding values ​​in the optimal parameter solution of the k-th round, thus obtaining the m-th set of perturbation parameter samples; where, Based on the multi-condition integrated fitness function and the m-th set of disturbance parameter samples, calculate the change in multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; calculate the uncertainty index of the j-th parameter based on the change in multi-condition integrated fitness caused by the j-th parameter under M local random disturbances.

[0157] In the specific implementation process, the calculation formula for the change in comprehensive fitness under multiple working conditions is as follows:

[0158]

[0159] In the formula, This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; Represents the multi-condition integrated fitness function; This represents the optimal parameter solution in the k-th round; This represents the corresponding values ​​of the remaining parameters (excluding the j-th parameter) in the optimal parameter solution of the k-th round; This represents the value of the j-th parameter after the m-th local random perturbation.

[0160] Based on the changes in the comprehensive fitness under multiple operating conditions obtained from M disturbances, the uncertainty index of the j-th parameter is calculated using the following formula:

[0161]

[0162] In the formula, The uncertainty index of the j-th parameter is represented by M; the number of local random perturbation samplings is represented by M; and α is the weighting coefficient between the average fitness change term and the fitness fluctuation term. This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; This represents the average value of the change in the comprehensive fitness of the j-th parameter under M local random disturbances.

[0163] The uncertainty index is used to evaluate the degree of influence of parameters on the multi-condition integrated fitness function in the neighborhood of the current optimal solution. The larger the value, the more significant the impact of this parameter on the fault transient response fitting results in the current optimization stage; The smaller the value, the weaker the influence of this parameter on the change of the fitness function in the current optimization stage.

[0164] In this embodiment, multiple local random perturbations are applied to each parameter around the optimal parameter solution in the k-th round. The change in fitness caused by each parameter in the neighborhood of the current optimal solution is quantitatively calculated, and an uncertainty index is constructed based on this. This objectively reflects the degree of influence of each parameter on the system response in the current optimization stage, providing a scientific and quantifiable decision-making basis for the dynamic updating of the dominant parameter set and avoiding the subjectivity of relying on experience-based judgments. Simultaneously, the uncertainty index comprehensively characterizes the overall influence intensity and nonlinear characteristics of the parameters by weightedly combining the average change in fitness caused by parameter perturbations and the degree of fitness fluctuation, making uncertainty assessment more accurate.

[0165] In one possible embodiment, step S8 includes:

[0166] If any parameter in the dominant parameter set of the kth round The uncertainty index is greater than the preset retention threshold τ low If so, then this parameter will be retained in the dominant parameter set for the (k+1)th round;

[0167] If any parameter in the dominant parameter set of the kth round The uncertainty index is less than or equal to the preset retention threshold τ low If so, then remove the parameter from the set of dominant parameters in round k;

[0168] When a parameter is removed from the dominant parameter set of round k, the parameters that do not belong to the dominant parameter set of round k among the multiple controller parameters to be identified are used as a candidate parameter pool. The parameter with the largest uncertainty index is selected from the candidate parameter pool and added to the dominant parameter set of round k+1, so that the number of dominant parameters in round k+1 is the same as that in round k.

[0169] If the difference between the maximum uncertainty index in the candidate parameter pool and the minimum uncertainty index in the dominant parameter set of the kth round is greater than the preset difference threshold, then the parameter with the minimum uncertainty index in the dominant parameter set of the (k+1)th round will be replaced with the parameter with the maximum uncertainty index in the candidate parameter pool.

[0170] In the actual implementation process, the degree of difference exceeding the preset difference threshold includes the following two situations:

[0171] In the first case, the difference between the maximum uncertainty index in the candidate parameter pool and the minimum uncertainty index in the dominant parameter set of the kth round is greater than the preset replacement threshold.

[0172] In the second scenario, the ratio between the maximum uncertainty index in the candidate parameter pool and the minimum uncertainty index in the dominant parameter set of the kth round is greater than a preset ratio threshold.

[0173] In this embodiment, the retention, removal, addition, and replacement of the dominant parameter set are dynamically implemented based on the uncertainty index of each parameter in the neighborhood of the current optimal solution. This ensures that the dominant parameter set always focuses on the parameters that have the most significant impact on the system response at the current stage, avoiding the problem of decreased identification accuracy caused by a fixed parameter set. By adding parameters from the candidate parameter pool of the same control link category, the proportion of various control parameters in the dominant parameter set remains stable during the iteration process, avoiding the problem of long-term absence of a certain type of parameter due to its exit from the dominant set. Through the active replacement mechanism, when there is an uncertainty index in the candidate parameter pool that is significantly better than the weakest parameter in the dominant set, replacement is performed, further improving the quality of the dominant parameter set.

[0174] In one possible embodiment, the preset iteration termination condition in S9 includes the following conditions:

[0175] The first condition, the stability condition of the dominant parameter set: the composition of the dominant parameter set no longer changes in several consecutive iterations.

[0176] The second condition, the fitness function convergence condition, is: the change in the multi-condition integrated fitness function value obtained from two adjacent iterations is less than the preset fitness convergence threshold.

[0177] The third condition, the maximum number of iterations, is: the current number of iterations reaches the preset maximum number of iterations.

[0178] When any one of the following conditions is met—the stability condition of the dominant parameter set, the convergence condition of the fitness function, or the maximum number of iterations—the parameter identification iteration process terminates, and the current optimal parameter solution is output as the controller parameter identification result of the photovoltaic grid-connected system.

[0179] Based on the same inventive concept, this application also provides a computer device, which includes a processor, a memory, and a computer program stored in the memory. The computer program is executed by the processor to implement the aforementioned method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment.

[0180] Based on the same inventive concept, this application also provides a computer storage medium storing a computer program, which is executed by a processor to implement the aforementioned method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment.

[0181] In some embodiments, the computer-readable storage medium may be a memory such as FRAM, ROM, PROM, EPROM, EEPROM, flash memory, magnetic surface memory, optical disk, or CD-ROM; or it may be a device including one or any combination of the above-mentioned memories. The computer may be a variety of computing devices, including smart terminals and servers.

[0182] In some embodiments, executable instructions may take the form of a program, software, software module, script, or code, written in any form of programming language (including compiled or interpreted languages, or declarative or procedural languages), and may be deployed in any form, including as a standalone program or as a module, component, subroutine, or other unit suitable for use in a computing environment.

[0183] As an example, executable instructions may, but do not necessarily, correspond to files in the file system. They may be stored as part of a file that holds other programs or data, for example, in one or more scripts in a Hyper Text Markup Language (HTML) document, in a single file dedicated to the program in question, or in multiple collaborative files (e.g., a file that stores one or more modules, subroutines, or code sections).

[0184] As an example, executable instructions can be deployed to execute on a single computing device, or on multiple computing devices located in one location, or on multiple computing devices distributed across multiple locations and interconnected via a communication network.

[0185] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0186] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying controller parameters of a photovoltaic grid-connected system based on uncertainty assessment, characterized in that, include: S1. Construct an electromagnetic transient simulation model of a photovoltaic grid-connected system and determine multiple controller parameters and their normalized parameter ranges; S2. Based on the triggering conditions and scope of action of different control links, the multiple controller parameters are grouped to obtain multiple parameter sets; S3. Construct a comprehensive fitness function based on the reference response data and the simulation response data output by the electromagnetic transient simulation model; S4. Calculate the local sensitivity of each controller parameter according to the comprehensive fitness function, and select parameters from the multiple parameter sets according to the local sensitivity to form the first round of dominant parameter set; S5. Let the iteration round k=1; S6. Using the parameters in the dominant parameter set of the kth round as optimization variables and the comprehensive fitness function as the optimization objective, perform parameter optimization within the normalized parameter range corresponding to each parameter to obtain the optimal parameter solution for the kth round. S7. Sample the multiple controller parameters based on the optimal parameter solution of the kth round, and calculate the uncertainty index of each parameter; S8. Update the dominant parameter set for the kth round according to the uncertainty index to obtain the dominant parameter set for the (k+1)th round. S9. Let k = k + 1, and repeat S6 to S8 until the preset iteration termination condition is met, and output the current optimal parameter solution.

2. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 1, characterized in that, The multiple controller parameters are grouped according to the triggering conditions and scope of action of different control links to obtain multiple parameter sets, including: Parameters that participate in control during normal operation, low voltage ride-through, and high voltage ride-through are divided into a shared parameter set; The parameters that only participate in the control during low voltage ride-through are divided into the low voltage ride-through parameter set; The parameters that only play a controlling role during high-voltage ride-through are classified as the high-voltage ride-through parameter set.

3. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 1, characterized in that, The comprehensive fitness function is obtained by weighting the dynamic trajectory matching term, the key transient feature matching term, and the inner loop dynamic characteristic matching term; wherein, the dynamic trajectory matching term is used to evaluate the overall similarity between the simulated response curve and the reference response curve; the key transient feature matching term is used to evaluate the similarity between the reference response and the simulated response for key physical features during low voltage ride-through and high voltage ride-through; and the inner loop dynamic characteristic matching term is used to evaluate the similarity between the reference response and the simulated response for the inner loop controller response characteristics during the initial stage of the fault.

4. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 1, characterized in that, The step involves calculating the local sensitivity of each controller parameter based on the comprehensive fitness function, and selecting parameters from the multiple parameter sets based on the local sensitivity to form the first round of dominant parameter set, including: A perturbation is applied to the midpoint of the normalized parameter range for each controller parameter to generate multiple sets of initial parameter samples; Calculate the comprehensive fitness function value of each set of initial parameter samples, and take the set of initial parameter samples with the highest comprehensive fitness function value as the initial point; The local sensitivity of each controller parameter at the initial point is calculated using the central difference method; According to the local sensitivity, all parameters in the multiple parameter sets are sorted in descending order, and the top-ranked parameters from each parameter set are selected to form the first round of dominant parameter set.

5. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 4, characterized in that, The formula for calculating the local sensitivity is: ； In the formula, Let j be the value of the j-th parameter in the initial point; To apply to parameters The disturbance The comprehensive fitness function is... This represents the fitness value after the j-th parameter has been positively perturbed; This represents the fitness value after the j-th parameter has been negatively perturbed; This represents the local sensitivity of the j-th parameter.

6. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 1, characterized in that, The optimization process involves using the parameters in the dominant parameter set of the k-th round as optimization variables and the comprehensive fitness function as the optimization objective. Parameter optimization is performed within the normalized parameter range corresponding to each parameter to obtain the optimal parameter solution for the k-th round, including: The comprehensive fitness function values ​​under low voltage ride-through and high voltage ride-through conditions are averaged with equal weights to construct a multi-condition comprehensive fitness function. Parameters that do not belong to the dominant parameter set of the kth round among the multiple controller parameters are regarded as non-dominant parameters, and soft regularization constraints are applied to the non-dominant parameters. Based on the multi-condition integrated fitness function and the soft regularization constraint term, a total loss function is constructed; Using the parameters in the dominant parameter set of the kth round as optimization variables and the total loss function as the optimization objective, the Nelder-Mead simplex optimization algorithm is used to optimize the parameters within the normalized parameter range corresponding to each parameter, thereby obtaining the optimal parameter solution for the kth round.

7. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 6, characterized in that, The step of sampling the multiple controller parameters based on the optimal parameter solution in the kth round and calculating the uncertainty index of each parameter includes: Each of the multiple controller parameters is subjected to M local random perturbations; In the m-th local random perturbation, only the value of the j-th parameter is changed, while the other parameters remain at their corresponding values ​​from the optimal parameter solution in the k-th round, thus obtaining the m-th set of perturbation parameter samples; where, ; Based on the multi-condition integrated fitness function and the m-th group of disturbance parameter samples, calculate the change in multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance. The uncertainty index of the j-th parameter is calculated based on the change in the comprehensive fitness of the j-th parameter under M local random disturbances.

8. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 7, characterized in that, The formula for calculating the change in comprehensive fitness under multiple operating conditions is as follows: ； In the formula, This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; This represents the multi-condition integrated fitness function; This represents the optimal parameter solution in the k-th round; This represents the corresponding values ​​of the remaining parameters (excluding the j-th parameter) in the optimal parameter solution of the k-th round; This represents the value of the j-th parameter after the m-th local random perturbation.

9. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 7, characterized in that, The formula for calculating the uncertainty index is as follows: ； In the formula, The uncertainty index of the j-th parameter is represented by M; the number of local random perturbation samplings is represented by M; and α is the weighting coefficient between the average fitness change term and the fitness fluctuation term. This represents the change in the multi-condition integrated fitness caused by the j-th parameter under the m-th local random disturbance; This represents the average value of the change in the comprehensive fitness of the j-th parameter under M local random disturbances.

10. The photovoltaic grid-connected system controller parameter identification method based on uncertainty assessment according to claim 1, characterized in that, The step of updating the dominant parameter set for the kth round based on the uncertainty index to obtain the dominant parameter set for the (k+1)th round includes: If the uncertainty index of any parameter in the dominant parameter set of round k is greater than the preset retention threshold, then the parameter is retained in the dominant parameter set of round k+1. If the uncertainty index of any parameter in the dominant parameter set of the kth round is less than or equal to the preset retention threshold, then the parameter is removed from the dominant parameter set of the kth round. When a parameter is removed from the dominant parameter set of round k, the parameters that do not belong to the dominant parameter set of round k among the multiple controller parameters to be identified are used as a candidate parameter pool. The parameter with the largest uncertainty index is selected from the candidate parameter pool and added to the dominant parameter set of round k+1, so that the number of dominant parameters in round k+1 is the same as that in round k. If the difference between the maximum uncertainty index in the candidate parameter pool and the minimum uncertainty index in the dominant parameter set of the kth round is greater than a preset difference threshold, then the parameter with the minimum uncertainty index in the dominant parameter set of the (k+1)th round will be replaced with the parameter with the maximum uncertainty index in the candidate parameter pool.

Images