A method and system for switching control of a reactive compensation capacitor

Abstract

The application relates to the technical field of intelligent scheduling, in particular to a reactive power compensation capacitor switching control method and system, which comprises the following steps: screening a representative weather station and constructing a waveform and peak value characteristic space; clustering and dividing scenes in the characteristic space, and screening typical periods according to the density in the cluster; constructing an extreme screening coefficient of an extreme working condition point based on the distance from the extreme working condition point to the surrounding box and the extreme adjacent point, so as to obtain an extreme scene period under any characteristic space, and form a representative period together with the typical period; and controlling the capacitor according to the switching strategy of the representative period. The application solves the problem that the existing control ignores photovoltaic timing fluctuation and extreme working conditions, leading to high capacitor switching frequency and compensation mismatch, and cooperatively improves voltage regulation accuracy, equipment life and operation safety.

Description

Technical Field

[0001] This application relates to the field of intelligent scheduling technology, specifically to a method and system for switching reactive power compensation capacitors. Background Technology

[0002] With the optimization and adjustment of the traditional energy structure, large-scale wind power, photovoltaic, and other new energy bases are being built across various regions. The voltage stability of new energy transmission systems highly depends on the precise control of reactive power. However, these systems lack synchronous power supply support at their near-end, and the dynamic regulation capacity of the static var generators (SVG) configured in photovoltaic power plants is limited. Furthermore, the reactive power losses in multi-stage AC transmission networks exhibit strong time-series variations with fluctuations in photovoltaic output, making voltage control of new energy transmission systems extremely difficult. Parallel capacitors in new energy transmission systems are key equipment for achieving daily voltage regulation, and the switching control strategy of capacitor banks directly determines the voltage quality, transmission losses, and lifespan of power equipment in the new energy transmission system.

[0003] Existing new energy transmission systems mostly focus on the probability distribution characteristics of photovoltaic power output, generating representative daily scenarios through random sampling. This ignores the strong temporal fluctuations in photovoltaic power output, making it difficult to effectively cover boundary conditions such as irradiance limits and extreme power output distributions throughout the entire life cycle. It also fails to balance voltage regulation accuracy with switching frequency constraints, resulting in frequent changes in the ideal compensation capacity of capacitor banks within a day, approaching or even exceeding the switching frequency limit of mechanical switches, which can easily accelerate equipment aging and damage. Summary of the Invention

[0004] To address the aforementioned technical problems, the purpose of this application is to provide a method and system for switching reactive power compensation capacitors, the specific technical solution of which is as follows: In a first aspect, embodiments of this application provide a method for switching reactive power compensation capacitors, the method comprising the following steps: The distance from any weather station to each photovoltaic power station and the installed capacity of the photovoltaic power stations are statistically analyzed to select two representative weather stations. The waveform and peak characteristics of solar irradiance intensity of representative meteorological stations in each preset period are extracted to construct the space of solar irradiance intensity waveform characteristics and peak characteristics respectively. In any feature space, one period corresponds to one data point. In any feature space, all data points are clustered, and the area of ​​each cluster is calculated to divide all cycles into multiple scenarios. The neighboring points of each data point are obtained, the proportion of its neighboring points in the cluster where each data point is located is calculated, and the intra-cluster density screening coefficient of each data point is constructed by combining the distance of each data point to all its neighboring points in its cluster, so as to screen out typical cycles from various scenarios under any feature space. The extreme working condition points in all data points in any feature space are extracted using the two-dimensional convex hull method; the bounding box of any feature space and the extreme neighbor points of the extreme working condition points are obtained; based on the distances of the extreme working condition points to the bounding box and the extreme neighbor points, the extreme screening coefficients of the extreme working condition points are constructed to obtain the extreme scene period under any feature space, and together with the typical period, they form a representative period. Obtain the capacitor switching strategy for each representative period to control the switching of capacitors in the current period.

[0005] Preferably, the selection of two representative weather stations includes: Based on the distance from any weather station to each photovoltaic power station, a distance weight is constructed between any weather station and each photovoltaic power station. The installed capacity of the photovoltaic power stations is then weighted, and the sum of the weighted results between any weather station and all photovoltaic power stations is used as the typical characteristic value of any weather station. The meteorological stations corresponding to the first two typical characteristic values ​​in the results of sorting the typical characteristic values ​​of a preset number of meteorological stations in descending order are taken as representative meteorological stations.

[0006] Preferably, the construction of the space regarding the waveform characteristics and peak characteristics of solar irradiance includes: The waveform characteristics and peak characteristics of two representative meteorological stations in each period are respectively formed into waveform characteristic binary groups and peak characteristic binary groups; By mapping the waveform feature tuple and peak feature tuple under the historical period to a two-dimensional space, the spatial representations of waveform features and peak features are obtained respectively.

[0007] Preferably, the division of all cycles into multiple scenarios includes: Obtain the first and third quartiles of the area of ​​all clusters, and classify the period corresponding to all data points in clusters whose area is less than or equal to the first quartile into a first-level scenario; The periods corresponding to all data points in clusters with an area greater than or equal to the third or fourth quartile are classified as Level 3 scenarios; the periods corresponding to all data points in the remaining clusters are classified as Level 2 scenarios.

[0008] Preferably, the process for constructing the intra-cluster density screening coefficients for each data point is as follows: Based on the distance of each data point to all its neighboring points in its cluster, the intra-cluster local density of each data point's cluster is constructed. The intra-cluster density screening coefficient of each data point is positively correlated with the proportion of its neighboring points in the cluster to which each data point belongs and the local density within the cluster.

[0009] Preferably, the step of selecting typical cycles from various scenarios under any feature space includes: For any feature space, in the first-level scenario, the period corresponding to the data point with the maximum intra-cluster density screening coefficient is denoted as the typical period; In the secondary scenario, the maximum half-side length of the bounding box of each cluster is divided into two equal parts, thus dividing the bounding box into two sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period. In the three-level scenario, the maximum half-side length of the bounding box of each cluster is divided into three equal parts, thereby dividing the bounding box into three sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period.

[0010] Preferably, the extreme screening coefficients of the extreme operating points are negatively correlated with the shortest distance from the extreme operating point to the bounding box and the distance from the extreme operating point to its extreme neighbor.

[0011] Preferably, obtaining the extreme scene period under any feature space includes: In any two-dimensional convex hull of a feature space, the extreme working condition point corresponding to the maximum extreme screening coefficient is taken as the starting point of the convex hull. Along the boundary of the two-dimensional convex hull, the two-dimensional convex hull is divided into a preset number of sub-segments. On each sub-segment, the period corresponding to the extreme working condition point with the maximum extreme screening coefficient is recorded as the extreme scene period.

[0012] Preferably, the switching control of the capacitor in the current cycle includes: Obtain the photovoltaic power forecast for the current period and the representative period closest to it, and control the switching of capacitors in the current period according to the capacitor switching strategy of the representative period.

[0013] Secondly, embodiments of this application also provide a reactive power compensation capacitor switching control system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any of the above-described reactive power compensation capacitor switching control methods.

[0014] This application has at least the following beneficial effects: Based on a dual weighting mechanism of geographical distance and installed capacity, this application selects two representative meteorological stations with the most cluster representativeness and reactive power influence. This effectively avoids the shortcomings of a single meteorological station being unable to reflect the spatial heterogeneity of irradiance, and also avoids the data dimension explosion problem caused by full access. This lays a reliable data foundation for the subsequent accurate and lightweight extraction of meteorological waveform features throughout the entire life cycle. Furthermore, this application extracts waveform and peak features of representative meteorological stations and constructs a two-dimensional feature space, decoupling and mapping high-dimensional and complex time-series meteorological data into an intuitive distribution of discrete data points. This not only accurately captures the two core dimensions of the frequency of switching the dominant capacitor and the compensation extreme value, but also significantly reduces the data scale and computational complexity, providing a clear and reliable topological basis for subsequent efficient clustering and screening of typical scenarios and boundary extreme conditions. Furthermore, this application adaptively controls the extraction quota of typical cycles through clustering and area grading, and accurately locates the meteorological consistency core in each region based on the cluster density screening coefficient. This effectively avoids the loss of boundary fluctuation information and the pollution of representative features by mixed modes in a broad scenario, ensuring that the selected typical cycles can both macroscopically cover diverse meteorological patterns and microscopically retain pure high-frequency waveform features. This provides a precise and reliable data foundation for the subsequent formulation of highly robust and low-frequency reactive power switching strategies. Furthermore, this application accurately captures low-density boundary points outside high-density clusters using the two-dimensional convex hull method, and uses extreme screening coefficients to quantitatively evaluate their destructive force approaching physical boundaries and the risk of omissions due to deviations from typical patterns. This allows for the extraction of extreme scenario cycles, which are then integrated with typical cycles to construct a complete representative cycle that takes into account both normal operation and extreme impacts. This fundamentally avoids the insufficiency or excess of reactive power compensation configuration under extreme conditions, ensuring the optimal synergy between voltage regulation safety and equipment lifespan throughout the entire life cycle. Ultimately, this application transforms the representative cycle into electrical control commands executable by the power grid. By constructing a time-series sample matrix to drive static reactive power optimization, it establishes a capacitor configuration and switching strategy library that takes into account both normal voltage regulation and extreme defense. In the online control phase, it adaptively calls the optimal strategy based on power sequence waveform matching, effectively limiting the daily operating frequency of the capacitor bank, avoiding insufficient or excessive compensation, and synergistically improving voltage regulation accuracy, equipment lifespan, and operational safety. Attached Figure Description

[0015] Figure 1 A flowchart illustrating the steps of a reactive power compensation capacitor switching control method provided in one embodiment of this application; Figure 2 This is a schematic diagram of scene division provided for one embodiment of this application. Detailed Implementation

[0016] The following description, in conjunction with the accompanying drawings, details the specific scheme of the reactive power compensation capacitor switching control method and system provided in this application.

[0017] Please see Figure 1 The diagram illustrates a flowchart of a reactive power compensation capacitor switching control method according to an embodiment of this application. The method includes the following steps: Step S1: Calculate the distance from any weather station to each photovoltaic power station and the installed capacity of the photovoltaic power stations in order to select two representative weather stations.

[0018] Large-scale renewable energy clusters have a high proportion of renewable energy, weak synchronous support, and long transmission distances, resulting in large reactive power losses and voltage instability. During the planning stage, it is necessary to ensure that reactive power compensation resources can cover long-term complex operating scenarios.

[0019] This embodiment takes a photovoltaic cluster sending-end system as an example. The system includes N photovoltaic power stations and M meteorological stations (N≥2, M≥2). The geographical latitude and longitude of the photovoltaic power stations and meteorological stations are obtained, and the latitude and longitude distance between them is calculated using the Haversine formula.

[0020] The planning stage of the new energy cluster transmission system needs to cover the entire life cycle of photovoltaic power plants. The theoretical power generation of actual photovoltaic power plants needs to be corrected for curtailment under the current transmission capacity limit. The TMY data of meteorological observation stations, as natural meteorological parameters, are not affected by the operation strategy of photovoltaic clusters. The data source is reliable and standardized, which not only ensures the typical meteorological characteristics under long-term coverage, but also avoids the distortion of power generation data caused by the operation strategy of photovoltaic power plants such as curtailment and power restriction.

[0021] Because photovoltaic clusters cover a wide area, the solar irradiance received by photovoltaic power stations in different geographical locations varies significantly due to meteorological factors such as cloud movement and aerosol distribution. If only a single representative meteorological station is selected, it cannot reflect the spatial unevenness of solar irradiance distribution. If all meteorological stations are selected, it will lead to an excessively high feature space dimension and an explosive amount of computational burden.

[0022] Therefore, this embodiment selects two representative weather stations by statistically analyzing the distance from any weather station to each photovoltaic power station and the installed capacity of the photovoltaic power stations. Specifically: In this embodiment, a distance weight is constructed between any weather station and each photovoltaic power station based on the distance from any weather station to each photovoltaic power station. The installed capacity of the photovoltaic power stations is then weighted, and the sum of the weighted results between any weather station and all photovoltaic power stations is used as the typical characteristic value of any weather station.

[0023] It should be noted that the process of constructing the typical characteristic values ​​of any weather station is as follows: First, the distance weight is calculated. In this embodiment, the expression for the distance weight between any weather station and each photovoltaic power station is: ; In the formula, This represents the distance weight between weather station m and photovoltaic power station n; This represents the distance from weather station m to photovoltaic power station n; represents the maximum distance among all weather stations to photovoltaic power station n; M represents the number of all weather stations; exp() represents an exponential function with the natural constant as the base.

[0024] Furthermore, the installed capacity of photovoltaic power stations is weighted, and the sum of the weighted results between any weather station and all photovoltaic power stations is used as the typical characteristic value of any weather station. Specifically: In this embodiment, the expression for the typical characteristic value of any weather station is: ; In the formula, This represents a typical characteristic value of weather station m; This represents the distance weight between weather station m and photovoltaic power station n; Let n represent the installed capacity of photovoltaic power station n; N represents the total number of photovoltaic power stations.

[0025] Based on the typical eigenvalue, it can be understood that the typical eigenvalue is used to characterize the comprehensive representativeness and importance of any meteorological station in the entire photovoltaic cluster space. Its dimension is megawatt (MW). This is because the distance weight, as a dimensionless proportionality coefficient, naturally inherits the dimension of the installed capacity of the photovoltaic power station after being multiplied by the megawatt installed capacity. The calculation of the typical eigenvalue is affected by both the geographical distance from the meteorological station to each photovoltaic power station and the installed capacity of each photovoltaic power station. The closer the meteorological station is to the key photovoltaic power station, the greater the distance weight. The larger the capacity, the larger the typical characteristic value of the corresponding meteorological station, reflecting that the irradiance data of the corresponding meteorological station can influence the photovoltaic reactive power regulation range and the more important its core position. If it is selected as the representative meteorological station, the reactive power spatial characteristics of the cluster can be preserved to the greatest extent. Conversely, when the meteorological station is far away from the high-capacity photovoltaic area, the distance weight and weighted result are both small. The smaller the typical characteristic value, the more it reflects that the corresponding meteorological station can only represent the edge of the cluster or a local area with low capacity, and its impact on the overall reactive power planning is weak. If it is mistakenly selected as the representative station, it will lead to the loss of the spatial heterogeneity of irradiance in key areas.

[0026] Furthermore, the meteorological stations corresponding to the top two typical feature values ​​in the descending sort of the typical feature values ​​of a preset number (M) of meteorological stations are used as representative meteorological stations for each photovoltaic power station. The two selected representative meteorological stations are not random geographical identifiers, but data anchors that carry most of the reactive power fluctuation information of the entire photovoltaic cluster. They respectively dominate the irradiance characteristics of different weighted areas within the cluster. This selection mechanism based on dual weighting of geographical distance and installed capacity ensures that the selected stations spatially cover the core areas with the highest installed capacity and the greatest reactive power impact. While effectively compressing the dimensionality of meteorological data and avoiding the explosion of subsequent feature search calculations, it preserves the spatial heterogeneity of solar irradiance to the greatest extent, so that the features extracted based on these two stations can truly and completely reflect the photovoltaic output fluctuations and extreme shocks faced by the entire sending-end system.

[0027] Thus, this embodiment, based on a dual weighting mechanism of geographical distance and installed capacity, selects two representative meteorological stations with the most cluster representativeness and reactive power influence. This effectively avoids the shortcomings of a single meteorological station in reflecting the spatial heterogeneity of irradiance, and also avoids the data dimension explosion problem caused by full access. This lays a reliable data foundation for the subsequent accurate and lightweight extraction of meteorological waveform features throughout the entire life cycle.

[0028] Step S2: Extract the waveform and peak characteristics of solar irradiance intensity of each representative meteorological station within each preset period to construct the space of solar irradiance intensity waveform characteristics and peak characteristics respectively. In any feature space, one period corresponds to one data point.

[0029] The switching frequency of reactive power compensation devices (especially parallel capacitors) in the power transmission system of new energy clusters is limited by the mechanical life of the switches, and there are strict constraints on the number of daily operations. Therefore, the switching decision of capacitors depends not only on the reactive power demand at the peak of photovoltaic output, but also on the daily fluctuation characteristics of photovoltaic output, which is mainly related to solar irradiance.

[0030] However, the typical meteorological year (TMY) data contains 8760 hours of observation data from various meteorological stations over 365 days a year, resulting in extremely high data dimensionality. Directly using this data for reactive power compensation configuration optimization would lead to an explosion in computational load and would fail to effectively extract the key meteorological patterns for the switching of the dominant capacitors. Therefore, it is necessary to reduce the meteorological days to extract a representative daily sample set that covers the entire lifecycle of the photovoltaic power plant. Since a single meteorological observation station cannot reflect the spatial heterogeneity of a broad area within a photovoltaic cluster, and a single-dimensional indicator cannot simultaneously characterize the "fluctuation frequency" and "peak intensity" of irradiance, this embodiment extracts the waveform and peak characteristics of solar irradiance intensity from representative meteorological stations within each preset period to construct spatial representations of the waveform and peak characteristics of solar irradiance intensity. This maps the annual time-series meteorological data into a spatial distribution of discrete data points. The specific construction process of the feature space is as follows: First, collect the previous year's TMY data from each representative meteorological station, including latitude and longitude, altitude, timestamp, solar irradiance, and temperature information.

[0031] Furthermore, the fractal dimension of the solar irradiance intensity of each representative meteorological station within each preset period is calculated as a quantitative result of the waveform characteristics of the solar irradiance intensity of each representative meteorological station within each period. Specifically, the fractal dimension can be calculated using the Hurst exponent method, the box counting method, or the Higuchi method. In this embodiment, the Hurst exponent method is selected as the method for calculating the fractal dimension. In practical applications, as other implementation methods, implementers may also choose other fractal dimension calculation methods according to specific circumstances. This embodiment does not impose any special restrictions.

[0032] It should be noted that the calculation process of fractal dimension is a well-known technique, and will not be elaborated further.

[0033] Meanwhile, the maximum value of solar irradiance intensity of each representative meteorological station within each period is extracted as the quantitative result of the peak characteristics of solar irradiance intensity of each representative meteorological station within each period.

[0034] Furthermore, two feature spaces related to waveform features and peak features are constructed respectively, specifically: In this embodiment, for the feature space of waveform features, the quantization result of the waveform feature of the representative meteorological station with the largest typical eigenvalue among the two representative meteorological stations is used as the horizontal axis variable of the two-dimensional coordinate system, and the quantization result of the waveform feature of the representative meteorological station with the smallest typical eigenvalue is used as the vertical axis variable of the two-dimensional coordinate system. This allows the waveform feature tuples and peak feature tuples under historical cycles to be mapped to the two-dimensional space in the construction process of the feature space of waveform features, where a data point is a tuple composed of the quantization results of the waveform features of the two representative meteorological stations under one cycle. Similarly, for the feature space of peak features, the same construction process as for the feature space of waveform features is used to obtain the feature space of peak features. In the feature space of peak features, a data point is a tuple composed of the quantization results of the peak features of the two representative meteorological stations under one cycle. Each data point in any feature space corresponds to one cycle.

[0035] It should be noted that in this embodiment, the length of one cycle is 1 day. In actual application, as other implementation methods, implementers can also set the length of the cycle according to the specific circumstances. This embodiment does not impose any special restrictions.

[0036] Based on the two feature spaces, it can be understood that mapping the annual time-series meteorological data into a discrete data point distribution in a two-dimensional space is essentially a decoupled representation of the meteorological scene from two orthogonal dimensions: "time-series fluctuation frequency" and "spatial peak intensity". The waveform feature space focuses on the complexity and volatility of irradiance changes within the period, which determines the frequency of operation that the capacitor bank may need within the period. The peak feature space focuses on the absolute extreme value of irradiance, which determines the maximum reactive power deficit or power backflow extreme value that may be faced. The construction of these two feature spaces allows scenes with similar fluctuation patterns or similar peak intensities to naturally cluster in space, thereby transforming the high-dimensional time-series optimization problem that is difficult to process directly into an intuitive two-dimensional spatial density and boundary identification problem. This lays the topological foundation for subsequent fine-grained differentiation of typical patterns and capture of extreme operating conditions.

[0037] Thus, this embodiment extracts waveform and peak features representing meteorological stations and constructs a two-dimensional feature space, decoupling and mapping high-dimensional and complex time-series meteorological data into an intuitive distribution of discrete data points. This not only accurately captures the two core dimensions of the frequency of switching the dominant capacitor and the compensation extreme value, but also significantly reduces the data scale and computational complexity, providing a clear and reliable topological basis for subsequent efficient clustering and screening of typical scenarios and boundary extreme conditions.

[0038] Step S3: In any feature space, cluster all data points and calculate the area of ​​each cluster to divide all cycles into multiple scenarios; obtain the neighboring points of each data point, count the proportion of its neighboring points in the cluster where each data point is located, and combine the distance of each data point to all its neighboring points in its cluster to construct the intra-cluster density screening coefficient of each data point, so as to screen out typical cycles from various scenarios under any feature space.

[0039] After mapping the annual meteorological cycle to discrete data points in a feature space, the clustering and dispersion of different data points in space essentially reflect the similarity and differences in solar irradiance fluctuation patterns. However, facing a massive and interwoven distribution of data points, directly selecting typical days presents two major challenges: First, data points of different fluctuation patterns are mixed together, making it impossible to directly extract representative pattern features. Clustering must first be used to group daily scenes with similar waveform features into the same cluster (i.e., the same scene). Second, even within the same scene, the distribution of data points is uneven. Some scenes cover extremely rich fluctuation patterns (wide distribution range), while others have relatively simple fluctuation patterns (concentrated distribution). If a coarse strategy of "selecting one period from one cluster" is adopted for all scenes, it will lead to the loss of boundary fluctuation information under broad scenes, failing to meet the requirements of the capacitor switching strategy for fine-grained matching of fluctuation frequencies within the cycle.

[0040] Therefore, this embodiment adopts a progressive screening logic of "first clustering to divide the scene, then classifying by scene width, and finally selecting by density". That is, in any feature space, all data points are clustered, and the area of ​​each cluster is calculated to divide all cycles into multiple scenes; the neighboring points of each data point are obtained, the proportion of its neighboring points in the cluster of each data point is counted, and the intra-cluster density screening coefficient of each data point is constructed by combining the distance of each data point to all its neighboring points in its cluster, so as to select typical cycles from various scenes under any feature space. The specific process is as follows: First, in this embodiment, all data points are used as input to the clustering algorithm in any feature space. The neighborhood radius ε and the minimum number of points MinPts can be adaptively determined by the K-dist graph. The distance metric is the Euclidean distance between data points. Finally, all clusters are output.

[0041] It should be noted that there are many commonly used clustering algorithms. In this embodiment, the DBSCAN density clustering algorithm is used to cluster the data points. In practical applications, as other implementation methods, implementers may also use other clustering methods such as the k-means clustering algorithm according to the specific situation. This embodiment does not impose any special restrictions on the selection of clustering algorithms.

[0042] The process of clustering data points using the DBSCAN density clustering algorithm, determining the neighborhood radius and minimum number of points using the K-dist graph, and calculating the Euclidean distance are all well-known techniques and will not be elaborated further.

[0043] Furthermore, this embodiment calculates the area of ​​each cluster to divide all periods into multiple scenarios, specifically: In this embodiment, the first and third quartiles of the area of ​​all clusters are obtained, and the periods corresponding to all data points in clusters with an area less than or equal to the first quartile are classified as first-level scenarios. The periods corresponding to all data points in clusters with an area greater than or equal to the third or fourth quartile are classified as Level 3 scenarios; the periods corresponding to all data points in the remaining clusters are classified as Level 2 scenarios.

[0044] In particular, if the number of clusters in any feature space is less than 3, then the period corresponding to all data points in each cluster is classified as a level 3 scenario.

[0045] Preferably, the scene division diagram provided in this embodiment is as follows: Figure 2 As shown.

[0046] It should be noted that the area of ​​each cluster in this embodiment is calculated using the OBB bounding box method. The OBB bounding boxes of all data points in each cluster are obtained, and the area of ​​the OBB bounding boxes is used as the area of ​​each cluster. The principle of the OBB bounding box is a well-known technology and will not be described in detail here.

[0047] Based on the above scenario division process, it can be understood that dividing daily scenarios into three levels based on the quartiles of cluster area is essentially an adaptive granular stratification of the scenario space according to the richness or divergence of photovoltaic power output fluctuation patterns. Level 1 scenarios correspond to small cluster areas, reflecting that the weather patterns of this type are highly convergent and concentrated (such as typical sunny days), and only one core representative day is needed to cover its features. On the other hand, Level 3 scenarios correspond to large cluster areas, reflecting that the irradiance fluctuations under this type of weather are extremely discrete and diverse (such as complex cloudy days). If a single representative period is still used, a large amount of boundary fluctuation information will be lost, resulting in a mismatch of capacitor switching strategies in this scenario. Therefore, it is necessary to divide multiple sub-regions to finely preserve their internal differences. This dynamic hierarchical mechanism ensures that the computational resources for typical day screening are tilted towards complex scenarios with varied patterns, taking into account both the completeness of scenario coverage and the conciseness of the sample set.

[0048] Furthermore, this embodiment obtains the neighboring points of each data point, calculates the proportion of its neighboring points in the cluster to which each data point belongs, and combines the distance of each data point to all its neighboring points in its cluster to construct the intra-cluster density screening coefficient for each data point. Specifically: In any feature space, calculate the distance from each data point to all other data points, sort all distances in ascending order, and take the first preset number of data points in the sorted result as the nearest neighbors of each data point, where the preset value is the distance from data point i. The value is Where, min() represents the minimum value function, This represents the number of all data points in the cluster containing data point i; The minimum number of points adaptively determined by the K-dist graph.

[0049] Furthermore, this embodiment constructs an intra-cluster density screening coefficient for each data point by statistically analyzing the proportion of its neighboring points within the cluster to which each data point belongs, and combining this with the distance from each data point to all its neighboring points within its cluster. Specifically: Based on the distances of each data point to all its neighboring points in its cluster, the intra-cluster local density of each data point's cluster is constructed, specifically: As one implementation method, in this embodiment, the local density within the cluster of the cluster where data point i is located is... The expression is: In the formula, This represents the mean distance from data point i to all its neighbors in its cluster. This represents a preset constant greater than 0, used to prevent the denominator from being 0. In this embodiment... The value of is 0.01. Provided that the denominator is not zero and does not excessively affect the calculation result, the implementer may also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0050] Based on the concept of intra-cluster local density, it can be understood that intra-cluster local density is used to characterize the consistency of meteorological patterns within the local neighborhood of a data point. It reflects whether there are a large number of daily scenes with almost identical intraday waveform characteristics around the data point. Its dimension is 1 / distance unit (i.e., the reciprocal of distance), because the denominator is the sum of distances (with the dimension of distance), and after division, the dimension is the reciprocal of distance. The calculation of intra-cluster local density is directly affected by the distance from the data point to its neighboring points. When the distance from the data point to its neighboring points in its cluster is smaller, the intra-cluster local density is larger, reflecting that the corresponding data point is in a high-density meteorological consistency core area. This kind of fluctuation pattern is frequent and stable throughout the entire life cycle, and the switching strategy generated by using it as a typical period has strong universality. Conversely, when the distance from the data point to its neighboring points in its cluster is larger, the intra-cluster local density is smaller, reflecting that the point is on the edge of the cluster or in the mode transition zone, and its waveform characteristics are weakly representative. If it is selected as a typical period, it is easy to cause the strategy to be mismatched with the core frequent scene.

[0051] Furthermore, in this embodiment, based on the proportion of neighboring points in the cluster where each data point is located, and the local density within the cluster, the intra-cluster density screening coefficient of each data point is determined. Specifically: In this embodiment, the intra-cluster density screening coefficient of the data points in each period is positively correlated with the proportion of its neighboring points in the cluster where the data points are located in each period and the intra-cluster local density.

[0052] It should be understood that a positive correlation means that the dependent variable increases as the independent variable increases, and the dependent variable decreases as the independent variable decreases. The specific relationship can be additive or multiplicative, etc., and is determined by the actual application. This application does not impose any special restrictions.

[0053] Preferably, as one implementation method, in this embodiment, the intra-cluster density screening coefficient of data point i is... The expression is: In the formula, This represents the proportion of neighboring points of data point i in all data points within the cluster it belongs to; This represents the local density within the cluster of data point i.

[0054] Based on the intra-cluster density screening coefficient, it can be understood that the intra-cluster density screening coefficient is used to characterize the comprehensive superiority of a data point as a typical representative day. It reflects the performance of the data point in taking into account both the dual attributes of "high consistency of intra-cluster patterns" and "low interference between global clusters". When the proportion of neighboring points in the cluster where the data point is located is larger and the local density within the cluster is larger, the intra-cluster density screening coefficient is larger. This reflects that the corresponding data point is not only at the core of the meteorological consistency of the cluster, but its characteristics are not contaminated by the mixed patterns of other clusters. Selecting its corresponding period as a typical period can extract the purest and most stable control benchmark. Conversely, when the proportion of neighboring points in the cluster where the data point is located is smaller and the local density within the cluster is smaller, the intra-cluster density screening coefficient is smaller. This reflects that the corresponding data point is either at the boundary of the cluster and is subject to multi-mode interference, or the mode is isolated and has an extremely low frequency. The corresponding period does not meet the conditions to be a typical period.

[0055] Furthermore, this embodiment uses the intra-cluster density screening coefficient to select typical cycles from various scenarios under any feature space, specifically: For any feature space, in the first-level scenario, the period corresponding to the data point with the maximum intra-cluster density screening coefficient is denoted as the typical period; In the secondary scenario, the maximum half-side length of the bounding box of each cluster is divided into two equal parts. The bounding box region corresponding to the maximum half-side length of each half is taken as a sub-region. Thus, the bounding box is divided into two sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period. In the three-level scenario, the maximum half-side length of the bounding box of each cluster is divided into three equal parts. The bounding box region corresponding to each third of the maximum half-side length is taken as a sub-region. Thus, the bounding box is divided into three sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period.

[0056] Based on the above typical cycle selection process, it can be understood that by using a progressive selection logic of "quota based on scene width and attribution based on density coefficient," the extraction of typical cycles achieves the best balance between "macro-scale model coverage" and "pure micro-scale features." For primary scenes with concentrated distribution, a single maximum density point is sufficient to anchor its features. For secondary and tertiary scenes with discrete distribution, the method of dividing the region equally along the OBB bounding box axis and then searching for peaks ensures that a "local core" can be selected in different sub-intervals of the divergent pattern space, thereby avoiding the masking of edge fluctuation patterns by a single center point. This hierarchical adaptive selection mechanism enables the final extracted typical cycles to not only encompass the mainstream meteorological patterns that occur frequently throughout the year, but also accurately reflect the local fluctuation differences within each pattern, providing highly robust data support for the subsequent generation of anti-interference and low-action capacitor switching strategies.

[0057] Thus, this embodiment effectively avoids the loss of boundary fluctuation information and the contamination of representative features by mixed modes in a broad scenario by using clustering and area grading to adaptively control the extraction quota of typical cycles, and by relying on the intra-cluster density screening coefficient to accurately locate the meteorological consistency core in each region. This ensures that the selected typical cycles can both macroscopically cover diverse meteorological patterns and microscopically retain pure high-frequency fluctuation characteristics, providing a precise and reliable data foundation for the subsequent formulation of highly robust and low-frequency reactive power switching strategies.

[0058] Step S4: Use the two-dimensional convex hull method to extract extreme working condition points from all data points in any feature space; obtain the bounding box of any feature space and the extreme neighbor points of the extreme working condition points; based on the distances of the extreme working condition points to the bounding box and the extreme neighbor points, construct the extreme screening coefficients of the extreme working condition points to obtain the extreme scene period under any feature space, and form a representative period with the typical period.

[0059] Based on the typical cycles of each scenario obtained in step S3, it is already possible to cover the set of typical cycle scenarios that occur frequently and have high meteorological consistency throughout the entire life cycle. However, the reactive power compensation configuration of the new energy cluster sending-end system must also consider extreme operating conditions (such as irradiance limits, extreme distribution of photovoltaic output, etc.). Otherwise, the switching control strategy generated solely based on typical scenarios is prone to insufficient or excessive reactive power compensation in extreme high-generation or extremely low-generation scenarios.

[0060] However, in the feature space, extreme operating condition data points exhibit significant spatial distribution characteristics of "low density and high isolation." On the one hand, they often exist outside of high-density clusters, and the typical period selection logic based on the intra-cluster density screening coefficient not only fails to capture these points but also excludes them due to their extremely low density. On the other hand, extreme operating conditions usually manifest as extreme combinations of two representative meteorological station irradiance characteristics (such as both stations simultaneously experiencing extremely low irradiance or one station experiencing extremely high irradiance and the other extremely low irradiance), which must lie on the overall boundary formed by all data points in the two-dimensional space. Therefore, it is necessary to shift the screening approach from "searching for high-density cores inward" to "extracting low-density boundaries outward," that is, using the two-dimensional convex hull method to extract extreme operating condition points from all data points in any feature space; obtaining the bounding box of any feature space; and constructing extreme screening coefficients for extreme operating condition points based on the distances from extreme operating condition points to the bounding box and extreme neighboring points, to obtain the extreme scene period under any feature space, which is then combined with the typical period to form a representative period. The specific process is as follows: First, this embodiment uses the two-dimensional convex hull method to extract extreme operating condition points from all data points in any feature space. Specifically: In this embodiment, the smallest convex polygon containing all data points in any feature space is obtained using a two-dimensional convex hull algorithm, which serves as the two-dimensional convex hull. Since discrete data points may not exist on the line segment boundaries of the two-dimensional convex hull, extracting only the vertices of the convex hull may easily miss actual working points in the neighborhood of extreme boundaries. Therefore, this embodiment sets a boundary threshold. The shortest distance from the feature space to the boundary of the two-dimensional convex hull is less than or equal to All data points and all vertices of the two-dimensional convex hull are defined as extreme working condition points in the feature space to comprehensively reflect boundary conditions such as irradiance limits and extreme output distributions; in this embodiment, the two-dimensional convex hull algorithm specifically adopts the Graham scan method.

[0061] in, , The adaptive neighborhood radius is used in the clustering process of all data points in the feature space. The preset boundary scaling factor (in this embodiment, it is taken as...) In practical applications, implementers can set boundary thresholds according to specific circumstances, and this embodiment does not impose any special restrictions.

[0062] The process of obtaining the two-dimensional convex hull using the Graham scanning method is a well-known technique and will not be described in detail here.

[0063] Furthermore, this embodiment obtains the bounding box of any feature space, and constructs extreme screening coefficients for extreme operating conditions based on the distances from the extreme operating point to the bounding box and the extreme neighboring points, specifically: First, obtain the extreme neighbor points of each extreme working condition point in any feature space, specifically: In any feature space, the distances from each extreme operating point to all data points are sorted in ascending order. The first preset number of data points in the sorted result are taken as the extreme neighbors of each extreme operating point, where the preset number is... In the formula, This represents the number of all typical periods under any given feature space; This represents the function for rounding up.

[0064] Furthermore, in this embodiment, the extreme screening coefficient of the extreme operating point is negatively correlated with the shortest distance from the extreme operating point to the bounding box and the distance from the extreme operating point to its extreme neighboring point.

[0065] Preferably, as one implementation method, in this embodiment, the extreme operating condition point under the g-th feature space Extreme screening coefficient The expression is: In the formula, Represents the extreme operating condition point in the g-th feature space. The shortest distance to the enclosure box; Represents the extreme operating condition point in the g-th feature space. The mean distance to all its extreme neighbors; norm[ ] represents the normalization function; This represents a preset constant greater than 0, used to prevent the denominator from being 0. In this embodiment... The value of is set to 0.01. Under the premise of ensuring that the denominator is not 0 and does not excessively affect the calculation result, the implementer can also set it according to the specific situation. This embodiment does not impose any special restrictions.

[0066] It should be noted that this embodiment addresses extreme operating conditions. Normalization is performed using maximum and minimum values, where all extreme operating points are normalized. The maximum and minimum values ​​in the data are used as the maximum and minimum values ​​in the maximum-minimum normalization method for normalization. In practical applications, as other implementation methods, implementers may also choose other normalization methods according to specific circumstances. This embodiment does not impose special restrictions on the selection of normalization methods.

[0067] The process of normalizing data using the maximum-minimum normalization method is a well-known technique and will not be elaborated further.

[0068] Based on the extreme screening coefficient, it can be understood that the extreme screening coefficient is used to characterize the severity of reactive power impact at extreme operating points and the risk of being missed by existing typical cycles. It is a dimensionless index. The calculation of the extreme screening coefficient is jointly affected by the shortest distance from the extreme operating point to the bounding box and the distance to its extreme neighbor. When the shortest distance from the extreme operating point to the bounding box is smaller and the distance from the extreme operating point to its extreme neighbor is larger, the extreme screening coefficient is larger. This reflects that the corresponding extreme operating point not only approaches the theoretical physical boundary of the irradiation combination (which is extremely destructive to the reactive power impact of the system) but is also far from the existing typical daily coverage range (which is difficult for existing strategies to cope with). It is more necessary to forcibly include it in the representative cycle to prevent voltage over-limit. Conversely, when the shortest distance from the extreme operating point to the bounding box is larger and the distance from the extreme operating point to its extreme neighbor is smaller, the extreme screening coefficient is smaller. This reflects that although the corresponding extreme operating point is at the convex hull boundary, the actual deviation is limited, and its mode has been approximately covered by the adjacent typical cycle. Its marginal impact on the system compensation strategy is weak and can be disregarded.

[0069] Furthermore, this embodiment uses extreme screening coefficients based on extreme operating conditions to obtain the extreme scenario period under any feature space, and combines it with the typical period to form a representative period. Specifically: In the two-dimensional convex hull of any feature space, the extreme working condition point corresponding to the maximum extreme screening coefficient is taken as the starting point of the convex hull. The two-dimensional convex hull is divided into a predetermined number of sub-segments along the boundary of the two-dimensional convex hull, following an equal total distance of boundary polylines (i.e., equal boundary perimeter). In each sub-segment, the period corresponding to the extreme working condition point with the maximum extreme screening coefficient is denoted as the extreme scenario period. In this embodiment, the predetermined number is set to one-third of the total number of typical periods. Furthermore, all typical periods and all extreme scenario periods under all feature spaces are used as representative periods.

[0070] Based on the extreme scenario cycle and representative cycle, these cycles are not routine operating conditions that occur frequently in daily life. Instead, they represent the most severe voltage instability risks and challenges faced by photovoltaic power-end systems throughout the year (such as severe reactive power deficiency caused by extreme cold and lack of sunlight, or power backflow and capacity rise effects caused by extreme power generation). Forcibly extracting them from the convex hull boundary is to establish a "safety bottom line" for reactive power configuration, ensuring that the capacitor bank capacity and switching logic still have sufficient adjustment margin under the worst weather combinations. Based on the representative cycle, it can be understood as the union of typical cycle and extreme scenario cycle, forming a complete scenario library that "takes into account both normal and extreme conditions". It has both high-frequency typical strategies to support daily economic operation and defensive plans to resist extreme shocks. This allows the reactive power compensation static configuration and dynamic switching strategy generated based on this set to truly achieve multi-objective synergistic optimization of voltage regulation accuracy, equipment life and operational safety throughout the entire life cycle.

[0071] Thus, this embodiment accurately captures low-density boundary points outside high-density clusters using the two-dimensional convex hull method, and uses extreme screening coefficients to quantitatively evaluate their destructive force approaching physical boundaries and the risk of omissions due to deviations from typical patterns. This allows for the extraction of extreme scenario cycles, which are then integrated with typical cycles to construct a complete representative cycle that balances normal operation and extreme impacts. This fundamentally avoids the insufficiency or excess of reactive power compensation configuration under extreme conditions, ensuring the optimal synergy between voltage regulation safety and equipment lifespan throughout the entire lifecycle.

[0072] Step S5: Obtain the capacitor switching strategy library for each representative period to control the switching of capacitors in the current period.

[0073] The representative periods extracted in the above steps cover typical and extreme irradiance fluctuation patterns throughout the entire lifecycle of a photovoltaic power plant. However, the dimensionality reduction screening of the meteorological feature space only provides a data foundation for reactive power compensation configuration. To achieve reasonable capacitor switching, the meteorological representative days must be converted into electrical control commands executable by the power grid. Since the intraday photovoltaic output curves of different representative periods vary, their impact patterns on system reactive power and the required compensation levels are also drastically different. If only a single or random switching strategy is adopted, insufficient compensation in some scenarios will inevitably lead to voltage exceeding limits, or excessive compensation in others will cause frequent equipment operation. Therefore, this embodiment obtains a capacitor switching strategy library for each representative period to control the switching of capacitors in the current period. Specifically: Extract the theoretical power generation of the nth photovoltaic power station at each sampling time within the a-th representative period, specifically: The meteorological station closest to each photovoltaic power station is used as the matching meteorological station for each photovoltaic power station. The solar irradiance and temperature data of the matching meteorological station of the nth photovoltaic power station in the a-th representative period at each sampling time are used as input. Based on the photoelectric conversion mechanism of the photovoltaic power station and combined with the installed capacity of the photovoltaic power station, the solar irradiance and temperature are converted into the theoretical power generation corresponding to each sampling time as the output. The sampling frequency is set to once per hour. In actual application, as other implementation methods, implementers can also set it according to specific circumstances. This embodiment does not impose special restrictions.

[0074] Furthermore, the theoretical power generation of the nth photovoltaic power station at all sampling times within the a-th representative period is arranged in chronological order to form the power generation sequence of the nth photovoltaic power station in the a-th representative period, denoted as . To meet the input dimension requirements of the reactive power optimization allocation method, the power generation sequences representing all periods are concatenated row-wise to construct a representative time-series sample matrix Z: Where A represents the total number of all representative cycles.

[0075] Furthermore, the time-series sample matrix Z, representing the period, is used as the input to the static reactive power optimization configuration model of the large-scale photovoltaic cluster sending-end power grid. At the same time, the power grid topology parameters, line impedance, and node voltage constraint boundaries of the new energy cluster sending-end system are also input. Through optimization, the output results are obtained as follows: First, the static configuration of reactive power compensation required to be installed at each site of the new energy cluster sending-end system (including the number of parallel capacitor banks and their individual capacities, and the number of reactor banks and their individual capacities); Second, the capacitor switching control strategy corresponding to each representative period. When controlling the switching of capacitors in the current cycle, the new energy cluster sending-end system obtains the photovoltaic power generation prediction sequence for the current cycle based on the photovoltaic power prediction model. Then, it calculates the Euclidean distance between the photovoltaic power generation prediction sequence and the power generation sequence of each representative cycle, and selects the representative cycle with the smallest distance, i.e. the highest waveform matching degree, as the optimal matching cycle for the current cycle. Finally, it extracts the capacitor switching control strategy corresponding to the optimal matching cycle as the capacitor switching control strategy for the current cycle.

[0076] The calculation process of theoretical power generation, the process of obtaining the capacitor switching control strategy corresponding to each representative period, and the process of obtaining the photovoltaic power generation prediction sequence are all well-known technologies and will not be elaborated further.

[0077] Thus, this embodiment transforms the representative cycle into electrical control commands executable by the power grid. By constructing a time-series sample matrix to drive static reactive power optimization, it establishes a capacitor configuration and switching strategy library that takes into account both normal voltage regulation and extreme defense. In the online control phase, it adaptively calls the optimal strategy based on power sequence waveform matching, effectively limiting the daily operating frequency of the capacitor bank, avoiding insufficient or excessive compensation, and synergistically improving voltage regulation accuracy, equipment lifespan, and operational safety.

[0078] Based on the same inventive concept as the above method, this application embodiment also provides a reactive power compensation capacitor switching control system, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it implements the steps of any one of the above-described reactive power compensation capacitor switching control methods.

Claims

1. A method for switching reactive power compensation capacitors, characterized in that, The method includes the following steps: The distance from any weather station to each photovoltaic power station and the installed capacity of the photovoltaic power stations are statistically analyzed to select two representative weather stations. The waveform and peak characteristics of solar irradiance intensity of representative meteorological stations in each preset period are extracted to construct the space of solar irradiance intensity waveform characteristics and peak characteristics respectively. In any feature space, one period corresponds to one data point. In any feature space, all data points are clustered, and the area of ​​each cluster is calculated to divide all cycles into multiple scenarios. The neighboring points of each data point are obtained, the proportion of its neighboring points in the cluster where each data point is located is calculated, and the intra-cluster density screening coefficient of each data point is constructed by combining the distance of each data point to all its neighboring points in its cluster, so as to screen out typical cycles from various scenarios under any feature space. The extreme working condition points in all data points in any feature space are extracted using the two-dimensional convex hull method; the bounding box of any feature space and the extreme neighbor points of the extreme working condition points are obtained; based on the distances of the extreme working condition points to the bounding box and the extreme neighbor points, the extreme screening coefficients of the extreme working condition points are constructed to obtain the extreme scene period under any feature space, and together with the typical period, they form a representative period. Obtain the capacitor switching strategy for each representative period to control the switching of capacitors in the current period.

2. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The selection process identified two representative weather stations, including: Based on the distance from any weather station to each photovoltaic power station, a distance weight is constructed between any weather station and each photovoltaic power station. The installed capacity of the photovoltaic power stations is then weighted, and the sum of the weighted results between any weather station and all photovoltaic power stations is used as the typical characteristic value of any weather station. The meteorological stations corresponding to the first two typical characteristic values ​​in the results of sorting the typical characteristic values ​​of a preset number of meteorological stations in descending order are taken as representative meteorological stations.

3. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The construction of the space regarding the waveform and peak characteristics of solar irradiance includes: The waveform characteristics and peak characteristics of two representative meteorological stations in each period are respectively formed into waveform characteristic binary groups and peak characteristic binary groups; By mapping the waveform feature tuple and peak feature tuple under the historical period to a two-dimensional space, the spatial representations of waveform features and peak features are obtained respectively.

4. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The process of dividing all cycles into multiple scenarios includes: Obtain the first and third quartiles of the area of ​​all clusters, and classify the period corresponding to all data points in clusters whose area is less than or equal to the first quartile into a first-level scenario; The periods corresponding to all data points in clusters with an area greater than or equal to the third or fourth quartile are classified as Level 3 scenarios; the periods corresponding to all data points in the remaining clusters are classified as Level 2 scenarios.

5. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The process of constructing the intra-cluster density screening coefficients for each data point is as follows: Based on the distance of each data point to all its neighboring points in its cluster, the intra-cluster local density of each data point's cluster is constructed. The intra-cluster density screening coefficient of each data point is positively correlated with the proportion of its neighboring points in the cluster to which each data point belongs and the local density within the cluster.

6. The reactive power compensation capacitor switching control method as described in claim 4, characterized in that, The process of selecting typical cycles from various scenarios under any feature space includes: For any feature space, in the first-level scenario, the period corresponding to the data point with the maximum intra-cluster density screening coefficient is denoted as the typical period; In the secondary scenario, the maximum half-side length of the bounding box of each cluster is divided into two equal parts, thus dividing the bounding box into two sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period. In the three-level scenario, the maximum half-side length of the bounding box of each cluster is divided into three equal parts, thereby dividing the bounding box into three sub-regions. The period corresponding to the data point with the highest density screening coefficient in each sub-region is recorded as the typical period.

7. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The extreme screening coefficients of the extreme operating points are negatively correlated with the shortest distance from the extreme operating point to the bounding box and the distance from the extreme operating point to its extreme neighbor.

8. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The acquisition of the extreme scene period under any feature space includes: In any two-dimensional convex hull of a feature space, the extreme working condition point corresponding to the maximum extreme screening coefficient is taken as the starting point of the convex hull. Along the boundary of the two-dimensional convex hull, the two-dimensional convex hull is divided into a preset number of sub-segments. On each sub-segment, the period corresponding to the extreme working condition point with the maximum extreme screening coefficient is recorded as the extreme scene period.

9. The reactive power compensation capacitor switching control method as described in claim 1, characterized in that, The switching control of the capacitor in the current cycle includes: Obtain the photovoltaic power forecast for the current period and the representative period closest to it, and control the switching of capacitors in the current period according to the capacitor switching strategy of the representative period.

10. A reactive power compensation capacitor switching control system, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the reactive power compensation capacitor switching control method as described in any one of claims 1-9.

Images