A method for identifying harmonic risk and inhibiting mis-switching of an intelligent metal comprehensive distribution box

Abstract

The present application belongs to the technical field of low-voltage reactive power compensation. A method for identifying harmonic risk and inhibiting misoperation of an intelligent metal comprehensive distribution box is disclosed. First, the present application collects harmonic-related operating state information and constructs a harmonic risk feature vector. Then, a light gradient boosting classification model is used to estimate the basic risk probability of future short-term harmonic risk. Next, the harmonic level is corrected at the combination level according to the basic risk probability and the current total harmonic distortion rate, combined with the equivalent compensation capacity of the candidate compensation combination, to generate a set of allowed branches. Finally, the comprehensive control cost is calculated in the allowed set, and the combination with the minimum cost is selected to execute the switching. The present application jointly uses the future harmonic risk probability, the candidate combination harmonic level estimate, the zero sequence current, the branch health state and the compensation capacity for switching decision, which can effectively reduce the number of harmonic over-limit windows and the number of misoperation actions, and can balance the inhibition effect and compensation performance under harmonic pollution conditions.

Description

Technical Field

[0001] This invention belongs to the field of harmonic risk control technology for low-voltage complete reactive power compensation devices, and particularly relates to a method for harmonic risk identification and erroneous switching suppression in intelligent metal integrated distribution boxes. It is applicable to JP intelligent metal integrated distribution boxes equipped with intelligent harmonic suppression reactive power compensation modules, composite switches and distribution monitoring and compensation controllers. Background Technology

[0002] In addition to its functions of power distribution and reactive power compensation within the transformer substation, the JP intelligent metal integrated distribution box also has harmonic suppression capabilities. However, in practical engineering applications, harmonic conditions are not constant. When the proportion of nonlinear loads increases, the zero-sequence current in the transformer substation increases, or individual branches age, conventional compensation strategies may still continue to apply compensation to the branches based on a fixed power factor threshold, leading to the following problems:

[0003] First, the large-capacity compensation investment in harmonic pollution scenarios may cause further amplification of harmonic currents, leading to an increase in the harmonic over-limit window.

[0004] Second, conventional control only focuses on the power factor before and after compensation, while ignoring whether there are high-risk periods that are not suitable for operation, which can easily lead to erroneous switching.

[0005] Third, for branches that already show signs of aging, frequent operation in scenarios with strong harmonics will further increase the damage to high-risk branches.

[0006] Fourth, there is a coupling relationship between harmonic risk, reactive power demand, zero-sequence current and branch health status. It is difficult to accurately distinguish between truly dangerous periods and periods that can be compensated normally based on a single THD threshold.

[0007] Therefore, there is an urgent need to provide a method that can simultaneously identify harmonic risks and suppress accidental switching in JP intelligent metal integrated distribution boxes. Summary of the Invention

[0008] To overcome the problems of existing compensation control methods that rely solely on fixed power factor or a single THD threshold for switching in harmonic pollution scenarios, easily leading to erroneous switching and inappropriate connection of high-capacity branches, this invention proposes a method for harmonic risk identification and erroneous switching suppression in intelligent metal integrated distribution boxes. This method first estimates the basic risk probability of high-risk harmonic events in the next 8 minutes based on the total harmonic distortion rate, characteristic subharmonics, zero-sequence current, reactive power demand, and branch health status over the past 30 minutes at each control moment. Then, it performs candidate combination-level correction of the harmonic level by combining the equivalent compensation capacity of candidate compensation combinations, generating a set of branches that can be connected. Within this set, it selects the combination with the lowest overall control cost, thereby achieving integrated control of harmonic risk identification and erroneous switching suppression.

[0009] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0010] A method for harmonic risk identification and erroneous switching suppression in an intelligent metal integrated distribution box includes the following steps: S1. Collect harmonic-related operating status information; S2. Construct a harmonic risk feature vector based on the collected operational status information; S3. Input the harmonic risk feature vector into the lightweight gradient boosting classification model to estimate the basic risk probability of a high-risk harmonic event occurring in the near future. S4. Based on the basic risk probability and the current total harmonic current distortion rate, the candidate compensation branch combinations are screened to generate a set of branches that can be put into operation. S5. Within the set of allowed input branches, calculate the comprehensive control cost of each candidate compensation branch combination, and select the candidate combination with the smallest comprehensive control cost as the current execution plan to perform erroneous input suppression control. S6. Output harmonic risk results and control recommendations.

[0011] Furthermore, the operating status information collected in S1 includes: total harmonic current distortion rate, fifth harmonic current component, seventh harmonic current component, zero-sequence current, active power, reactive power, power factor, main circuit current, three-phase imbalance, current compensation branch activation status, the most recent operation time of each branch, the rated compensation capacity of each branch, the actual response compensation capacity of each branch after the most recent successful activation, and ambient temperature.

[0012] Furthermore, the construction of the harmonic risk feature vector in S2 further includes: based on the harmonic and load sequences within the most recent 30-minute historical window, constructing the following features according to a 1-minute sampling period: current total harmonic current distortion rate feature, mean feature of total harmonic current distortion rate of the most recent 5 sampling points, current fifth harmonic current component feature, current seventh harmonic current component feature, current zero-sequence current feature, mean feature of zero-sequence current of the most recent 5 sampling points, current target reactive power compensation demand feature, mean feature of target reactive power compensation demand of the most recent 5 sampling points, current power factor, current main circuit current, current three-phase imbalance, ambient temperature, and maximum branch health status quantity.

[0013] Furthermore, the basic risk probability of a high-risk harmonic event occurring in the near future in S3 refers to the probability of a high-risk harmonic event occurring within the next 8-minute window; the high-risk harmonic event is defined as meeting any of the following conditions at any time within the subsequent 8 sampling points: Condition 1: The total harmonic current distortion rate is greater than or equal to 6.6%, and meets either "target reactive power compensation demand greater than or equal to 95 kvar" or "zero-sequence current greater than or equal to 100 A"; Condition 2: The total harmonic current distortion rate is greater than or equal to 5.6%, and the maximum branch health status quantity is greater than or equal to 1.30, while the target reactive power compensation demand is greater than or equal to 110 kvar.

[0014] Furthermore, the lightweight gradient boosting classification model in S3 consists of 60 to 100 CART classification trees, each with a maximum depth of 2 to 4, a learning rate of 0.03 to 0.10, a minimum number of leaf node samples of 3 to 10, and a model storage footprint of no more than 256 kilobytes.

[0015] Furthermore, the generation of the allowed input branch set in S4 specifically includes: for any candidate combination, first calculate its equivalent compensation capacity and the currently invested equivalent compensation capacity, then calculate the harmonic level estimate and combination-level harmonic risk probability corresponding to the candidate combination; include candidate combinations that meet the condition that the combination-level harmonic risk probability is less than the first threshold and the harmonic level estimate is less than the allowed harmonic threshold into the directly allowed set; include candidate combinations that meet the condition that the combination-level harmonic risk probability is between the first threshold and the second threshold and the equivalent compensation capacity is not greater than the capacity limit threshold into the capacity-limited allowed set; and list the remaining candidate combinations as prohibited or delayed input targets.

[0016] Furthermore, the estimated harmonic level corresponding to the candidate combination is calculated based on the current total harmonic current distortion rate, the increase in the equivalent compensation capacity of the candidate combination relative to the currently implemented equivalent compensation capacity, the maximum branch health status, the zero-sequence current, and the main circuit current; the combined-level harmonic risk probability is calculated based on the basic risk probability, the increase in the estimated harmonic level relative to the current total harmonic current distortion rate, and the increase in the equivalent compensation capacity of the candidate combination relative to the currently implemented equivalent compensation capacity.

[0017] Furthermore, the comprehensive control cost in S5 includes at least a combination of multiple of the following cost items: reactive power residual cost, switching action number cost, harmonic risk cost, overcompensation cost, undercompensation cost, zero-sequence current additional cost, and composite switch action interval end cost; each cost item is multiplied by its corresponding non-negative weight coefficient and then summed to obtain the comprehensive control cost, and the sum of all weight coefficients is 1.

[0018] Furthermore, the harmonic risk cost is calculated based on the combined harmonic risk probability and the degree to which the harmonic level estimate exceeds the allowable harmonic threshold; the zero-sequence current additional cost is positive only when the candidate combination contains a single-phase compensation branch and the current zero-sequence current exceeds the reference threshold, otherwise it is zero; the composite switch operation interval end cost is calculated by accumulating the costs based on whether the time between the most recent operation and the end of the operation of each branch in the candidate combination is less than the minimum operation interval.

[0019] Furthermore, the output of harmonic risk results and control suggestions in S6 includes: when the basic risk probability is less than the first threshold and the current total harmonic current distortion rate is less than the first distortion rate threshold, outputting a normal state; when the basic risk probability is between the first threshold and the second threshold, or when the current total harmonic current distortion rate is between the first distortion rate threshold and the second distortion rate threshold, outputting a level one risk warning; when the basic risk probability is greater than or equal to the second threshold or the current total harmonic current distortion rate is greater than or equal to the second distortion rate threshold, outputting a level two risk warning and recording the erroneous switching suppression control result; when the combined harmonic risk probability of a candidate combination is greater than or equal to the second threshold and the estimated harmonic level is greater than or equal to the second distortion rate threshold, outputting a level three risk warning and implementing a prohibition or delay control for the candidate combination.

[0020] The method for harmonic risk identification and erroneous switching suppression of an intelligent metal integrated distribution box according to the present invention has the following advantages:

[0021] This invention does not simply issue an alarm after harmonic exceedances, nor does it solely rely on whether the current THD exceeds a threshold to lock out all compensation branches. Instead, it combines the basic risk probability of high-risk harmonic events in the next 8 minutes, the estimated harmonic levels corresponding to candidate combinations, zero-sequence current, branch health status, and compensation capacity combinations into a single switching control decision. This achieves integrated control of harmonic risk identification and erroneous switching suppression. Compared to fixed-threshold compensation control methods, this invention reduces the number of harmonic exceedance windows and unnecessary switching actions during high-risk periods, while also maintaining compensation effectiveness. Attached Figure Description

[0022] Figure 1 A flowchart for harmonic risk identification and erroneous switching suppression in intelligent metal integrated distribution boxes.

[0023] Figure 2 This is a comparison chart of risk probability and THD under harmonic pollution conditions.

[0024] Figure 3 This is a comparison chart showing the effect of suppressing accidental switching under harmonic pollution conditions.

[0025] Figure 4 This is a comparison chart of the number of harmonic exceedance windows.

[0026] Figure 5 This is a comparison chart of the total number of throwing and cutting actions.

[0027] Figure 6 This is a comparison chart of power factors after suppressing erroneous switching under harmonic pollution conditions. Detailed Implementation

[0028] To better understand the purpose, structure, and function of this invention, the following detailed description of a method for identifying harmonic risks and suppressing accidental switching in an intelligent metal integrated distribution box is provided in conjunction with the accompanying drawings.

[0029] This invention discloses a method for harmonic risk identification and erroneous switching suppression in an intelligent metal integrated distribution box. The method is applied to an intelligent metal integrated distribution box, which at least includes a voltage acquisition unit, a current acquisition unit, a harmonic acquisition unit, a zero-sequence current acquisition unit, a power distribution monitoring and compensation control unit, a storage unit, and a control output unit. Figure 1 As shown, the method includes the following steps:

[0030] S1. Collect harmonic-related operating status information;

[0031] The operating status information includes the total harmonic current distortion rate. Fifth harmonic current component Seventh harmonic current component Zero-sequence current Active power reactive power Power factor Main circuit current Three-phase imbalance Current status of compensation branch road deployment The most recent action time for each branch road Rated compensation capacity of each branch road The actual response compensation capacity of each branch after its most recent successful commissioning and ambient temperature . Indicates the first One sampling point.

[0032] S2. Construct the harmonic risk characteristic vector;

[0033] A harmonic risk feature vector is constructed based on the harmonic and load sequences within the most recent 30-minute historical window, using a 1-minute sampling period; where the target reactive power compensation demand is... , The target power factor is preferably 0.98; the current effective capacity factor of each branch is calculated as follows: , The preferred value is 0.70; the branch health status coefficient is calculated as follows: Calculate the maximum branch health status quantity. Health status coefficient of each branch The maximum value in; the harmonic risk characteristics include , , , , , Current power factor Current main circuit current Current three-phase imbalance Ambient temperature and the maximum branch health status quantity .in, The current total harmonic current distortion rate characteristics, This represents the mean characteristic of the total harmonic current distortion rate over the most recent 5 sampling points. and These are the characteristics of the current fifth harmonic current component and the current seventh harmonic current component, respectively. The current zero-sequence current characteristics are as follows. The mean characteristic of the zero-sequence current at the most recent 5 sampling points is... Given the characteristics of reactive power compensation demand for the current target, This represents the mean characteristics of the target reactive power compensation demand over the most recent 5 sampling points. This indicates the number of sampling points traced back to the past.

[0034] S3. Estimate the probability of future short-term harmonic risks;

[0035] Input the harmonic risk feature vector obtained in step S2 into the lightweight gradient boosting classification model to obtain the basic risk probability of a high-risk harmonic event occurring within the next 8-minute window. High-risk harmonic events are defined as events that satisfy any of the following conditions at any time within the subsequent 8 sampling points: Condition 1, total harmonic current distortion rate. The total harmonic current distortion rate is greater than or equal to 6.6%, and meets one of the following conditions: "target reactive power compensation demand greater than or equal to 95kvar" or "zero-sequence current greater than or equal to 100A"; Condition 2: total harmonic current distortion rate. The target reactive power compensation requirement is greater than or equal to 110 kvar, and the maximum branch health status quantity is greater than or equal to 1.30. The lightweight gradient boosting classification model consists of 60 to 100 CART classification trees, each with a maximum depth of 2 to 4, a learning rate of 0.03 to 0.10, a minimum number of leaf node samples of 3 to 10, and a model storage footprint of no more than 256 kB.

[0036] S4. Generate the set of allowed branch inputs;

[0037] Based on the base risk probability output in step S3 Based on the current total harmonic current distortion rate, candidate compensation branch combinations are screened to generate a set of allowable branches; for any candidate combination First, calculate its equivalent compensation capacity. and the equivalent compensation capacity already in place Then calculate the estimated harmonic levels corresponding to the candidate combinations:

[0038]

[0039] in, The rated current of the main circuit.

[0040] Harmonic risk probability:

[0041] ;

[0042] Will satisfy and Candidate combinations will be included in the directly allowed set, satisfying and Candidate combinations that meet the minimum allowable threshold are included in the set of allowed combinations, while the remaining candidate combinations are listed as prohibited or deferred. Among them, Indicate candidate combinations At any moment Estimated total harmonic current distortion rate under the given conditions.

[0043] The set of allowed input branches is based on the harmonic risk probability of candidate combinations. Harmonic level estimates corresponding to candidate combinations Joint generation:

[0044] .

[0045] S5. Implement mis-dispensing suppression control;

[0046] Within the set of permissible input branches obtained in step S4, the comprehensive control cost is calculated for each candidate combination. And select the candidate combination with the lowest overall control cost as the current execution plan, where, The reactive power residual cost corresponding to the candidate combination. , Cost of switching action count; , The cost of harmonic risk; , To allow for harmonic threshold values, the preferred value is... ; To pay the price of unpaid compensation, , Select the reactive residual dead zone threshold. ; Adding a cost to zero-sequence current , As the sequence current reference threshold, take , In candidate combinations The value is 1 if it includes a single-phase compensation branch, otherwise it is 0. The cost of ending the operation interval of the composite switch. , This is an indicator function; it takes the value 1 if the condition within the parentheses is true, and 0 otherwise. To determine the minimum operating interval for closing the switch, take... ; This refers to the current control moment; when a branch is less than the time since its most recent action... When the time is right, the corresponding branch is counted as 1; otherwise, it is counted as 0. In order to compensate for the cost, ; To normalize the reference compensation capacity, the preferred option is... ; to The weight coefficients are non-negative and satisfy the following conditions: When the set of allowed input branches is empty, maintain the current input state or only allow selection to make... The exit strategy is declining.

[0047] S6. Output harmonic risk results and control recommendations;

[0048] Output harmonic risk level, permitted branch set, prohibited or delayed branch combination, harmonic exceedance warning, and maintenance re-inspection suggestions; among which, when and Output normal state when or Output a level 1 risk warning when or Output secondary risk warnings and record the results of erroneous injection suppression control when candidate combinations meet the requirements. and The system will issue a Level 3 risk warning and implement a ban or delay on investment for the candidate portfolio.

[0049] The following description is based on specific embodiments.

[0050] Step 1: Determine the target audience and sampling scope

[0051] A JP intelligent metal integrated distribution box with a total compensation capacity of 180kvar was selected as the representative implementation object. Branch C4 is a 20kvar single-phase compensation branch, while branches C5, C6, C7, and C8 are all 40kvar three-phase compensation branches. The sampling period was preferably set to 1 minute. The training, verification, and testing datasets mainly consisted of physical prototype test data, historical field operation data, and type test data, supplemented by boundary condition augmentation data generated by the digital prototype.

[0052] Step 2: Construction of Harmonic Risk Characteristics

[0053] Extract the most recent value of total harmonic current distortion from the most recent 30-minute window. Average total harmonic current distortion rate over the last 5 minutes Fifth harmonic current component Seventh harmonic current component Current zero-sequence current Average zero-sequence current over the last 5 minutes Current target reactive power compensation requirements Average reactive power compensation demand of the target over the last 5 minutes Current power factor, main circuit current, three-phase imbalance, ambient temperature, and maximum branch health status. Together, they constitute the harmonic risk characteristic vector; among which , , , Health status coefficient of each branch The maximum value in.

[0054] Step 3: Harmonic Risk Probability Estimation

[0055] In this embodiment, a lightweight gradient boosting classification model is used to identify high-risk harmonic events within an 8-minute window. The model preferably consists of 80 CART classification trees with a maximum depth of 3, a learning rate of 0.05, a minimum leaf node sample size of 5, and a total model storage footprint not exceeding 256kB. The model training labels are obtained by labeling the actual measurement values ​​of the subsequent 8 sampling points in the physical prototype's measured sequence and the historical field sequence according to preset high-risk event rules.

[0056] If the THDI is greater than or equal to 6.6% at any of the subsequent 8 sampling points, and either the target reactive power compensation requirement is greater than or equal to 95kvar or the zero-sequence current is greater than or equal to 100A, then the current sample will be marked as a high-risk positive sample for harmonics.

[0057] If at any of the subsequent 8 sampling points the THDI is greater than or equal to 5.6%, the maximum branch health status quantity is greater than or equal to 1.30, and the target reactive power compensation demand is greater than or equal to 110kvar, the current sample will also be marked as a high-risk positive sample for harmonics. The above thresholds are determined jointly based on GB / T 14549, capacitor type test data, on-site historical over-limit statistics, and validation set grid search, rather than being set by the simulation model itself.

[0058] Model outputs base risk probability .when When the value is greater than or equal to 0.42, it is considered that the current moment already has a clear trend of high harmonic risk; when Greater than or equal to 0.72 or current When the risk level is greater than or equal to 6.3%, the system enters a high-risk suppression and control mode.

[0059] Step 4: Generate the set of allowed input branches and perform cost evaluation.

[0060] In this embodiment, all possible input states of the five compensation branches C4, C5, C6, C7, and C8 are uniformly grouped into a candidate combination set, forming a total of 32 candidate combinations. For any candidate combination S, the equivalent compensation capacity at time k is... The equivalent compensation capacity has already been put into use. ,in This represents the current effective capacity coefficient of the branch.

[0061] For each candidate combination, the estimated harmonic level is further calculated:

[0062]

[0063] Harmonic risk probability:

[0064] ;

[0065] When the candidate combination satisfies and When, include it in the directly allowed set; when the candidate combination satisfies And the equivalent compensation capacity is no greater than When, include it in the tolerance-allowed set; when and And the equivalent compensation capacity of the candidate combination is greater than At that time, a high-risk investment ban is imposed on the candidate combination, making it no longer a preferred option.

[0066] Within the allowed set of branch lines, according to Calculate the overall control cost, where , , , , , , coefficient to First, the cost terms are normalized, and then determined by joint optimization of grid search and coordinate descent on the validation set consisting of physical prototype test data, field historical data and type test data.

[0067] When the harmonic risk is low, the controller is allowed to select the capacity combination with the lower overall cost from the candidate combinations based on the compensation target. When the harmonic risk is high, a higher cost is imposed on high-capacity combinations and single-phase combinations sensitive to zero-sequence current, and lower-risk branch combinations are prioritized to limit unnecessary large-capacity deployment. For Level 3 risk status, the controller directly outputs the control result of prohibiting, delaying, or maintaining the current state, and simultaneously generates the harmonic risk level, branch control record, and operation and maintenance review suggestions.

[0068] Step Six: Joint Data Validation and Result Analysis

[0069] In the development phase of this embodiment, training, verification, and testing datasets were constructed using physical prototype test data, historical field operation data, and type test data. A system-level digital prototype built using Python was used to supplement boundary samples under normal operating conditions, low power factor fluctuation conditions, harmonic pollution conditions, and branch aging disturbance conditions. The digital prototype data is only used to expand upon operating conditions that are not easily covered by physical data over a long period and is not used as the sole basis for the final results.

[0070] Joint test results show that, within the test set comprised of physical prototype test data, field operation history data, and digital prototype boundary augmentation data, the harmonic risk identification model of the present invention achieves an accuracy of approximately 0.9744, a recall of approximately 0.9500, and an F1 score of approximately 0.9620. Under harmonic pollution conditions, the present invention reduces the number of harmonic exceedance windows from 240 to 172, a reduction of approximately 28.33%; simultaneously, it reduces the total number of switching operations from 860 to 478, a reduction of approximately 44.42%. Regarding the power factor, the present invention can still maintain the compensated power factor at approximately 0.9954 under harmonic pollution conditions, while the fixed threshold method achieves approximately 0.9605.

[0071] like Figure 2 As shown, the basic risk probability and candidate combination risk probability output by the method of this invention can be significantly increased during the harmonic pollution stage, and maintain a corresponding relationship with THD changes; as Figure 3 , Figure 4 Figure 5As shown, the method of the present invention can effectively suppress excessive switching in harmonic pollution scenarios and reduce the number of harmonic over-limit windows; as Figure 6 As shown, while controlling mis-switching, a high compensation effect can still be maintained.

[0072] Therefore, the method of the present invention can directly link the candidate combination-level harmonic risk estimation results to the compensation control strategy based on harmonic risk identification.

[0073] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. 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 harmonic risk identification and erroneous switching suppression in an intelligent metal integrated distribution box, characterized in that, Includes the following steps: S1. Collect harmonic-related operating status information; S2. Construct a harmonic risk feature vector based on the collected operational status information; S3. Input the harmonic risk feature vector into the lightweight gradient boosting classification model to estimate the basic risk probability of a high-risk harmonic event occurring in the near future. S4. Based on the basic risk probability and the current total harmonic current distortion rate, the candidate compensation branch combinations are screened to generate a set of branches that can be put into operation. S5. Within the set of allowed input branches, calculate the comprehensive control cost of each candidate compensation branch combination, and select the candidate combination with the smallest comprehensive control cost as the current execution plan to perform erroneous input suppression control. S6. Output harmonic risk results and control recommendations.

2. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The operating status information collected in S1 includes: total harmonic current distortion rate, fifth harmonic current component, seventh harmonic current component, zero-sequence current, active power, reactive power, power factor, main circuit current, three-phase imbalance, current compensation branch activation status, the most recent operation time of each branch, the rated compensation capacity of each branch, the actual response compensation capacity of each branch after the most recent successful activation, and ambient temperature.

3. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The construction of the harmonic risk feature vector in S2 further includes: based on the harmonic and load sequences within the most recent 30-minute historical window, constructing the following features according to a 1-minute sampling period: current total harmonic current distortion rate feature, mean feature of total harmonic current distortion rate of the most recent 5 sampling points, current fifth harmonic current component feature, current seventh harmonic current component feature, current zero-sequence current feature, mean feature of zero-sequence current of the most recent 5 sampling points, current target reactive power compensation demand feature, mean feature of target reactive power compensation demand of the most recent 5 sampling points, current power factor, current main circuit current, current three-phase imbalance, ambient temperature, and maximum branch health status quantity.

4. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The basic risk probability of a high-risk harmonic event occurring in the future within a short period of time in S3 refers to the probability of a high-risk harmonic event occurring within the next 8-minute window. The high-risk harmonic event is defined as meeting any of the following conditions at any time within the subsequent 8 sampling points: Condition 1: The total harmonic current distortion rate is greater than or equal to 6.6%, and meets either "target reactive power compensation demand is greater than or equal to 95 kvar" or "zero-sequence current is greater than or equal to 100 A"; Condition 2: The total harmonic current distortion rate is greater than or equal to 5.6%, and the maximum branch health status quantity is greater than or equal to 1.30, while the target reactive power compensation demand is greater than or equal to 110 kvar.

5. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The lightweight gradient boosting classification model in S3 consists of 60 to 100 CART classification trees, each with a maximum depth of 2 to 4, a learning rate of 0.03 to 0.10, a minimum number of leaf node samples of 3 to 10, and a model storage footprint of no more than 256 kilobytes.

6. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The generation of the permitted branch set in S4 specifically includes: for any candidate combination, first calculate its equivalent compensation capacity and the currently invested equivalent compensation capacity, then calculate the estimated harmonic level and the combination-level harmonic risk probability corresponding to the candidate combination; include candidate combinations that meet the condition that the combination-level harmonic risk probability is less than the first threshold and the estimated harmonic level is less than the permitted harmonic threshold into the directly permitted set; include candidate combinations that meet the condition that the combination-level harmonic risk probability is between the first threshold and the second threshold and the equivalent compensation capacity is not greater than the capacity limit threshold into the capacity-limited permitted set; and list the remaining candidate combinations as prohibited or delayed investment targets.

7. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 6, characterized in that, The estimated harmonic level corresponding to the candidate combination is calculated based on the current total harmonic current distortion rate, the increase in the equivalent compensation capacity of the candidate combination relative to the currently implemented equivalent compensation capacity, the maximum branch health status, the zero-sequence current, and the main circuit current; the combined harmonic risk probability is calculated based on the basic risk probability, the increase in the estimated harmonic level relative to the current total harmonic current distortion rate, and the increase in the equivalent compensation capacity of the candidate combination relative to the currently implemented equivalent compensation capacity.

8. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The comprehensive control cost in S5 includes at least a combination of several of the following cost items: reactive power residual cost, switching action number cost, harmonic risk cost, overcompensation cost, undercompensation cost, zero-sequence current additional cost, and composite switch action interval end cost; each cost item is multiplied by its corresponding non-negative weight coefficient and then summed to obtain the comprehensive control cost, and the sum of all weight coefficients is 1.

9. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 8, characterized in that, The harmonic risk cost is calculated based on the combined harmonic risk probability and the degree to which the harmonic level estimate exceeds the allowable harmonic threshold; the zero-sequence current additional cost is positive only when the candidate combination contains a single-phase compensation branch and the current zero-sequence current exceeds the reference threshold, otherwise it is zero; the composite switch operation interval end cost is calculated by accumulating the costs based on whether the time between the most recent operation and the end of the operation of each branch in the candidate combination is less than the minimum operation interval.

10. The method for harmonic risk identification and erroneous switching suppression of intelligent metal integrated distribution boxes according to claim 1, characterized in that, The harmonic risk results and control suggestions output in S6 include: when the basic risk probability is less than the first threshold and the current total harmonic current distortion rate is less than the first distortion rate threshold, a normal state is output; when the basic risk probability is between the first threshold and the second threshold, or the current total harmonic current distortion rate is between the first distortion rate threshold and the second distortion rate threshold, a level one risk warning is output; when the basic risk probability is greater than or equal to the second threshold or the current total harmonic current distortion rate is greater than or equal to the second distortion rate threshold, a level two risk warning is output and the erroneous switching suppression control result is recorded; when the combined harmonic risk probability of a candidate combination is greater than or equal to the second threshold and the estimated harmonic level is greater than or equal to the second distortion rate threshold, a level three risk warning is output and a ban or delay control is implemented for the candidate combination.

Images