Low-voltage capacitor control method based on compensation system
By analyzing the correlation and prediction reliability between power grid parameters and reactive power, calculating weighting coefficients, and combining historical reference values to correct the reactive power prediction results, the problem of low reactive power prediction accuracy is solved, and accurate capacitor compensation and power grid stability are achieved.
Patent Information
- Application Number
- CN202511524903.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-24
- Publication Date
- 2025-11-21
AI Technical Summary
Existing technologies fail to effectively consider the dynamic coupling relationship between grid parameters and reactive power in reactive power prediction, resulting in low prediction accuracy and affecting the compensation effect of capacitors and grid stability.
By analyzing the correlation between power grid parameters and reactive power and the reliability of the prediction, weighting coefficients are calculated, and the prediction results of reactive power are corrected by combining historical reference values. Low-voltage capacitors are used for real-time compensation.
It improves the accuracy of reactive power prediction, enhances the compensation effect of capacitors, and improves the operational stability of the power grid.
Smart Images
Figure CN120999664A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application relates to the technical field of reactive power compensation, in particular to a low-voltage capacitor control method based on a compensation system. BACKGROUND
[0002] In a power system, accurate prediction and dynamic compensation of reactive power are crucial for maintaining grid voltage stability, reducing line loss and improving power quality. Fluctuations in reactive power can lead to voltage instability, reduced power factor, and even equipment failure and grid collapse. In order to achieve efficient reactive power compensation and optimal control, accurate prediction of reactive power is required.
[0003] Due to the dynamic changes and uncertainties of power system load, the prediction of reactive power becomes more complex. Traditional methods mainly rely on the trend characteristics of reactive power for prediction modeling, without considering the dynamic coupling relationship between the remaining grid parameters and reactive power, and ignoring the reference value of historical periods with similar change trends, which makes it difficult to accurately capture the nonlinear, random and time-varying change trend characteristics of reactive power, resulting in large deviations in the prediction of reactive power, low prediction accuracy of reactive power, and affecting the compensation effect of capacitors on reactive power and the stability of grid operation. SUMMARY
[0004] In order to solve the above technical problems, a low-voltage capacitor control method based on a compensation system is provided to solve the existing problems.
[0005] The technical problem of the application is solved by providing a low-voltage capacitor control method based on a compensation system, comprising the following steps: Obtain each grid parameter at each time in the current monitoring period and each historical period, including reactive power; Predict each grid parameter in the preset local period at each time, analyze the difference between reactive power at different times and its corresponding prediction result, and calculate the prediction credibility of the current time; Analyze the correlation between reactive power and the remaining grid parameters in the local period, as well as the correlation between the prediction result of reactive power and the prediction result of the remaining grid parameters, determine the prediction correlation degree of the current time, and determine the weight coefficient of the current time in combination with the prediction credibility; Based on the difference between the grid parameters at different times in the current monitoring period, all times in the current monitoring period are divided into categories; analyze the continuity of time in each category to obtain a target period; The evaluation coefficient is calculated according to the difference between the target period and the same period in each historical period, so as to obtain the reference period corresponding to each historical period; and the historical reference value corresponding to the next moment of the current moment is obtained based on the reactive power of the reference period in the next moment; The reactive power is predicted, the prediction result of the reactive power is corrected in combination with the historical reference value and the weight coefficient, and the reactive power is compensated in real time through the capacitor.
[0006] Preferably, the calculation of the prediction reliability of the current moment comprises: determining the prediction error between the reactive power and the predicted value of the reactive power at each moment; calculating the average value of the prediction error of the current moment and a plurality of moments before the current moment, and performing negative mapping on the average value as the prediction reliability of the current moment.
[0007] Preferably, the determination of the prediction correlation degree of the current moment comprises: the correlation degree between each grid parameter and the reactive power at each moment and at all moments in the local period is recorded as the first correlation degree; analyzing the correlation change between each grid parameter and the reactive power at each moment and at all moments in the local period, and calculating the second correlation degree; the difference between the first correlation degree and the second correlation degree is taken as the relative difference of each grid parameter at each moment; the cumulative sum of the relative difference of all grid parameters at each moment is calculated, and the average value of the cumulative sum of the current moment and a plurality of moments before the current moment is negatively mapped as the prediction correlation degree of the current moment.
[0008] Preferably, the calculation of the second correlation degree comprises: the predicted parameter sequence of each grid parameter at each moment and at all moments in the local period is formed; all the reactive power at all moments in the local period of each moment and the predicted value of the reactive power at each moment are formed into a predicted reactive sequence; the second correlation degree is the correlation degree between the predicted parameter sequence of each grid parameter at each moment and the predicted reactive sequence.
[0009] Preferably, the weight coefficient is the normalized result of the average value of the prediction reliability and the prediction correlation degree.
[0010] Preferably, the acquisition process of the target period comprises: the data of all grid parameters at each moment is formed into a feature vector; The feature vectors of all time points in a current monitoring period are clustered, a cluster to which the current time point belongs is selected, and the cluster is recorded as a target cluster; in the target cluster, the current time point and a plurality of time points continuous to the current time point are taken as a target time period.
[0011] Preferably, the calculation of the evaluation coefficient comprises: A same time period corresponding to the target time period in each historical period is intercepted as a matching time period. A normalization result of an inverse of a distance between the feature vectors of all time points between the matching time period corresponding to each historical period and the target time period is taken as an evaluation coefficient of the matching time period corresponding to each historical period.
[0012] Preferably, the acquisition of the reference time period corresponding to each historical period comprises: taking the matching time period corresponding to all historical periods whose evaluation coefficients are greater than a preset threshold as a reference time period.
[0013] Preferably, the historical reference value is a mean value of a reactive power of a next time point of a last time point in the reference time period corresponding to all historical periods.
[0014] Preferably, the correction of the prediction result of the reactive power comprises: The reactive power at the time point corresponds to a corrected prediction value The calculation formula of the corrected prediction value is: wherein, is a weight coefficient of the current time point is a prediction value of the reactive power at the time point is a historical reference value of the time point
[0015] The application has at least the following beneficial effects: The application predicts the reactive power in each local period, analyzes the difference between the predicted value of the reactive power at each time and the actually monitored reactive power, calculates the prediction reliability of the current time, and has the beneficial effect of considering the prediction deviation of the reactive power, being able to capture the reliability of the prediction result of the short-term dynamic change of the reactive power by the prediction algorithm, and reflecting that the prediction algorithm performs well under the current condition; determining the prediction correlation degree of the current time, which has the beneficial effect of considering the dynamic coupling relationship between the remaining power grid parameters and the reactive power, and explaining the maintenance of the correlation between different power grid parameters and the reactive power at each time, so as to reflect the reliability of the prediction result of the reactive power; obtaining the weight coefficient of the current time, which has the beneficial effect of comprehensively evaluating the prediction accuracy of the reactive power at the current time, reflecting the capture of the change trend of the reactive power by the prediction algorithm at the current time, and explaining that the prediction result is reliable; obtaining the target period, which has the beneficial effect of selecting the time segment where the current time is located, so as to identify the data of the historical period with reference value in the subsequent; calculating the evaluation coefficient to obtain the reference period corresponding to each historical period, which has the beneficial effect of selecting the reference period in the historical period similar to the change trend of the target period through the difference of various parameters in the target period and the corresponding period in the historical period, and then obtaining the historical reference value corresponding to the next time of the current time through the reactive power of the next time of the reference period, which has the beneficial effect of considering the average level of the reactive power of the next time of the reference period in the historical period similar to the trend, and providing an important reference for the prediction of the current time; based on the prediction result of the reactive power, combining the historical reference value and the weight coefficient, correcting the prediction result of the reactive power, and compensating the reactive power in real time through the capacitor, which has the beneficial effect of dynamically analyzing the reliability of the prediction algorithm for the prediction result of the reactive power at different times, correcting the prediction value of the reactive power, increasing the dependence on the prediction value of the reactive power in the case of reliable prediction, increasing the dependence on the historical reference value in the case of unreliable prediction, accurately capturing the change trend characteristics of the nonlinearity, randomness and time variation of the reactive power, improving the prediction accuracy of the reactive power, making the compensation control of the capacitor more accurate, and improving the compensation effect of the capacitor on the reactive power and the stability of the power grid operation. BRIEF DESCRIPTION OF DRAWINGS
[0016] The application will be further described in detail below in combination with the drawings.
[0017] Figure 1 A step flow chart of a low-voltage capacitor control method based on a compensation system provided by the embodiment of the application; Figure 2A step flowchart of a method for acquiring a prediction correlation degree of a current time moment is provided in the embodiments of the present application. DETAILED DESCRIPTION
[0018] In order to make the purposes, technical solutions and advantages of the present application clearer, the present application is further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and not to limit the present application.
[0019] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the present application belongs.
[0020] Referring to Figure 1 , a step flowchart of a method for controlling a low-voltage capacitor based on a compensation system is shown, which comprises the following steps: Step 1, acquiring each power grid parameter at each time moment in a current monitoring period and each historical period, wherein the reactive power is included.
[0021] In a power system, transformers, asynchronous motors and other devices have electromagnetic coils. These devices need to establish a magnetic field when working, which consumes a large amount of reactive power. If the reactive power loss cannot be compensated in time, it will lead to voltage fluctuation in the power grid, power factor reduction, and further affect the safe and stable operation of the power grid. Therefore, capacitors are usually used as reactive power compensation devices to provide capacitive reactive power to offset inductive reactive power, thereby real-time compensating the reactive power in the power grid, improving the power factor of the power system, reducing line loss, and improving voltage quality.
[0022] Based on the above analysis, the intelligent electric meter is used to acquire each power grid parameter at each time moment in different monitoring periods in real time, wherein the power grid parameters include current, voltage, active power and reactive power. In the present embodiment, the length of the monitoring period is one day, and the time interval of data acquisition is 1s. As other implementation manners, the implementer can set it according to the actual situation.
[0023] The monitoring period in which the current time moment is located is recorded as the current monitoring period, and all monitoring periods before the current monitoring period are recorded as each historical period. Thus, each power grid parameter at each time moment in the current monitoring period and each historical period is obtained.
[0024] Step 2, predicting each power grid parameter in a preset local period at each time moment, analyzing the difference between the reactive power at different time moments and the corresponding prediction results, and calculating the prediction reliability of the current time moment.
[0025] Since the reactive power in different historical periods has certain variation trend, the difference between the prediction result of the reactive power in each historical period and the actual data is analyzed to calculate the prediction credibility and evaluate the accuracy of the reactive power prediction, specifically: The multiple time points before each time point are recorded as a local period. In this embodiment, 1000 time points before each time point are recorded as a local period. As an alternative, the implementer can set it according to the actual situation.
[0026] For the current monitoring period, the prediction algorithm is used to predict the reactive power at all time points in the local period to obtain the predicted value of the reactive power at each time point. In this embodiment, the exponential smoothing algorithm is used for prediction, wherein the exponential smoothing algorithm is a known technology and will not be described here. Since the local period does not include each time point, when predicting the reactive power at all time points in the local period, the predicted value of the reactive power at each time point is obtained.
[0027] The difference between the reactive power at each time point in the current monitoring period and the predicted value of the reactive power is recorded as a prediction error. In this embodiment, the absolute value of the difference between the reactive power at each time point in the current monitoring period and the predicted value of the reactive power is recorded as a prediction error.
[0028] The average value of the prediction error of the current time point and the multiple time points before it is calculated, and the average value is negatively mapped as the prediction credibility of the current time point. In this embodiment, the average value of the prediction error of the current time point and the 60 time points before it is calculated. As an alternative, the implementer can set it according to the actual situation. Secondly, the specific process of negative mapping is: negative mapping is performed through an exponential function, assuming that the average value is recorded as The result of is taken as the prediction credibility, wherein is an exponential function with a natural constant as the base.
[0029] It should be noted that the smaller the prediction error, the smaller the difference between the prediction result of the reactive power and the actual reactive power, and the greater the obtained prediction credibility, indicating that the accuracy of the prediction result after predicting the reactive power is relatively high.
[0030] Thus, the prediction credibility of the current time point is obtained.
[0031] Step 3: Analyze the correlation between reactive power and other grid parameters within a local time period, as well as the correlation between the predicted reactive power and the predicted results of other grid parameters, determine the prediction correlation at the current moment, and determine the weighting coefficient at the current moment based on the prediction reliability.
[0032] Secondly, the changes in reactive power in the power grid are coupled with the changes in other power grid parameters to a certain extent. Therefore, it is necessary to analyze the correlation between reactive power and other power grid parameters and calculate the predictive correlation degree. The flowchart of the method for obtaining the predictive correlation degree at the current moment provided in this application embodiment is as follows: Figure 2 As shown, specifically: The degree of correlation between the data of each power grid parameter at each time point and all times within the local time period and the reactive power is denoted as the first correlation degree; In this embodiment, the correlation is measured by calculating the Pearson correlation coefficient between the data of each power grid parameter at each time and all times within the local time period and the reactive power. The Pearson correlation coefficient is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other methods of the prior art, such as Spearman correlation coefficient, cosine similarity, etc. This embodiment does not impose any special restrictions on this.
[0033] The data of each power grid parameter at all times within the local time period are predicted to obtain the predicted value of each power grid parameter at each time. In this embodiment, the exponential smoothing algorithm is used for prediction. The exponential smoothing algorithm is a well-known technology and will not be described in detail here.
[0034] The predicted values of each power grid parameter at each time point, along with the data from all times within the local time period, are combined to form a predicted parameter sequence. The predicted reactive power at each moment is combined with the reactive power at all moments within the local time period to form a predicted reactive power sequence. The correlation between the predicted parameter sequence and the predicted reactive power sequence of each power grid parameter at each time point is calculated and denoted as the second correlation degree. In this embodiment, the degree of correlation is measured by calculating the Pearson correlation coefficient between the predicted parameter sequence of each power grid parameter and the predicted reactive power sequence. The Pearson correlation coefficient is a well-known technique and will not be described in detail here. As other implementation methods, implementers may use other methods of the prior art, such as Spearman correlation coefficient, cosine similarity, etc. This embodiment does not impose any special restrictions on this.
[0035] The difference between the first correlation and the second correlation is taken as the relative difference of each power grid parameter at each time point; In this embodiment, the absolute value of the difference between the first correlation and the second correlation is denoted as the relative difference.
[0036] calculating the accumulation sum of the relative difference of all power grid parameters at each time, and performing negative mapping on the average of the accumulation sum of the current time and a plurality of times before the current time as the prediction correlation degree of the current time; In the embodiment, the average of the accumulation sum of the current time and 60 times before the current time is negatively mapped as the other implementation, and the implementer can set it according to the actual situation. Secondly, the specific process of negative mapping is: the negative mapping is performed through an exponential function, and it is assumed that the accumulation sum is denoted as , the result of which is taken as the prediction correlation degree, wherein is an exponential function with a natural constant as the base.
[0037] It should be noted that the greater the first correlation degree, the higher the correlation between the power grid parameter and the reactive power at each time and the local period thereof. The greater the second correlation degree, the better the correlation between the power grid parameter and the reactive power after the prediction of the power grid parameter and the reactive power respectively, and the prediction result has a high credibility. The smaller the relative difference, the greater the prediction correlation degree, which means that the correlation between the power grid parameter and the reactive power before and after the prediction is relatively consistent, the prediction algorithm can capture the correlation between the power grid parameter and the reactive power well at the current time, the prediction result can maintain the correlation between the power grid parameter and the reactive power, and the prediction result of the reactive power is highly reliable.
[0038] Further, based on the prediction credibility and the prediction correlation degree, a weight coefficient is determined, specifically: The normalized result of the average of the prediction credibility and the prediction correlation degree is taken as the weight coefficient of the current time. In the embodiment, a sigmoid function is used for normalization processing, wherein the sigmoid function is a known technology, and will not be described here. As other implementation, the implementer can use other methods of prior art, for example, a softmax function, and the embodiment does not specially limit this.
[0039] It should be noted that the greater the weight coefficient, the higher the prediction accuracy of the reactive power at the current time, and the prediction algorithm can capture the change trend of the reactive power well at the current time, and the prediction result has a high credibility.
[0040] Thus, the weight coefficient of the current time is obtained.
[0041] Step 4, based on the difference of the power grid parameters at different time points in the current monitoring period, all time points in the current monitoring period are divided into categories; the continuity of the time in each category is analyzed to obtain a target period; the evaluation coefficient is calculated through the difference between the target period and the same period corresponding to each historical period to obtain the reference period corresponding to each historical period; and the historical reference value corresponding to the next time point of the current time point is obtained based on the reactive power of the reference period at the next time point.
[0042] Further, in the power system, the data changes of different power grid parameters have periodicity, for example, the power load increases at the peak of night electricity consumption; therefore, by analyzing the similar situation of data changes in the historical period and the current time point, the reference period is selected, and the historical reference value is calculated, specifically: The data of all power grid parameters at each time point is combined to form a feature vector. In this embodiment, the current, voltage, active power and reactive power at each time point are combined to form a feature vector.
[0043] The feature vectors of all time points in the current monitoring period are clustered, and the cluster to which the current time point belongs is selected, which is denoted as a target cluster. In this embodiment, the DBSCAN clustering algorithm is used to obtain multiple clustering clusters, wherein the DBSCAN clustering algorithm is a known technology and will not be described here. As other embodiments, the implementer can use other methods of prior art, for example, K-means clustering algorithm, and this embodiment does not specially limit this.
[0044] In the target cluster, the current time point and multiple time points continuous to the current time point are taken as a target period. The same period corresponding to the target period in each historical period is taken as a matching period. The normalization result of the inverse of the distance of the feature vectors of all time points between the matching period corresponding to each historical period and the target period is calculated, which is denoted as an evaluation coefficient. In this embodiment, the distance is measured by calculating the DTW distance of the feature vectors of all time points between the matching period corresponding to each historical period and the target period, wherein the DTW distance is a known technology and will not be described here. Secondly, the sigmoid function is used for normalization processing, wherein the sigmoid function is a known technology and will not be described here. As other embodiments, the implementer can use other methods of prior art, for example, softmax function, and this embodiment does not specially limit this.
[0045] The matching period corresponding to all historical periods with the evaluation coefficient greater than a preset threshold is denoted as a reference period. The average of the reactive power of the next moment of the last moment in the reference period corresponding to all historical periods is taken as the historical reference value of the next moment corresponding to the current moment; In the embodiment, the preset threshold value is 0.7, and as another implementation manner, the implementer can set it according to the actual situation.
[0046] It should be noted that the greater the evaluation coefficient, the closer the change trend of the power grid data of the matching period corresponding to each historical period to the target period, and the higher the prediction reference of the reactive power of the historical period to the reactive power of the next moment of the current moment.
[0047] Step 5, predicting the reactive power, combining the historical reference value and the weight coefficient, correcting the prediction result of the reactive power, and performing real-time compensation on the reactive power through the capacitor.
[0048] Further, based on the weight coefficient and the historical reference value, the prediction value of the reactive power of the next moment of the current moment is corrected, specifically: The reactive power of the next moment of the current moment is predicted, and the prediction value of the reactive power of the next moment of the current moment is obtained; In the embodiment, the exponential smoothing algorithm is used for prediction, wherein the exponential smoothing algorithm is a known technology and will not be described here.
[0049] wherein, is the corrected prediction value of the reactive power at the moment, is the weight coefficient of the current moment, is the prediction value of the reactive power at the moment, is the historical reference value at the moment. It should be noted that the greater the weight coefficient, the relatively higher the prediction accuracy of the reactive power at the current moment, and therefore, the dependence on the prediction value of the reactive power is increased at this time, and vice versa, that is, the prediction accuracy of the reactive power at the current moment is relatively low, and the change of the reactive power of the historical period should be relied on more.
[0050] Based on the corrected prediction value of the reactive power at the moment, the reactive power of the power grid is compensated through the low-voltage capacitor, and the specific process is as follows: The difference between the corrected prediction value of the reactive power at the moment and the reactive power of the current moment is taken as the compensation amount. According to the compensation amount, the switching operation of the capacitor bank is controlled, if the compensation amount is greater than 0, the corresponding capacitor bank needs to be put into operation to increase the reactive power compensation; if the compensation amount is less than 0, the corresponding capacitor bank needs to be cut off to reduce the reactive power; if the compensation amount is equal to 0, the switching operation of the capacitor bank does not need to be controlled.
[0051] It should be understood that, although Figure 1 The steps in the flowcharts of the above embodiments are displayed in sequence according to the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other orders. Moreover, Figure 1 At least part of the steps in the above embodiments can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution order of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with at least part of other steps or sub-steps or stages of other steps.
[0052] The technical features of the above embodiments can be combined in any manner. In order to make the description concise, all possible combinations of the technical features in the above embodiments are not described, but as long as the combination of the technical features does not exist contradictory, it should be considered as the scope of the present application.
[0053] The above embodiments only express several implementation manners of the present application, and the description is more specific and detailed, but it should not be understood as a limitation of the present application. It should be pointed out that, for ordinary skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, therefore, any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application, without departing from the technical scheme of the present application, all belong to the protection scope of the technical scheme of the present application.
Claims
1. A low-voltage capacitor control method based on a compensation system, characterized in that, The method includes the following steps: Acquire data on power grid parameters at each time point within the current monitoring cycle and each historical cycle, including reactive power; Predict the parameters of each power grid within a preset local time period at each time, analyze the difference between the reactive power at different times and the corresponding prediction results, and calculate the prediction reliability at the current time. Analyze the correlation between reactive power and other grid parameters within a local time period, as well as the correlation between the predicted reactive power and the predicted reactive power and other grid parameters, determine the prediction correlation at the current moment, and combine the prediction reliability to determine the weighting coefficient at the current moment. Based on the differences in power grid parameters at different times within the current monitoring period, all times within the current monitoring period are divided into categories; the continuity of time within each category is analyzed to obtain the target time period; By analyzing the differences in grid parameters between the target time period and the corresponding time period in each historical cycle, an evaluation coefficient is calculated to obtain the reference time period for each historical cycle. Based on the reactive power of the reference time period at the next moment, the historical reference value for the next moment corresponding to the current moment is obtained. Reactive power is predicted, and the prediction results are corrected by combining historical reference values and weighting coefficients. Reactive power is then compensated in real time through capacitors.
2. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, The calculation of the prediction confidence at the current moment includes: Determine the prediction error between the reactive power at each time point and its predicted value; Calculate the average of the prediction errors at the current time and at multiple previous times, and apply a negative mapping to the average as the prediction confidence at the current time.
3. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, Determining the predictive correlation at the current moment includes: The degree of correlation between the data of each power grid parameter at each time point and all times within its local time period and reactive power is denoted as the first correlation degree; Analyze the correlation changes between various power grid parameters and the predicted values of reactive power at each time point, as well as the data at all times within a local time period, and calculate the second correlation degree; The difference between the first correlation and the second correlation is taken as the relative difference of each power grid parameter at each time point; Calculate the cumulative sum of the relative differences of all power grid parameters at each time point, and negatively map the mean of the cumulative sums at the current time point and multiple previous times as the prediction correlation degree at the current time point.
4. The low-voltage capacitor control method based on a compensation system as described in claim 3, characterized in that, The calculation of the second relevance includes: The predicted values of each power grid parameter at each time point, along with the data from all times within the local time period, are combined to form a predicted parameter sequence. The reactive power at all times within a local time period at each time point, and the predicted reactive power at each time point, are combined to form a predicted reactive power sequence. The second correlation is the degree of correlation between the predicted parameter sequence and the predicted reactive power sequence of each power grid parameter at each time point.
5. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, The weighting coefficients are the normalized results of the mean values of the prediction confidence and the prediction correlation.
6. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, The process of obtaining the target time period is as follows: The data of all power grid parameters at each time point are used to form a feature vector; Cluster the feature vectors of all times within the current monitoring period, select the cluster to which the current time belongs, and denote it as the target cluster; Within the target cluster, the current time and multiple consecutive times are defined as the target time period.
7. The low-voltage capacitor control method based on a compensation system as described in claim 6, characterized in that, The calculation of the evaluation coefficients includes: Extract the time period that corresponds to the target time period within each historical period as the matching time period; The normalized result of the reciprocal of the distance between the matching time period corresponding to each historical period and the target time period at all times is used as the evaluation coefficient of the matching time period corresponding to each historical period.
8. The low-voltage capacitor control method based on a compensation system as described in claim 7, characterized in that, The step of obtaining the reference time period corresponding to each historical period includes: recording the matching time periods corresponding to all historical periods with an evaluation coefficient greater than a preset threshold as reference time periods.
9. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, The historical reference value is the average reactive power at the moment following the last moment within the reference period corresponding to all historical cycles.
10. The low-voltage capacitor control method based on a compensation system as described in claim 1, characterized in that, The correction of the reactive power prediction results includes: reactive power in The corrected predicted value at each time point The calculation formula is: ,in, For the current moment The weighting coefficients, For reactive power in Predicted value at time, for Historical reference value for a given moment.