Island detection method, device, equipment, medium and product

Abstract

The embodiment of the invention provides an island detection method, device and equipment, a medium and a product, and is applied to the technical field of power grid fault detection and protection. The method comprises the steps of collecting an electrical signal of a target power grid grid-connected point based on a preset frequency; performing signal processing and multi-time scale feature extraction based on the electrical signal to obtain a multi-dimensional feature parameter corresponding to each time scale; based on a preset timeliness weight corresponding to each time scale, performing weighted fusion calculation on the multi-dimensional feature parameter to obtain a fusion feature value of each time scale; based on the current operation state of the target power grid, determining a fusion coefficient of each time scale, and based on the fusion coefficient and the fusion feature value of each time scale, calculating to obtain fusion discriminant values corresponding to the plurality of time scales; and carrying out island judgment based on the fusion judgment value, the multi-dimensional feature parameter corresponding to each time scale and a preset non-island feature database to obtain an island judgment result. The technical effect of improving the island detection accuracy is achieved.

Description

Technical Field

[0001] This application relates to the field of power grid fault detection and protection technology, and in particular to an islanding detection method, device, equipment, medium and product. Background Technology

[0002] In grid-connected power generation systems for distributed energy sources such as photovoltaics and wind power, islanding protection is required at the connection point between the distributed energy source and the public power grid. By monitoring the grid operation status at the connection point in real time, when a grid fault causes an islanding effect between the distributed energy source and the local load, the islanding effect can be identified in a timely manner and quickly disconnected, thereby ensuring the safety of grid equipment and personnel.

[0003] In existing technologies, islanding detection mainly involves extracting voltage or current characteristics of the grid connection point at a certain time scale, and then performing model detection or rule base detection based on the extracted characteristics to determine whether an islanding effect exists.

[0004] Because existing island detection methods rely on feature extraction in a single dimension and at a single time scale, they cannot take into account multiple island characteristics, resulting in low accuracy in island detection. Summary of the Invention

[0005] This application provides island detection methods, apparatus, equipment, media, and products to achieve the technical effect of improving the accuracy of island detection.

[0006] In a first aspect, embodiments of this application provide an island detection method, including:

[0007] Electrical signals at the target power grid connection point are collected based on a preset frequency; these electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals.

[0008] Signal processing and multi-timescale feature extraction based on electrical signals are performed to obtain multi-dimensional feature parameters corresponding to each timescale; among which, the multi-timescale includes millisecond-level timescales and second-level timescales;

[0009] For each time scale, the multi-dimensional feature parameters corresponding to the time scale are weighted and fused based on the preset timeliness weights corresponding to the time scale to obtain the fused feature value corresponding to each time scale.

[0010] Based on the current operating status of the target power grid, the fusion coefficients corresponding to each time scale are determined, and based on the fusion coefficients and the fusion characteristic values ​​corresponding to each time scale, the fusion discrimination values ​​corresponding to multiple time scales are calculated.

[0011] Based on the fusion discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, island determination is performed to obtain the island determination result.

[0012] In one possible implementation, signal processing and multi-timescale feature extraction are performed based on electrical signals to obtain multi-dimensional feature parameters corresponding to each time scale, including:

[0013] The current total harmonic distortion rate at the grid connection point is generated based on the electrical signal.

[0014] The length of the filter window is determined based on the total harmonic distortion.

[0015] Based on the filter window length, preset interference compensation factor, and preset weight attenuation factor, the electrical signal is subjected to sliding filtering to obtain an electrical signal with interference suppression.

[0016] Based on preset outlier judgment conditions, abnormal data is detected in the electrical signals that have been suppressed for interference, and abnormal data is removed to obtain the electrical signals with outliers removed.

[0017] Based on a preset trend compensation factor, the electrical signals that have been removed due to anomalies are linearly interpolated to obtain standardized electrical signals.

[0018] Feature extraction is performed on the standardized electrical signal at each time scale to obtain multiple initial feature parameters corresponding to each time scale;

[0019] For each time scale, multiple initial feature parameters are corrected to obtain multiple standardized feature parameters after correction;

[0020] Based on multiple standardized feature parameters, the discrimination coefficient of each standardized feature parameter is calculated, and standardized feature parameters with discrimination coefficients lower than the first preset threshold are removed to obtain multiple target feature parameters.

[0021] Multiple target feature parameters at each time scale are integrated in a preset order, and each target feature parameter is mapped to a preset interval to obtain multi-dimensional feature parameters corresponding to each time scale.

[0022] In one possible implementation, for each time scale, a weighted fusion calculation is performed on the multi-dimensional feature parameters corresponding to the time scale based on a preset timeliness weight, to obtain the fused feature value corresponding to each time scale, including:

[0023] For each time scale, the discriminative weight of the feature parameter is calculated based on the discriminative coefficient corresponding to each feature parameter in the multi-dimensional feature parameters.

[0024] The weighted fusion calculation is performed based on the discriminative weight of each feature parameter, the parameter value of each feature parameter in the multi-dimensional feature parameters, and the preset timeliness weight corresponding to the time scale, to obtain the fused feature value corresponding to the time scale.

[0025] In one possible implementation, the fusion coefficients corresponding to each time scale are determined based on the current operating state of the target power grid, and based on the fusion coefficients and the fusion characteristic values ​​corresponding to each time scale, fusion discriminant values ​​corresponding to multiple time scales are calculated, including:

[0026] Based on the operating status of the target power grid, the fusion coefficient corresponding to each time scale is calculated; where the operating status includes the current voltage change rate, photovoltaic power change rate, and islanding type identification factor of the target power grid.

[0027] Based on the fusion coefficient, the fusion feature values ​​corresponding to each time scale, and the preset cross-time scale correction terms, a weighted fusion calculation is performed to obtain the fusion discrimination value; among which, the preset cross-time scale correction terms include preset cross-time scale compensation terms, preset stability adjustment terms, and preset error correction terms.

[0028] In one possible implementation, island determination is performed based on the fusion discriminant value, multi-dimensional feature parameters corresponding to each time scale, and a preset non-island feature database to obtain island determination results, including:

[0029] The target time scale for similarity matching is determined based on the operating status of the target power grid;

[0030] Similarity matching is performed based on the multi-dimensional feature parameters corresponding to the target time scale and a pre-set non-isolated feature database to obtain similarity matching values;

[0031] When the similarity matching value is greater than or equal to the second preset threshold, it is determined that the current target power grid has a non-islanding disturbance;

[0032] When the similarity matching value is less than the second preset threshold, and the fusion discrimination value and the island state data meet the first preset discrimination condition, it is determined that the grid connection point has an island effect; wherein, the first preset discrimination condition includes: the fusion discrimination value is greater than or equal to the sum of the third preset threshold and the preset discrimination correction term, the duration of the island effect is greater than the fourth preset threshold, and the difference between the island effect discrimination time and the triggering time is less than the fifth preset threshold.

[0033] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the second preset discrimination condition, the target power grid is determined to be operating normally; wherein, the second preset discrimination condition includes: the fusion discrimination value is less than the difference between the third preset threshold and the preset discrimination correction term;

[0034] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the third preset discrimination condition, the target power grid is determined to be in an uncertain state; wherein, the third preset discrimination condition includes: the fusion discrimination value is greater than or equal to the difference between the third preset threshold and the preset discrimination correction term, and is less than the sum of the third preset threshold and the preset discrimination correction term.

[0035] In one possible implementation, after obtaining the island determination result, the method further includes:

[0036] When the islanding determination result indicates that there is a non-islanding disturbance in the target power grid, the blocking protection action is executed;

[0037] When the islanding determination result indicates that the target power grid has an islanding effect, the protection action is executed;

[0038] When the islanding determination result indicates that the target power grid is in an uncertain state, the monitoring time is extended and the determination is re-performed.

[0039] Secondly, embodiments of this application provide an island detection device, comprising:

[0040] The acquisition module is used to acquire electrical signals at the target power grid connection point based on a preset frequency; wherein the electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals;

[0041] The first processing module is used to perform signal processing and multi-timescale feature extraction based on electrical signals to obtain multi-dimensional feature parameters corresponding to each timescale; wherein, the multi-timescale includes millisecond-level timescale and second-level timescale;

[0042] The second processing module is used to perform weighted fusion calculation on the multi-dimensional feature parameters corresponding to each time scale based on the preset timeliness weight of the time scale, so as to obtain the fusion feature value corresponding to each time scale.

[0043] The third processing module is used to determine the fusion coefficients corresponding to each time scale based on the current operating status of the target power grid, and to calculate the fusion discrimination values ​​corresponding to multiple time scales based on the fusion coefficients and the fusion feature values ​​corresponding to each time scale.

[0044] The fourth processing module is used to determine islands based on the fusion discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, and to obtain the island determination result.

[0045] In one possible implementation, the first processing module is further configured to:

[0046] The current total harmonic distortion rate at the grid connection point is generated based on the electrical signal.

[0047] The length of the filter window is determined based on the total harmonic distortion.

[0048] Based on the filter window length, preset interference compensation factor, and preset weight attenuation factor, the electrical signal is subjected to sliding filtering to obtain an electrical signal with interference suppression.

[0049] Based on preset outlier judgment conditions, abnormal data is detected in the electrical signals that have been suppressed for interference, and abnormal data is removed to obtain the electrical signals with outliers removed.

[0050] Based on a preset trend compensation factor, the electrical signals that have been removed due to anomalies are linearly interpolated to obtain standardized electrical signals.

[0051] Feature extraction is performed on the standardized electrical signal at each time scale to obtain multiple initial feature parameters corresponding to each time scale;

[0052] For each time scale, multiple initial feature parameters are corrected to obtain multiple standardized feature parameters after correction;

[0053] Based on multiple standardized feature parameters, the discrimination coefficient of each standardized feature parameter is calculated, and standardized feature parameters with discrimination coefficients lower than the first preset threshold are removed to obtain multiple target feature parameters.

[0054] Multiple target feature parameters at each time scale are integrated in a preset order, and each target feature parameter is mapped to a preset interval to obtain multi-dimensional feature parameters corresponding to each time scale.

[0055] In one possible implementation, the second processing module is further configured to:

[0056] For each time scale, the discriminative weight of the feature parameter is calculated based on the discriminative coefficient corresponding to each feature parameter in the multi-dimensional feature parameters.

[0057] The weighted fusion calculation is performed based on the discriminative weight of each feature parameter, the parameter value of each feature parameter in the multi-dimensional feature parameters, and the preset timeliness weight corresponding to the time scale, to obtain the fused feature value corresponding to the time scale.

[0058] In one possible implementation, the third processing module is further configured to:

[0059] Based on the operating status of the target power grid, the fusion coefficient corresponding to each time scale is calculated; where the operating status includes the current voltage change rate, photovoltaic power change rate, and islanding type identification factor of the target power grid.

[0060] Based on the fusion coefficient, the fusion feature values ​​corresponding to each time scale, and the preset cross-time scale correction terms, a weighted fusion calculation is performed to obtain the fusion discrimination value; among which, the preset cross-time scale correction terms include preset cross-time scale compensation terms, preset stability adjustment terms, and preset error correction terms.

[0061] In one possible implementation, the fourth processing module is further configured to:

[0062] The target time scale for similarity matching is determined based on the operating status of the target power grid;

[0063] Similarity matching is performed based on the multi-dimensional feature parameters corresponding to the target time scale and a pre-set non-isolated feature database to obtain similarity matching values;

[0064] When the similarity matching value is greater than or equal to the second preset threshold, it is determined that the current target power grid has a non-islanding disturbance;

[0065] When the similarity matching value is less than the second preset threshold, and the fusion discrimination value and the island state data meet the first preset discrimination condition, it is determined that the grid connection point has an island effect; wherein, the first preset discrimination condition includes: the fusion discrimination value is greater than or equal to the sum of the third preset threshold and the preset discrimination correction term, the duration of the island effect is greater than the fourth preset threshold, and the difference between the island effect discrimination time and the triggering time is less than the fifth preset threshold.

[0066] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the second preset discrimination condition, the target power grid is determined to be operating normally; wherein, the second preset discrimination condition includes: the fusion discrimination value is less than the difference between the third preset threshold and the preset discrimination correction term;

[0067] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the third preset discrimination condition, the target power grid is determined to be in an uncertain state; wherein, the third preset discrimination condition includes: the fusion discrimination value is greater than or equal to the difference between the third preset threshold and the preset discrimination correction term, and is less than the sum of the third preset threshold and the preset discrimination correction term.

[0068] In one possible implementation, the fourth processing module is further configured to:

[0069] When the islanding determination result indicates that there is a non-islanding disturbance in the target power grid, the blocking protection action is executed;

[0070] When the islanding determination result indicates that the target power grid has an islanding effect, the protection action is executed;

[0071] When the islanding determination result indicates that the target power grid is in an uncertain state, the monitoring time is extended and the determination is re-performed.

[0072] Thirdly, embodiments of this application provide an electronic device, including: a memory and a processor;

[0073] The memory stores the instructions that the computer executes;

[0074] The processor executes computer execution instructions stored in memory, causing the processor to perform the first aspect above and various possible implementations of the first aspect.

[0075] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, are used to implement the first aspect and various possible implementations thereof.

[0076] Fifthly, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the first aspect and various possible implementations thereof.

[0077] This application provides an islanding detection method, apparatus, device, medium, and product. The method acquires multi-dimensional electrical signals from the target power grid connection point at a preset frequency. Signal processing and feature extraction at multiple time scales are performed on the electrical signals to obtain multi-dimensional feature parameters corresponding to each time scale. Based on the multi-dimensional feature parameters of each time scale, a weighted fusion calculation is performed on the multi-dimensional feature parameters at each time scale using preset time-sensitivity weights corresponding to the time scales to obtain a fused feature value for each time scale. The fusion coefficients corresponding to each time scale are determined based on the operating state of the target power grid. A comprehensive fusion discrimination value corresponding to multiple time scales is calculated using the fusion discrimination value, the multi-dimensional feature parameters corresponding to each time scale, and a preset non-islanding feature database. Islanding determination is performed on the target power grid using the fusion discrimination value, the multi-dimensional feature parameters corresponding to each time scale, and a preset non-islanding feature database to obtain the corresponding islanding determination result. Compared to existing technologies, this method improves the coverage of feature data by acquiring multi-dimensional signals; it ensures that the extracted features cover both instantaneous and steady-state islanding scenarios by utilizing multi-timescale feature extraction; and it achieves multi-timescale and multi-dimensional feature extraction and fusion through feature parameter fusion within and across timescales. The fusion coefficients are determined using the target power grid's operating state, improving the real-time performance of the fused feature extraction and ensuring its adaptability to the scenario. By combining multi-dimensional and multi-timescale feature extraction and fusion, and taking into account various islanding features, the technical effect of improving the accuracy of islanding detection is achieved. Attached Figure Description

[0078] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0079] Figure 1 Flowchart of the island detection method provided in this application Figure 1 ;

[0080] Figure 2 Flowchart of the island detection method provided in this application Figure 2 ;

[0081] Figure 3 Flowchart of the island detection method provided in this application Figure 3 ;

[0082] Figure 4 This is a schematic diagram of the island detection device provided in this application;

[0083] Figure 5 A schematic diagram of the structure of the electronic device provided in this application.

[0084] The accompanying drawings illustrate specific embodiments of this application, which will be described in more detail below. These drawings and descriptions are not intended to limit the scope of the concept in any way, but rather to illustrate the concept of this application to those skilled in the art through reference to particular embodiments. Detailed Implementation

[0085] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings denote the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this application as detailed in the appended claims.

[0086] In existing technologies, islanding detection mainly involves real-time monitoring of single-dimensional features such as voltage or current at the grid connection point to provide further anti-islanding protection. Specifically, this involves extracting voltage or current features at a certain time scale, and then performing model detection or rule-based detection based on the extracted features to determine whether an islanding effect exists.

[0087] However, existing technologies for island detection and protection rely on single-dimensional features and a single time scale for feature extraction, and feature fusion is mainly linear. This results in the extracted features failing to fully reflect the current state of the grid connection point. Furthermore, a single time scale cannot simultaneously match the characteristics of instantaneous and steady-state islands, leading to low accuracy in island detection in existing technologies.

[0088] To address the aforementioned technical challenges, this application proposes the following technical concept: Employing a dual-timescale (millisecond and second) hierarchical extraction of multi-dimensional features from grid-connected points, capturing instantaneous islanding abrupt changes at the millisecond level, and identifying steady-state islanding shifts at the second level, thus achieving comprehensive coverage of both types of islanding features. A hierarchical fusion architecture is constructed, encompassing both same-timescale and cross-timescale fusion. First, weighted fusion of features within the same scale is performed, followed by cross-scale collaborative fusion. Finally, based on the real-time operating status of the power grid, the fusion coefficient for cross-timescale fusion is dynamically determined, ensuring the fusion result adapts to the current power grid state. This achieves efficient and accurate identification of both instantaneous and steady-state islanding, thereby improving the technical accuracy of islanding detection.

[0089] The technical solution of this application and how the technical solution of this application solves the above-mentioned technical problems are described in detail below with specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments. The embodiments of this application will now be described with reference to the accompanying drawings.

[0090] Figure 1 Flowchart of the island detection method provided in this application Figure 1 ,like Figure 1 As shown, the method includes:

[0091] S101. Acquire electrical signals at the target power grid connection point based on a preset frequency.

[0092] In this step, the electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals. The preset frequency refers to a fixed sampling frequency that can capture instantaneous islanding abrupt changes; the grid connection point refers to the physical connection node between the distributed power source and the public power grid.

[0093] Optionally, the electrical signals can be acquired by using high-precision sensors pre-deployed at the grid connection point to collect the electrical signals at a preset frequency and generate a sequence of time and signal values ​​for further signal processing. The pre-deployed sensors include, but are not limited to, voltage transformers, current transformers, and data acquisition devices.

[0094] For example, the preset frequency is set to 1kHz, meaning data is collected every 1ms, and the specific value and collection time are recorded. If further feature extraction is required, a continuous electrical signal needs to be collected. This can be done by collecting a continuous 100ms of phase current signal, three-phase voltage signal, zero-sequence current signal, and zero-sequence voltage signal, and then using the time-aligned signal as the electrical signal for subsequent feature extraction.

[0095] Optionally, the preset frequency can be adjusted according to the actual interference intensity of the target power grid. When the interference intensity increases, the preset frequency can be increased; when the interference decreases, the sampling frequency can be decreased. For example, the latest preset frequency is 1kHz. When the interference of the target power grid is detected to increase, the preset frequency can be increased to 2kHz; when the interference of the target power grid is detected to decrease, the preset frequency can be decreased to 500Hz.

[0096] Optionally, the electrical signal includes signals of multiple dimensions. When acquiring the signal, a multi-channel acquisition method can be adopted. For example, a dual-channel synchronous signal acquisition method can be deployed, using the same preset frequency for signal acquisition. The two acquired signals are then time-aligned to obtain an electrical signal composed of multi-dimensional signals.

[0097] S102. Based on electrical signals, perform signal processing and multi-time-scale feature extraction to obtain multi-dimensional feature parameters corresponding to each time scale.

[0098] In this step, signal processing refers to filtering, outlier removal, interpolation, parameter correction, and redundant feature removal of the original electrical signal to achieve high-precision feature extraction. Multiple time scales include millisecond-level and second-level time scales. The millisecond-level time scale is used to capture instantaneous islanding features, while the second-level time scale is used to capture steady-state islanding features. Multi-dimensional feature parameters refer to the core features extracted from the electrical signal that reflect the islanding state. For example, multi-dimensional feature parameters include, but are not limited to: voltage amplitude deviation, current amplitude deviation, frequency deviation, phase deviation, voltage change rate, current change rate, total harmonic distortion rate, zero-sequence current deviation, and zero-sequence voltage deviation.

[0099] Alternatively, signal processing and multi-timescale feature extraction can be performed as follows:

[0100] The electrical signal undergoes a series of processing steps, including harmonic calculation, filtering, anomaly removal, and interpolation, to obtain a standardized electrical signal. Initial feature parameters corresponding to the standardized electrical signal are extracted at both millisecond and second time scales. These initial feature parameters are then processed through parameter correction, redundancy removal, and standardization mapping to obtain seven categories of multi-dimensional feature parameters for each of the two time scales. These multi-dimensional feature parameters are updated in real time as data acquisition progresses: every 10ms for the millisecond time scale and every 1 second for the second-scale scale.

[0101] Optionally, in addition to millisecond and second-level time scales, microsecond or minute-level time scales can be added. The microsecond-level time scale is suitable for ultra-high-speed sudden changes in UHV grid connection scenarios, while the minute-level time scale is appropriately suitable for long-term shifts in the steady-state islanding of microgrids.

[0102] It should be noted that the signal processing and feature extraction in this step are described below. Figure 2 Further detailed explanations are provided in the embodiments shown, and will not be repeated here.

[0103] S103. For each time scale, the multi-dimensional feature parameters corresponding to the time scale are weighted and fused based on the preset timeliness weight corresponding to the time scale to obtain the fused feature value corresponding to each time scale.

[0104] In this step, the preset timeliness weight refers to the pre-set weight coefficient for the core functions at different time scales. For example, the millisecond-level time scale focuses on real-time feature extraction, and its corresponding preset timeliness weight can be set closer to 1; the second-level time scale focuses on stable feature extraction, and its corresponding preset timeliness weight can be set slightly lower; thus distinguishing the two time scales and clarifying the time value of features at different time scales. The fused feature value refers to the comprehensive representation value of multi-dimensional features at each time scale, which can reflect the grid islanding characteristics at the corresponding time scale.

[0105] Alternatively, one possible implementation for calculating the fusion feature values ​​for each time scale is as follows:

[0106] S1031. For the multi-dimensional feature parameters at each time scale, calculate the discriminative weight of the feature parameters based on the discriminative coefficient corresponding to each feature parameter in the multi-dimensional feature parameters.

[0107] In this step, the discrimination coefficient refers to a coefficient that quantifies the degree of difference between a certain feature parameter in isolated and non-isolated scenarios. A larger discrimination coefficient indicates a stronger ability of the feature parameter to identify isolated scenarios and lower redundancy. The discrimination weight refers to the weighted coefficient of each feature parameter calculated using the discrimination parameter as the core, and the sum of the weighted coefficients of multiple feature parameters at a given time scale is 1.

[0108] It should be noted that the discrimination coefficient is calculated during the feature selection stage, and the discrimination coefficient corresponding to each feature parameter in the multi-dimensional feature parameters can be used directly here.

[0109] Optionally, the discriminative weight of each feature parameter is calculated as shown in Formula 1:

[0110]

[0111] in, It refers to the discrimination coefficient of the i-th feature parameter at time t; This refers to the scene adaptation factor at time t. This refers to the weighted smoothing factor at time t. t refers to the weight of the i-th feature parameter at time t-1; m refers to the number of feature parameters.

[0112] For example, the scenario adaptation factor is a dynamic factor determined based on the photovoltaic power and load type corresponding to the current target power grid, and the value range of the scenario adaptation factor can be set to [0.9, 1.1]. The weight smoothing factor is used to avoid excessive weight fluctuations, thereby improving the stability of the discrimination weight, and the weight smoothing factor can be set to 0.1. The discriminant coefficients for each feature parameter in the multi-dimensional feature parameters at a certain time scale are: ΔU=0.62, ΔI=0.65, Δf=0.68, Δφ=0.61, du / dt=0.71, di / dt=0.69, and THD=0.73. After correction by a scene adaptation factor of 1.0 and a weight smoothing term of 0.1, the normalized discriminant weights are: ΔU=0.132, ΔI=0.139, Δf=0.145, Δφ=0.130, du / dt=0.151, di / dt=0.147, and THD=0.156, with a total weight of 1. Here, ΔU refers to voltage amplitude deviation, ΔI refers to current amplitude deviation, Δf refers to frequency deviation, Δφ refers to phase deviation, du / dt refers to voltage change rate, di / dt refers to current change rate, and THD refers to total harmonic distortion.

[0113] It should be noted that the discrimination coefficient in this step is calculated based on the real-time operating status of the target power grid, and the discrimination coefficient is updated in real time in response to the updates of the multi-dimensional characteristic parameters. When allocating discrimination weights for the multi-dimensional characteristic parameters, the rationality of the discrimination weights for each characteristic parameter can also be monitored in real time. When an abnormal change in the discrimination weights is detected, the weights are corrected.

[0114] S1032. Based on the discrimination weight of each feature parameter, the parameter value of each feature parameter in the multi-dimensional feature parameters, and the preset timeliness weight corresponding to the time scale, a weighted fusion calculation is performed to obtain the fusion feature value corresponding to the time scale.

[0115] In this step, the fusion feature value corresponding to each time scale is calculated as the first-level fusion calculation. The purpose is to sum the product of the discrimination weight, feature parameter value and the preset timeliness weight, and calculate the comprehensive representation of multi-dimensional features at each time scale by adding an error compensation term.

[0116] Optionally, the calculation method for the fused feature values ​​is as shown in Formula 2:

[0117]

[0118] in, This represents the fusion feature value at the k-th time scale at time t; This refers to the discriminative weight of the i-th feature parameter at time t; This refers to the preset timeliness weight of the k-th time scale at time t; This refers to the parameter value of the i-th feature parameter at the k-th time scale at time t. This refers to the error compensation term, which is calculated as 0.01 × (1 - F) k (t-1)). It should be noted that the preset timeliness weight is also a dynamically updated parameter. Specifically, when the time scale is in the millisecond range, =1.0-0.001×t / 10; when the time scale is on the order of seconds, =0.8+0.0005×t / 100.

[0119] For example, when calculating the fusion feature value, calculations are performed at both millisecond and second time scales. The preset timeliness weight for the millisecond time scale is 1.0, and the preset timeliness weight for the second time scale is 0.8. Assuming the current time is t and the previous time is t-1, the error compensation term for the current time, calculated based on the fusion feature value from the previous time, is -0.025. The parameter values ​​and discrimination weights of the multi-dimensional feature parameters corresponding to the millisecond time scale at the current time t are as follows: parameter value 0.8, discrimination weight 0.132; parameter value 1.0, discrimination weight 0.139; parameter value 0.8, discrimination weight 0.145; parameter value 1.0, discrimination weight 0.130; parameter value 1.0, discrimination weight 0.151; parameter value 0.88, discrimination weight 0.147; parameter value 0.9, discrimination weight 0.156. Therefore, the calculated fusion feature value at time t at the millisecond time scale is approximately 3.55.

[0120] It should be noted that, in this step, the calculation of fused feature values ​​can be performed using either linear weighted fusion or neural network fusion. Neural networks are used to learn the non-linear relationships between features at different time scales, thereby fusing multiple feature parameters.

[0121] Optionally, after calculating the fusion feature value corresponding to each time scale, it is also necessary to perform a reasonableness check on the fusion feature value. Using a pre-set range of values ​​for the fusion feature value, it is determined whether the calculated fusion feature value exceeds this range. If it exceeds this range, the updates of multi-dimensional feature parameters and discrimination weights are triggered, and then the fusion feature value is recalculated. The range of values ​​for the fusion feature value can be set to 0-m, where m refers to the number of feature parameters retained in the multi-dimensional feature parameters.

[0122] Optionally, the calculated millisecond-level and second-level fusion feature values ​​need to be dynamically updated over time. The update period for the fusion feature values ​​at the millisecond time scale can be set to 10ms, and the update period for the fusion feature values ​​at the second time scale can be set to 1s.

[0123] S104. Determine the fusion coefficients corresponding to each time scale based on the current operating status of the target power grid, and calculate the fusion discrimination values ​​corresponding to multiple time scales based on the fusion coefficients and the fusion characteristic values ​​corresponding to each time scale.

[0124] In this step, the grid operating status refers to the core parameters corresponding to the implementation conditions of the target grid, such as the current voltage change rate, photovoltaic power change rate, and islanding type identification factor. The fusion coefficient refers to the weighted coefficient corresponding to different time scales. This weighted coefficient is assigned to the fusion feature value at each time scale, and the sum of the weighted coefficients at each time scale is 1. The fusion discriminant value is the final feature value obtained after fusion across time scales, used for further islanding determination.

[0125] Alternatively, one possible implementation for calculating the fusion discriminant value corresponding to multiple time scales is as follows:

[0126] S1041. Based on the operating status of the target power grid, calculate the fusion coefficient corresponding to each time scale.

[0127] In this step, the operating status includes the target grid's current voltage change rate, photovoltaic power change rate, and islanding type identification factor. The voltage change rate refers to the instantaneous speed of the voltage at the grid connection point, used to determine whether a sudden change has occurred in the target grid; the photovoltaic power change rate refers to the rate of change of photovoltaic power output during grid outages, used to reflect fluctuations in distributed generation output; the islanding type identification factor refers to the islanding type identifier determined based on the target grid's operating status. For example, the islanding type identification factor for an instantaneous islanding type can be set to 1.2; the islanding type identification factor for a steady-state islanding type can be set to 0.8; and the islanding type identification factor for an uncertain state can be set to 1.0.

[0128] Optionally, the fusion coefficient is calculated as follows: The real-time voltage change rate and photovoltaic power change rate of the target power grid are obtained to determine the islanding type identification factor; these three factors are substituted into the fusion coefficient calculation formula to obtain the millisecond-level fusion coefficient; based on the millisecond-level fusion coefficient, the second-level fusion coefficient is calculated, and a coefficient smoothing term is added for correction to ensure that the sum of the fusion coefficients corresponding to the two time scales is 1. The calculation of the fusion coefficient is shown in Formulas 3 and 4 below:

[0129]

[0130]

[0131] in, This refers to the millisecond-level fusion coefficient; This refers to the second-level fusion coefficient; This refers to the rate of change of voltage at time t; This refers to the instantaneous maximum voltage change rate of the island; This refers to the rate of change of photovoltaic power at time t; This refers to the maximum rate of change in photovoltaic power. This refers to the island type identification factor at time t; This refers to the smoothing term of the fusion coefficient. The calculation method for the smoothing term of the fusion coefficient is as follows: This item is used to avoid excessive fluctuations in the fusion coefficient and improve fusion stability.

[0132] For example, the maximum rate of change of instantaneous islanding voltage is preset to 0.5 kV / s; the maximum rate of change of photovoltaic power is preset to 500 kW / s. In the instantaneous islanding scenario: du / dt(t) ≥ 0.2 kV / s, α(t) is adjusted to 0.6-0.7, and β(t) is adjusted to 0.3-0.4 to accelerate the response speed of instantaneous characteristics and ensure rapid identification of instantaneous islanding. For example: du / dt = 0.4 kV / s, dP pv / dt=300kW / s, γ island =1.2. Calculate α(t) = 0.5 + 0.1 × (0.4 / 0.5) + 0.05 × (300 / 500) × 1.2 = 0.5 + 0.08 + 0.036 = 0.616, β(t) = 1 - 0.616 + 0.01 × (0.42 - 0.384) = 0.384 + 0.00036 ≈ 0.384.

[0133] In steady-state islanding scenarios: when du / dt(t) < 0.2 kV / s, α(t) is adjusted to below 0.5, and β(t) is adjusted to above 0.5 to increase the weight of steady-state features and ensure reliable identification of steady-state islands. For example: when du / dt = 0.06 kV / s, dP pv / dt=50kW / s, γisland =0.8. Calculate α(t) = 0.5 + 0.1 × (0.06 / 0.5) + 0.05 × (50 / 500) × 0.8 = 0.5 + 0.012 + 0.004 = 0.516, β(t) = 1 - 0.516 + 0.01 × (0.488 - 0.484) = 0.484 + 0.00004 ≈ 0.484.

[0134] In normal operating scenarios: du / dt(t) < 0.05kV / s, α(t) = 0.45 + 0.01 × α(t-1), β(t) = 0.55 - 0.01 × α(t-1). The fusion coefficients at this time are used to balance instantaneous and steady-state characteristics and avoid misjudgment.

[0135] S1042. Based on the fusion coefficient, the fusion feature values ​​corresponding to each time scale, and the preset cross-time scale correction term, a weighted fusion calculation is performed to obtain the fusion discrimination value.

[0136] In this step, the preset cross-timescale correction terms include a preset cross-timescale compensation term, a preset stability adjustment term, and a preset error correction term. The preset cross-timescale compensation term is used to balance the differences in feature values ​​across different timescales, the preset stability adjustment term is used to ensure consistent fusion of discriminant values, and the preset error correction term is used to correct the fusion error from the previous time step.

[0137] Optionally, the fusion discriminant value can be calculated as shown in Formula 5:

[0138]

[0139] in, This refers to the fusion discriminant value at time t; This refers to the fused feature value corresponding to a millisecond-level time scale; This refers to the fused feature values ​​corresponding to a timescale of seconds; This refers to the millisecond-level fusion coefficient; This refers to the second-level fusion coefficient. This refers to the preset cross-timescale compensation term at time t, calculated as follows: △ cross (t)=0.02×(F1(t)-F2(t)). This refers to the preset stability adjustment term, calculated as: ζ fusion2 (t)=0.01×(K(t-1)-K(t-2)). This refers to the preset error correction term, calculated as: η error (t) = 0.005 × (1 - K(t-1)).

[0140] For example, in a transient island scenario, at time t, the calculated fusion feature value F1(t) for the millisecond timescale is 0.132×0.8+0.139×1.0+0.145×0.8+0.130×1.0+0.151×1.0+0.147×0.88+0.156×0.9=3.57, and the fusion feature value F2(t) for the second timescale is 0.6. The corresponding millisecond-level fusion coefficient α(t) is 0.616, and the second-level fusion coefficient β(t) is 0.384. The calculated preset cross-timescale compensation term Δ... cross (t) = 0.02 × (3.57 - 0.6) = 0.0594, preset stability adjustment term ζ fusion2 (t) = 0.01 × (2.2 - 2.1) = 0.001, preset error correction term η error (t) = 0.005 × (1 - 2.1) = -0.0055. Therefore, the final fusion discriminant value K(t) is calculated as follows: K(t) = 0.616 × 3.57 × (1 + 0.0594) + 0.384 × 0.6 × (1 - 0.0594) + 0.001 - 0.0055 ≈ 0.616 × 3.78 + 0.384 × 0.564 + 0.001 - 0.0055 ≈ 2.32 + 0.217 + 0.001 - 0.0055 ≈ 2.53.

[0141] In a steady-state island scenario, at time t, the calculated fusion feature value F1(t) for the millisecond timescale is 0.8, and the fusion feature value F2(t) for the second timescale is 3.2. The corresponding millisecond-level fusion coefficient α(t) is 0.516, and the second-level fusion coefficient β(t) is 0.484. The calculated preset cross-timescale compensation term Δ... cross (t) = 0.02 × (0.8 - 3.2) = -0.048, with the preset stability adjustment term ζ. fusion2 (t) = 0.01 × (1.9 - 1.8) = 0.001, preset error correction term η error (t) = 0.005 × (1 - 1.9) = -0.0045. Therefore, the final fusion discriminant value K(t) is calculated as follows: K(t) = 0.516 × 0.8 × (1 - 0.048) + 0.484 × 3.2 × (1 + 0.048) + 0.001 - 0.0045 ≈ 0.516 × 0.762 + 0.484 × 3.354 + 0.001 - 0.0045 ≈ 0.393 + 1.623 + 0.001 - 0.0045 ≈ 2.01.

[0142] It should be noted that the calculation error of the fusion discriminant value needs to be controlled within 3% to ensure the accuracy of the discrimination result.

[0143] S105. Based on the fusion discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, island determination is performed to obtain the island determination result.

[0144] In this step, a pre-set non-islanding feature database stores common non-islanding disturbance characteristics of the power grid, including standardized feature parameter ranges corresponding to disturbances such as voltage sags, frequency fluctuations, and load surges. Islanding determination refers to performing islanding detection on the target power grid connection point through non-islanding disturbance identification, fusion of discriminant value correspondence, and multi-condition comprehensive verification to determine whether the target power grid is currently in a non-islanding disturbance, islanding effect, normal operation, or uncertain state.

[0145] Optionally, the islanding determination result is obtained by determining the target time scale based on the operating status of the target power grid, performing similarity matching between its multi-dimensional feature parameters and the non-islanding feature database, and determining whether it is a non-islanding disturbance. If not, the fused discriminant value is compared with the dynamic threshold, and a comprehensive verification is performed based on conditions such as islanding duration and discrimination lag time to finally obtain the islanding determination result. If the target power grid is determined to be a non-islanding disturbance, and its multi-dimensional feature parameters do not match the existing features in the non-islanding feature database, then a new disturbance type and features are determined and added to the non-islanding feature database.

[0146] It should be noted that the method for determining the island determination result is as follows: Figure 3 Further explanation will be provided in the embodiments shown, and will not be repeated here.

[0147] The islanding detection method provided in this application collects electrical signals from multiple dimensions at a target power grid connection point at a preset frequency. Signal processing and feature extraction at multiple time scales are performed on the electrical signals to obtain multi-dimensional feature parameters corresponding to each time scale. Based on the multi-dimensional feature parameters at each time scale, a weighted fusion calculation is performed on the multi-dimensional feature parameters at each time scale using preset time-sensitivity weights corresponding to the time scale, resulting in a fused feature value for each time scale. The fusion coefficients for each time scale are determined based on the operating state of the target power grid, and a comprehensive fusion discriminant value for multiple time scales is calculated using the fusion coefficients and fusion feature values. Using the fusion discriminant value, the multi-dimensional feature parameters at each time scale, and a preset non-islanding feature database, islanding determination is performed on the target power grid, yielding the corresponding islanding determination result. Compared with existing technologies, this method improves the coverage of feature data through multi-dimensional signal acquisition; ensures that the extracted features can cover both instantaneous and steady-state islanding scenarios through multi-time scale feature parameter fusion and cross-time scale feature fusion; and achieves multi-time scale and multi-dimensional feature extraction and fusion through feature parameter fusion within and across time scales. The fusion coefficients are determined by utilizing the target power grid's operating status, improving the real-time performance of fusion feature extraction and ensuring its adaptability to the scenario. By combining multi-dimensional and multi-timescale feature extraction and fusion, and taking into account various island features, the technical effect of improving the accuracy of island detection is achieved.

[0148] Figure 2 Flowchart of the island detection method provided in this application Figure 2 ,like Figure 2 As shown, the method includes:

[0149] S201. Generate the current total harmonic distortion rate of the grid connection point based on the electrical signal.

[0150] In this step, the total harmonic distortion rate (THD) refers to the ratio between the total effective value of harmonic components in the voltage signal at the grid connection point and the effective value of the fundamental frequency, reflecting the harmonic interference intensity of the target power grid. The THD is generated by removing anomalies from the voltage signal in the electrical signal, extracting the effective values ​​of the fundamental and harmonic frequencies using Fast Fourier Decomposition (FFT), calculating the THD based on the extracted effective values, and then adding harmonic attenuation compensation and fundamental frequency fluctuation correction to obtain the current THD.

[0151] For example, the total harmonic distortion rate can be calculated as shown in Formula 6:

[0152]

[0153] Where THD refers to the total harmonic distortion rate at the grid connection point at the current time t; n refers to the order of the harmonic voltage, and the range of the harmonic voltage here is the 2nd to 21st harmonic voltage in the Fourier decomposition, with the corresponding fundamental frequency being the fundamental frequency of the first decomposition. U n (t) refers to the effective value of the nth harmonic voltage at time t; λ n This refers to the harmonic attenuation coefficient, calculated as λ. n =0.01×n; U1(t) refers to the effective value of the fundamental voltage at time t; △ U1 This refers to the correction factor for fundamental voltage fluctuations, which can be set to 0.02 to correct for THD calculation deviations caused by fundamental voltage fluctuations; U1(e) refers to the effective value of the rated fundamental voltage. It should be noted that when the target power grid is operating normally, THD ≤ 2%, and when the grid is detected to be in an islanded state, THD ≥ 3%.

[0154] It should be noted that the total harmonic distortion rate in this step refers to the total harmonic distortion rate of voltage. The calculation logic of formula 6 can also be used to calculate the total harmonic distortion rate of current. The total harmonic distortion rates of voltage and current are then weighted and averaged to obtain the comprehensive total harmonic distortion rate, which is then used for subsequent calculations.

[0155] S202. Determine the length of the filter window based on the total harmonic distortion rate.

[0156] In this step, the filter window length refers to the number of window sampling points for the adaptive moving average filter.

[0157] For example, the filter window length is calculated as shown in Formula 7:

[0158]

[0159] Where N(t) refers to the length of the filter window at time t, and 3≤N(t)≤7; THD(t) refers to the total harmonic distortion rate at time t; int(·) represents the rounding operation.

[0160] For example, the total harmonic distortion rate at the current moment, calculated based on the electrical signal, is 2.8%, and the filter window length calculated according to Formula 7 is 5.

[0161] S203. Based on the filter window length, preset interference compensation factor, and preset weight attenuation factor, the electrical signal is subjected to sliding filtering to obtain an electrical signal with interference suppression.

[0162] In this step, the sliding filter process refers to performing a sliding average on the sampled values ​​of the continuous electrical signal based on the length of the filtering window, generating the filtered electrical signal point by point, with the aim of reducing high-frequency interference in the signal. The preset interference compensation factor is a coefficient used to compensate for high-frequency interference deviations in the power grid, and the preset weight attenuation factor is a coefficient used to weaken the interference of old sampling points, with the aim of giving higher weight to the sampling points updated at each time point, thereby improving the real-time performance of the filtered signal.

[0163] For example, the sliding filter processing method is shown in Equation 8:

[0164]

[0165] Where, x filter (t) refers to the filtered signal value at time t; x(t−k·T) s ) refers to tk·T in electrical signals s The original sampled value at time T; s This refers to the sampling period, which can be set to 1ms. △x disturb (t) refers to the high-frequency interference deviation value at time t, calculated as: Δx disturb (t)=x(t)−x avg (t), x avg (t) represents the mean of the first 5 sampling points, and N(t) refers to the length of the filtering window at time t. λ is the preset weight attenuation factor, and γ is the preset interference compensation factor.

[0166] For example, the filter window length N=5, the preset weight attenuation factor λ=0.15, the preset interference compensation factor γ=0.08, and the voltage sampling values ​​within the window at t=105ms are 100.2V, 100.1V, 99.9V, 99.8V, and 99.7V. Substituting these values ​​into the filter formula, the filtered voltage value is calculated to be 99.92V.

[0167] S204. Based on the preset abnormal value judgment conditions, abnormal data detection is performed on the electrical signal with interference suppression, and abnormal data is removed to obtain the electrical signal with abnormal data removed.

[0168] In this step, the preset outlier determination criteria refer to the electrical signal sampling point's acquired value exceeding the rated value by ±20%, or the difference between the acquired value and the values ​​of the three adjacent sampling points exceeding the average value by ±10%. Abnormal data refers to signal values ​​that deviate from the normal range due to acquisition failures, electromagnetic interference, signal transmission errors, or other reasons. The outlier removal method involves removing the signal value from the electrical signal sequence when it is determined to be abnormal.

[0169] S205. Based on the preset trend compensation factor, the electrical signals that have been removed from the anomaly are linearly interpolated to obtain standardized electrical signals.

[0170] In this step, linear interpolation refers to using linear fitting to fill in missing values ​​at the locations of discarded outlier data, based on the values ​​of adjacent valid sampling points, to ensure the continuity of the signal sequence. The preset trend compensation factor is a coefficient used to compensate for the changing trends of the electrical signal, preventing trend deviations between the interpolated data and the actual signal values.

[0171] Alternatively, the linear interpolation process can be performed as shown in Equation 9:

[0172]

[0173] Where x(t) refers to the interpolated signal value at time t; x(t-1), x(t-2), and x(t-3) refer to the effective sampled values ​​at times t-1, t-2, and t-3, respectively; and x(t+1) refers to the effective sampled value at time t+1. θ is a preset trend compensation factor, which can be set to 0.05.

[0174] For example, the voltage value at t=105ms is considered outlier and is discarded. The valid sampling points before and after this point are: t=104ms, voltage signal value 99.8V; t=103ms, voltage signal value 90.9V; t=102ms, voltage signal value 100.0V; t=106ms, voltage signal value 99.7V. With a preset trend compensation factor θ=0.05, substituting these values ​​into Formula 9 yields an interpolated signal value of 99.75V.

[0175] S206. Extract features from the standardized electrical signal at each time scale to obtain multiple initial feature parameters corresponding to each time scale.

[0176] In this step, the initial feature parameters refer to the raw feature parameters extracted from the standardized electrical signal, including various categories such as voltage amplitude deviation, current amplitude deviation, frequency deviation, phase deviation, voltage or current change rate, and total harmonic distortion rate. Feature extraction on a time scale refers to short-duration feature extraction from 1ms to 100ms when the time scale is in the millisecond range, and long-duration feature extraction from 1s to 5s when the time scale is in the second range.

[0177] Optionally, when performing feature extraction for different time scales, it is necessary to determine the duration of each time scale. The method for determining the duration of the time scale is shown in Equations 10 and 11:

[0178]

[0179]

[0180] Where, τ ms (t) represents the duration of the adaptive millisecond-level time scale; τ s (t) represents the duration on an adaptive second-level time scale; f s The sampling frequency can be set to 1kHz; N ms (t) represents the number of adaptive sampling points in milliseconds, N s (t) represents the number of adaptive sampling points on a second-by-second basis. The calculation methods for the two sampling point numbers are shown in Formulas 12 and 13:

[0181]

[0182]

[0183] Where, N ms (t) represents the number of adaptive sampling points in milliseconds, with a base value of 100, dynamically adjusted according to the photovoltaic power change rate. N s (t) represents the number of adaptive sampling points per second, with a base value of 1000 / second, dynamically adjusted according to the voltage deviation; dP pv (t) / dt is the rate of change of photovoltaic power at time t, in kW / s, (dP) pv )max is the maximum rate of change of photovoltaic power, which is preset to 500 kW / s. △U(t) is the voltage deviation at time t, U e This is the rated voltage of the power grid, for example, 10kV.

[0184] Optionally, feature extraction is performed for different time scales after determining the corresponding time scale duration. Specifically, millisecond-level feature extraction is based on N... ms Using (t) consecutive sampling points, six core features are extracted in real time: voltage and current amplitude deviation, frequency deviation, phase deviation, voltage and current change rate, and total harmonic distortion rate. The system focuses on capturing the abrupt changes in transient islanding characteristics, providing support for rapid identification of transient islands. Furthermore, feature values ​​are updated every 10ms to ensure fast response. A transient change amplification factor is added during the extraction process to enhance the characteristic differences of transient islands.

[0185] Second-level feature extraction is based on N s (t) consecutive sampling points are used, and the above six types of feature parameters are extracted once every 1 second. The weighted average value within 5 seconds is taken as the steady-state feature value. The focus is on capturing the offset characteristics of steady-state islands, providing support for reliable identification of steady-state islands and avoiding misjudgments caused by instantaneous disturbances. A steady-state offset compensation factor is added during the weighted averaging process to weaken the impact of instantaneous disturbances. Among them, steady-state islands refer to the long-term balance between photovoltaic power and load power when the grid-connected switch is manually disconnected.

[0186] For example, the current standardized electrical signal acquisition range is 0ms to 10000ms. The feature extraction for the millisecond time scale is as follows: from the voltage signal at t=200ms-300ms, the voltage amplitude deviation ΔU=2.8% and the voltage change rate du / dt=0.32kV / s are extracted. After adding a sudden change amplification factor, the initial feature parameters are ΔU=3.36% and du / dt=0.384kV / s. The feature extraction for the second time scale is as follows: from the signal at t=0-5s, the weighted average of the total harmonic distortion rate is extracted as 3.2%. After adding a steady-state compensation factor, the initial feature parameter total harmonic distortion rate is 3.04%.

[0187] S207. For each time scale, perform parameter correction on multiple initial feature parameters to obtain multiple standardized feature parameters after correction.

[0188] In this step, parameter correction refers to multi-dimensional compensation correction of the initial characteristic parameters, including but not limited to temperature drift compensation, grid impedance correction, transformer error correction, and harmonic phase correction.

[0189] Optionally, the initial characteristic parameters include: voltage amplitude deviation, current amplitude deviation, frequency deviation, phase deviation, voltage or current change rate, and total harmonic distortion rate.

[0190] The correction method for voltage amplitude deviation is shown in Formula 14:

[0191]

[0192] Where △U refers to the voltage amplitude deviation, expressed as a percentage, and ranging from 0 to 100%. U(t) is the effective value of the filtered grid-connected point voltage at time t. e This is the effective value of the rated voltage of the power grid, for example, 10kV; the corresponding effective value on the secondary side would be 100V. α T The temperature drift coefficient is calculated as follows: α T =1.2×10 -4 / ℃, used to correct voltage measurement deviations caused by temperature drift in the protection device hardware. T(t) is the internal temperature of the protection device at time t, and 25 is the standard reference temperature; all units are degrees Celsius. β Z This is the impedance correction factor for the power grid, and its value can be β. Z =0.02, this coefficient is used to correct voltage deviation caused by changes in grid impedance, and is a dimensionless value. ΔZ(t) is the grid impedance deviation at time t, calculated as: ΔZ(t) = Z(t) - Z eq Z(t) is the real-time impedance of the power grid at time t. eq This is the equivalent impedance of the power grid.

[0193] The correction method for current amplitude deviation is shown in Formula 15:

[0194]

[0195] Where △I refers to the current amplitude deviation, expressed as a percentage, and ranging from 0 to 100%. I(t) is the effective value of the filtered grid-connected current at time t. e This is the effective value of the rated current at the grid connection point. γ L This is the load fluctuation correction factor, and its value can be γ. L =0.03, this coefficient is used to correct for current deviations caused by load power fluctuations. △P L (t) represents the load power deviation at time t, P L (e) represents the rated load power. △ T This is the error correction coefficient for the current transformer, and its value can be Δ. T =0.015, this coefficient is used to correct the measurement error of the current transformer. △f(t) is the frequency deviation at time t.

[0196] The calculation method for frequency deviation is shown in Formula 16:

[0197]

[0198] Here, Δf refers to the frequency deviation, in Hz, and its value ranges from 0 to 1 Hz. f(t) is the real-time frequency at time t, in Hz; e The rated frequency of the power grid is, for example, 50Hz. Δφ(t) represents the voltage phase at time t, in rad. s (t) represents the voltage phase at time t-Ts. s The sampling period is, for example, 1 ms. △φ harm (t) represents the phase deviation caused by power grid harmonics at time t, calculated as: Δφ harm (t)=0.01×sum(n×U n (t) / U1(t)), where n is the harmonic order, U n Un(t) is the effective value of the nth harmonic voltage, and U1(t) is the effective value of the fundamental voltage. ξ is the voltage deviation correction factor, which can be set to ξ=0.1 to correct frequency calculation deviations caused by voltage deviations. ζ is the frequency drift correction factor, which can be set to ζ=0.005 to correct frequency drift caused by long-term operation. t is the operating time, in seconds. max The maximum continuous running time can be set to 86400s, or 24 hours.

[0199] The correction method for phase deviation is shown in Equation 17:

[0200]

[0201] Here, Δφ refers to the phase deviation, with units of rad (or °), and a value range of 0-π rad (0-180°). U (t) represents the voltage phase at time t, φ I (t) represents the current phase at time t. φ e This represents the rated phase difference between voltage and current during normal operation, for example, 18.19°, corresponding to 0.317 rad. Δφ of Δφ comp (t) represents the phase compensation amount at time t, used to correct grid harmonics and transformer phase errors. It is calculated as: Δφ comp (t) = 0.008 × THD(t)). η is the power factor correction coefficient, which can be set to η = 0.02. cosφ(t) is the actual power factor angle at time t. e This is the rated power factor angle.

[0202] The correction methods for the rate of change of voltage and the rate of change of current are shown in Equations 18 and 19:

[0203]

[0204]

[0205] Where τ=T s =1ms, representing the time interval. U(t)-U(t-τ) is the voltage difference between time t and time t-τ; I(t)-I(t-τ) is the current difference between time t and time t-τ. du / dt is the voltage change rate in kV / s, and di / dt is the current change rate in A / s. μ is the amplification factor for sudden changes, which can be set to μ=1.2 to amplify instantaneous sudden change signals and improve the sensitivity of instantaneous islanding identification. sgn(·) is the sign function, with 1 for positive sudden changes, -1 for negative sudden changes, and 0 for no sudden changes. κ is the interference suppression coefficient, which can be set to κ=0.05 to suppress the deviation in the rate of change calculation caused by high-frequency noise. △U noise (t), △I noise (t) represents the voltage and current noise deviation values ​​at time t, respectively.

[0206] The correction method for the total harmonic distortion rate is shown in Formula 6 above.

[0207] S208. Based on multiple standardized feature parameters, calculate the discrimination coefficient of each standardized feature parameter, and remove standardized feature parameters with discrimination coefficients lower than the first preset threshold to obtain multiple target feature parameters.

[0208] Optionally, the discrimination coefficient of each standardized feature parameter is calculated as shown in Formula 20:

[0209]

[0210] Where D refers to the discrimination coefficient of the i-th feature parameter; This refers to the mean value of this feature parameter in an isolated scenario. This refers to the mean value of the feature parameter in non-isolated scenarios. This refers to the standard deviation of the feature parameter in an isolated scenario. This refers to the standard deviation of the feature parameter in non-isolated scenarios. This refers to the feature importance weight corresponding to the i-th feature parameter. This refers to the scene adaptation factor. This refers to interference inhibitory factors. This refers to the similarity of non-island perturbations.

[0211] For example, the first preset threshold is set to 0.5; the calculated discrimination coefficients of various features are △U=0.62, △I=0.65, △f=0.68, △φ=0.61, du / dt=0.71, di / dt=0.69, and THD=0.73, all of which are ≥0.5 and are all retained as target feature parameters; the discrimination coefficient of the zero-sequence voltage feature is 0.38<0.5, which is determined to be a redundant feature and is removed.

[0212] S209. Integrate multiple target feature parameters at each time scale in a preset order, and map each target feature parameter to a preset interval to obtain multi-dimensional feature parameters corresponding to each time scale.

[0213] In this step, the preset order refers to the fixed arrangement of the target feature parameters, such as ΔU, ΔI, Δf, Δφ, du / dt, di / dt, and THD. This order ensures consistency in feature integration. The preset interval refers to the numerical range of 0-1, which is a unified interval for feature standardization, used to eliminate differences in the dimensions and value ranges of different features. Mapping refers to converting the original values ​​of the target feature parameters into values ​​within the 0-1 interval, achieving dimensionless feature representation.

[0214] Alternatively, the mapping method is as shown in Formula 21:

[0215]

[0216] Where, x norm x(t) represents the standardized feature parameter value at time t, ranging from 0 to 1. x(t) represents the original calculated value of a certain feature parameter at time t. min (t), x max(t) represents the minimum and maximum values ​​of the feature parameter within 100 ms before time t, respectively. These two values ​​are dynamically updated every 100 ms to adapt to changes in the scene. ε is a small positive number, and its value can be ε=10. -6 This is used to avoid cases where the denominator is 0, and also to prevent the standardized value from approaching 0, thus improving the stability of the fusion operation. △ offset (t) represents the dynamic offset at time t, calculated as Δ. offset (t) = 0.001 × t / 100, used to correct long-term drift of characteristic parameters, adjusted every 100 ms. γ dim (i) is the dimension correction factor for the i-th characteristic parameter, which is dynamically allocated according to the characteristic dimension, such as γ of ΔU. dim =1.0, γ of du / dt dim =0.98, used to balance the standardized results for different dimensional characteristics. ζ stab ζstab(t) is a stability adjustment term, calculated as: ζstab(t) = 0.005 × (1 - xnorm(t)), used to suppress fluctuations in the standardized results and improve stability. After standardization, the standardization error of all characteristic parameters must be ≤0.1%.

[0217] For example, the original values ​​of the millisecond-level target feature parameters are △U=3.32%, △I=2.8%, △f=0.06Hz, and THD=3.04%. After being integrated in a preset order and substituted into Formula 21, the multi-dimensional feature parameters obtained after mapping to the 0-1 interval are: △U=0.8, △I=1.0, △f=0.8, and THD=0.9.

[0218] Figure 3 Flowchart of the island detection method provided in this application Figure 3 ,like Figure 3 As shown, the method includes:

[0219] S301. Determine the target time scale for similarity matching based on the operating status of the target power grid.

[0220] In this step, the target grid operating state refers to the core state of the grid as reflected in real time at the grid connection point. In this embodiment, voltage change rate and photovoltaic power change rate are the core judgment indicators, divided into three categories: instantaneous change state, steady-state operating state, and attribute ambiguity state. This is the sole basis for selecting the time scale. The target time scale refers to a single time scale specifically selected for similarity matching of non-islanded disturbances. It is the sole feature data source for similarity calculation, avoiding matching interference caused by the mixing of features from two time scales. When determining the time scale, the adaptability of the time scale needs to be considered. A millisecond-level time scale adapts to instantaneous changes in the grid, while a second-level time scale adapts to steady-state operation. Essentially, this ensures that the feature data source for similarity matching is highly consistent with the temporal characteristics of the disturbance, thereby improving matching accuracy.

[0221] For example, the target time scale can be determined in the following ways:

[0222] a1. Extract the real-time voltage change rate and photovoltaic power change rate at the grid connection point.

[0223] a2. Determine the operating status of the power grid based on multiple preset thresholds:

[0224] Transient sudden change state: voltage change rate ≥ 0.2kV / s or photovoltaic power change rate ≥ 200kW / s, this parameter is the typical threshold for photovoltaic grid connection;

[0225] Steady-state operation: Voltage change rate < 0.05 kV / s and photovoltaic power change rate < 50 kW / s;

[0226] Attribute ambiguity state: an intermediate state between the above thresholds.

[0227] a3. Select the target time scale according to the adaptation rules of state and scale: select millisecond level for instantaneous changes; select second level for steady-state operation; for ambiguous attribute states, take the weighted fusion value of millisecond and second level features as the matching data source. Here, the adaptation rule refers to one state corresponding to one time scale.

[0228] S302. Based on the multi-dimensional feature parameters corresponding to the target time scale and the preset non-isolated feature database, perform similarity matching to obtain the similarity matching value.

[0229] In this step, the multi-dimensional feature parameters corresponding to the target time scale refer to the highly discriminative feature parameters after standardization and redundancy removal at the selected time scale. The preset non-islanding feature database refers to a database used to store standardized feature parameters of all common non-islanding disturbances in photovoltaic grid-connected scenarios. These standardized features include, but are not limited to, feature parameters corresponding to scenarios such as voltage sags, instantaneous frequency fluctuations, sudden load changes, and grid electromagnetic interference. Furthermore, each disturbance corresponds to a set of standard feature values ​​consistent with the dimensions of the feature to be matched, used for subsequent similarity calculations. Similarity matching refers to quantitatively calculating the similarity between the feature to be matched and the reference feature, converting it into a value between 0 and 1, thus achieving quantitative identification of non-islanding disturbances. The closer the similarity matching value is to 1, the more similar the feature to be matched is to the non-islanding disturbance reference feature; the closer it is to 0, the greater the difference.

[0230] Optionally, the similarity matching value is calculated as shown in Formula 22:

[0231]

[0232] Where S(t) is the feature similarity at time t, with a value ranging from 0 to 1. norm,i (t) represents the standardized value of the i-th feature parameter extracted at time t, x i,disturb (t) represents the standardized value of the i-th feature parameter of the corresponding perturbation in the non-islanded perturbation feature library at time t. m is the number of feature parameters retained, for example, m=7. W i (t) represents the weight of the i-th feature parameter at time t, i.e., the discrimination weight, used to highlight the role of high-discrimination features in similarity recognition. α disturb (i) is the disturbance type adaptation factor for the i-th feature parameter. This parameter is dynamically allocated according to the disturbance type, such as α of ΔU in the voltage sag scenario. disturb =1.2, other features α disturb =0.9, used to improve compatibility. △ similar (t) is the similarity correction term at time t, calculated as follows: △ similar (t)=0.02×(1-S(t-1)), which is used to correct the similarity calculation error in the previous time step and improve the recognition stability.

[0233] S303. When the similarity matching value is greater than or equal to the second preset threshold, it is determined that there is a non-islanding disturbance in the current target power grid.

[0234] In this step, the second preset threshold refers to the critical threshold used for similarity matching judgment. Non-island perturbation judgment refers to the qualitative identification of non-island perturbations through threshold comparison.

[0235] For example, if the second preset threshold is 0.8, and the calculated similarity matching value is greater than or equal to 0.8, then it is determined that the current power grid characteristic change is caused by a non-islanding disturbance, and a non-islanding disturbance is confirmed to exist. If S(t) < 0.8, then the non-islanding disturbance is excluded, and the process proceeds to the subsequent islanding effect judgment.

[0236] S304. When the similarity matching value is less than the second preset threshold, and the fusion discrimination value and the island state data meet the first preset discrimination condition, it is determined that there is an island effect at the grid connection point.

[0237] In this step, the first preset discrimination conditions include: the fusion discrimination value is greater than or equal to the sum of the third preset threshold and the preset discrimination correction term; the islanding effect duration is greater than the fourth preset threshold; and the difference between the islanding effect discrimination time and the triggering time is less than the fifth preset threshold. The third preset threshold refers to the basic judgment threshold for the fusion discrimination value, which can be dynamically optimized according to the grid status. The preset discrimination correction term refers to the dynamic correction value of the third preset threshold, used to adapt to different grid operating states and improve the accuracy of threshold judgment. Islanding status data includes the duration of the islanding effect, as well as the islanding effect discrimination time and the triggering time. The islanding effect duration refers to the duration from the first occurrence of the islanding characteristic change at the grid connection point to the current moment.

[0238] For example, when it is determined that there are no non-islanding disturbances in the power grid, the calculated fusion discrimination value K(t) = 0.82; the third preset threshold K th =0.7; Preset discrimination correction term △ judge =0.08; duration of the islanding effect t last =35ms; the difference between the islanding effect discrimination time and the triggering time Δt judge (t) = 3ms. The fourth preset threshold is 20ms, and the fifth preset threshold is 5ms. The first preset discrimination condition is verified. If the first preset condition is met, it can be determined that the current target power grid's grid connection point exhibits an islanding effect.

[0239] S305. When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the second preset discrimination condition, the target power grid is determined to be operating normally.

[0240] In this step, the second preset discrimination condition includes: the difference between the fusion discrimination value and the third preset threshold and the preset discrimination correction term. This is used to distinguish between three states: normal power grid, islanding effect, and uncertainty. The uncertain state can also be called a suspected islanding effect.

[0241] For example, the third preset threshold K th =0.7; Preset discrimination correction term △ judge=0.08; fusion discrimination value K(t)=0.55; calculation determines 0.55<0.7-0.08; then it is determined that the second preset discrimination condition is met, and the power grid is judged to be operating normally.

[0242] S306. When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the third preset discrimination condition, the target power grid is determined to be in an uncertain state.

[0243] In this step, the third preset discrimination condition includes: the fusion discrimination value is greater than or equal to the difference between the third preset threshold and the preset discrimination correction term, and less than the sum of the third preset threshold and the preset discrimination correction term. This condition is used for intermediate interval judgment, that is, the fusion discrimination value is between the lower limit of normal operation and the upper limit of islanding effect, and it is impossible to directly determine whether it is normal or islanded. At this time, the target power grid has neither clear normal operation characteristics nor islanding effect characteristics that meet all conditions. It is mostly caused by slight disturbances in the power grid and small fluctuations in characteristic parameters. It is an intermediate result of islanding judgment and needs to be further monitored and confirmed.

[0244] For example, the third preset threshold K th =0.7; Preset discrimination correction term △ judge =0.08; fusion discrimination value K(t)=0.68; calculation determines 0.7-0.08<0.68<0.7+0.08; then it is determined that the third preset discrimination condition is met, and the current target power grid is in an uncertain state.

[0245] Optionally, when the target power grid is in an uncertain state, the level of uncertainty can be classified according to the trend of the fusion discrimination value. When the fusion discrimination value is close to the upper limit of the islanding effect, the level of uncertainty of the current target power grid is classified as high; when the fusion discrimination value is close to the lower limit of normal operation, the level of uncertainty is classified as low; when the fusion discrimination value is in the middle of the range between the lower limit of normal operation and the upper limit of the islanding effect, the level of uncertainty is classified as medium. This level can be used to determine the extension of the monitoring time.

[0246] S307. When the islanding determination result indicates that there is a non-islanding disturbance in the target power grid, execute the blocking protection action.

[0247] In this step, the blocking protection action refers to the shielding operation of the protection function for the target power grid, that is, temporarily disabling the triggering mechanism of islanding protection and not sending any protection trip command. This action is a measure to prevent accidental activation during non-islanding disturbances. Because non-islanding disturbances are generally caused by normal changes in the operating conditions of the power grid, it is not necessary to disconnect the photovoltaic power station from the power grid at this time. Executing the blocking protection action at this time can prevent the photovoltaic power station from disconnecting from the grid due to malfunction of the protection device of the target power grid.

[0248] Optionally, the execution duration of the blocking protection action can be determined according to the type of non-islanded disturbance. For example, when the non-islanded disturbance is a voltage sag, the execution duration of the blocking protection action can be set to 1 second; when the non-islanded disturbance is a load change, the execution duration of the blocking protection action can be set to 500 ms.

[0249] S308. When the islanding determination result indicates that the target power grid has an islanding effect, execute the protection action.

[0250] In this step, the protection action is performed by sending trip commands to the photovoltaic power station and the grid connection point switch, and driving the photovoltaic power station and the grid connection point switch to execute the trip commands, thereby disconnecting the photovoltaic power station from the grid and achieving the purpose of eliminating islanding.

[0251] Optionally, based on the islanding determination results, the type of islanding currently existing in the target power grid can be identified, such as transient islanding and steady-state islanding. Based on the islanding type, tiered protection actions can be executed, thereby reducing the impact of photovoltaic power plant disconnection on the power grid.

[0252] For example, when the islanding type is instantaneous islanding, the trip command is directly executed when the protection action is performed; when the islanding type is steady-state islanding, the load needs to be reduced for the target power grid before the trip command is executed.

[0253] S309. When the islanding determination result indicates that the target power grid is in an uncertain state, extend the monitoring time and re-determine the determination.

[0254] In this step, extending the monitoring duration refers to the continuous monitoring time required for delayed re-judgment under uncertain conditions of the target power grid. By accumulating more feature data, sufficient basis is provided for further re-judgment. The extended monitoring duration can be determined according to the level of uncertainty. For example, when the level of uncertainty is low, the extended monitoring duration is set to 50ms; when the level of uncertainty is medium, the extended monitoring duration is set to 100ms; and when the level of uncertainty is high, the extended monitoring duration is set to 200ms.

[0255] Figure 4 This is a schematic diagram of the island detection device provided in this application, as shown below. Figure 4 As shown, the island detection device provided in this embodiment includes:

[0256] The acquisition module 401 is used to acquire electrical signals of the target power grid connection point based on a preset frequency; wherein the electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals and zero-sequence voltage signals.

[0257] The first processing module 402 is used to perform signal processing and multi-timescale feature extraction based on electrical signals to obtain multi-dimensional feature parameters corresponding to each timescale; wherein, the multi-timescale includes millisecond-level timescale and second-level timescale.

[0258] The second processing module 403 is used to perform weighted fusion calculation on the multi-dimensional feature parameters corresponding to each time scale based on the preset timeliness weight corresponding to the time scale, so as to obtain the fusion feature value corresponding to each time scale.

[0259] The third processing module 404 is used to determine the fusion coefficients corresponding to each time scale based on the current operating status of the target power grid, and to calculate the fusion discrimination values ​​corresponding to multiple time scales based on the fusion coefficients and the fusion feature values ​​corresponding to each time scale.

[0260] The fourth processing module 405 is used to determine islands based on the fusion discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, and to obtain the island determination result.

[0261] Optionally, in one possible implementation, the first processing module 402 is further configured to:

[0262] The total harmonic distortion rate at the current grid connection point is generated based on the electrical signal.

[0263] The length of the filtering window is determined based on the total harmonic distortion rate.

[0264] Based on the filter window length, preset interference compensation factor, and preset weight attenuation factor, the electrical signal is subjected to sliding filtering to obtain an electrical signal with suppressed interference.

[0265] Based on preset outlier judgment conditions, abnormal data is detected in the electrical signals that have been suppressed for interference, and abnormal data is removed to obtain the electrical signals with outliers removed.

[0266] Based on a preset trend compensation factor, the electrical signals that have been removed from the anomaly process are linearly interpolated to obtain standardized electrical signals.

[0267] Feature extraction is performed on the standardized electrical signal at each time scale to obtain multiple initial feature parameters corresponding to each time scale.

[0268] For each time scale, multiple initial feature parameters are corrected to obtain multiple standardized feature parameters after correction.

[0269] Based on multiple standardized feature parameters, the discrimination coefficient of each standardized feature parameter is calculated, and standardized feature parameters with discrimination coefficients lower than a first preset threshold are removed to obtain multiple target feature parameters.

[0270] Multiple target feature parameters at each time scale are integrated in a preset order, and each target feature parameter is mapped to a preset interval to obtain multi-dimensional feature parameters corresponding to each time scale.

[0271] Optionally, in one possible implementation, the second processing module 403 is further configured to:

[0272] For each time scale, the discriminative weight of the feature parameter is calculated based on the discriminative coefficient corresponding to each feature parameter.

[0273] The weighted fusion calculation is performed based on the discriminative weight of each feature parameter, the parameter value of each feature parameter in the multi-dimensional feature parameters, and the preset timeliness weight corresponding to the time scale, to obtain the fused feature value corresponding to the time scale.

[0274] Alternatively, in one possible implementation, the third processing module 404 is further configured to:

[0275] Based on the operating status of the target power grid, the fusion coefficient corresponding to each time scale is calculated; where the operating status includes the current voltage change rate, photovoltaic power change rate, and islanding type identification factor of the target power grid.

[0276] Based on the fusion coefficient, the fusion feature values ​​corresponding to each time scale, and the preset cross-time scale correction terms, a weighted fusion calculation is performed to obtain the fusion discrimination value; among which, the preset cross-time scale correction terms include preset cross-time scale compensation terms, preset stability adjustment terms, and preset error correction terms.

[0277] Alternatively, in one possible implementation, the fourth processing module 405 is further configured to:

[0278] The target timescale for similarity matching is determined based on the operating status of the target power grid.

[0279] Similarity matching is performed based on the multi-dimensional feature parameters corresponding to the target time scale and a pre-set non-isolated feature database to obtain similarity matching values.

[0280] When the similarity matching value is greater than or equal to the second preset threshold, it is determined that there is a non-islanding disturbance in the current target power grid.

[0281] When the similarity matching value is less than the second preset threshold, and the fusion discrimination value and the island state data meet the first preset discrimination condition, it is determined that there is an island effect at the grid connection point; wherein, the first preset discrimination condition includes: the fusion discrimination value is greater than or equal to the sum of the third preset threshold and the preset discrimination correction term, the duration of the island effect is greater than the fourth preset threshold, and the difference between the island effect discrimination time and the triggering time is less than the fifth preset threshold.

[0282] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the second preset discrimination condition, the target power grid is determined to be operating normally; wherein, the second preset discrimination condition includes: the fusion discrimination value is less than the difference between the third preset threshold and the preset discrimination correction term.

[0283] When the similarity matching value is less than the second preset threshold and the fusion discrimination value meets the third preset discrimination condition, the target power grid is determined to be in an uncertain state; wherein, the third preset discrimination condition includes: the fusion discrimination value is greater than or equal to the difference between the third preset threshold and the preset discrimination correction term, and is less than the sum of the third preset threshold and the preset discrimination correction term.

[0284] Alternatively, in one possible implementation, the fourth processing module 405 is further configured to:

[0285] When the islanding determination result indicates that there is a non-islanding disturbance in the target power grid, the blocking protection action is executed.

[0286] When the islanding determination result indicates that the target power grid has an islanding effect, a protection action is executed.

[0287] When the islanding determination result indicates that the target power grid is in an uncertain state, the monitoring time is extended and the determination is re-performed.

[0288] The apparatus provided in this embodiment can execute the method provided in the above method embodiment. Its implementation principle and technical effect are similar, and will not be described in detail here.

[0289] Figure 5 A schematic diagram of the structure of the electronic device provided in this application. Figure 5 As shown, the electronic device provided in this embodiment includes at least one processor 501 and a memory 502. Optionally, the device further includes a communication component 503. The processor 501, memory 502, and communication component 503 are connected via a bus 504.

[0290] In the specific implementation process, at least one processor 501 executes computer execution instructions stored in memory 502, causing at least one processor 501 to execute the above-mentioned island detection method or approach.

[0291] The specific implementation process of processor 501 can be found in the above method embodiments, and its implementation principle and technical effect are similar. It will not be repeated here.

[0292] In the above embodiments, it should be understood that the processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), etc. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the method disclosed in this invention can be directly implemented by a hardware processor, or implemented by a combination of hardware and software modules within the processor.

[0293] The memory may include random access memory (RAM) and may also include non-volatile memory (NVM), such as at least one disk storage device.

[0294] The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of illustration, the buses shown in the accompanying drawings are not limited to a single bus or a single type of bus.

[0295] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0296] This application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-described method.

[0297] The aforementioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk. The readable storage medium can be any available medium accessible to a general-purpose or special-purpose computer.

[0298] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can reside in an Application Specific Integrated Circuit (ASIC). Alternatively, the processor and the readable storage medium can exist as discrete components in the device.

[0299] The division of units is merely a logical functional division; in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be indirect coupling or communication connection through some interfaces, devices, or units, and may be electrical, mechanical, or other forms.

[0300] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0301] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0302] If a function is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0303] Those skilled in the art will understand that all or part of the steps of the above-described method embodiments can be implemented by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above-described method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0304] Finally, it should be noted that other embodiments of the invention will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This invention is intended to cover any variations, uses, or adaptations of the invention that follow the general principles of the invention and include common knowledge or customary techniques in the art not disclosed herein, and is not limited to the precise structures described above and shown in the accompanying drawings, and various modifications and changes can be made without departing from its scope. The scope of the invention is limited only by the appended claims.

Claims

1. An island detection method, characterized in that, include: Electrical signals at the target power grid connection point are acquired based on a preset frequency; wherein, the electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals; Signal processing and multi-timescale feature extraction are performed based on the electrical signal to obtain multi-dimensional feature parameters corresponding to each timescale; wherein, the multi-timescale includes millisecond-level timescale and second-level timescale; For each time scale, the multi-dimensional feature parameters corresponding to the time scale are weighted and fused based on the preset timeliness weight corresponding to the time scale to obtain the fused feature value corresponding to each time scale; Based on the current operating status of the target power grid, the fusion coefficients corresponding to each time scale are determined, and based on the fusion coefficients and the fusion feature values ​​corresponding to each time scale, the fusion discrimination values ​​corresponding to multiple time scales are calculated. Based on the fusion discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, island determination is performed to obtain the island determination result.

2. The method according to claim 1, characterized in that, The signal processing and multi-timescale feature extraction based on the electrical signal yields multi-dimensional feature parameters corresponding to each time scale, including: The current total harmonic distortion rate of the grid connection point is generated based on the electrical signal; The length of the filter window is determined based on the total harmonic distortion. Based on the filter window length, the preset interference compensation factor, and the preset weight attenuation factor, the electrical signal is subjected to sliding filtering to obtain an electrical signal with interference suppression. Based on preset outlier determination conditions, abnormal data is detected in the electrical signal that has been suppressed by interference, and abnormal data is removed to obtain the electrical signal with outlier removed. The abnormal electrical signals are linearly interpolated based on a preset trend compensation factor to obtain standardized electrical signals. Feature extraction is performed on the standardized electrical signal at each time scale to obtain multiple initial feature parameters corresponding to each time scale; For each time scale, the plurality of initial feature parameters are corrected to obtain the corrected plurality of standardized feature parameters; Based on multiple standardized feature parameters, the discrimination coefficient of each standardized feature parameter is calculated, and standardized feature parameters with discrimination coefficients lower than a first preset threshold are removed to obtain multiple target feature parameters. Multiple target feature parameters at each time scale are integrated in a preset order, and each target feature parameter is mapped to a preset interval to obtain multi-dimensional feature parameters corresponding to each time scale.

3. The method according to claim 2, characterized in that, For each time scale, a weighted fusion calculation is performed on the multi-dimensional feature parameters corresponding to the time scale based on a preset timeliness weight, to obtain a fused feature value for each time scale, including: For each of the multi-dimensional feature parameters at each time scale, the discriminative weight of the feature parameter is calculated based on the discriminative coefficient corresponding to each feature parameter in the multi-dimensional feature parameters. The fusion feature value corresponding to the time scale is obtained by weighting and fusing the discriminative weight of each feature parameter, the parameter value of each feature parameter in the multi-dimensional feature parameters, and the preset timeliness weight corresponding to the time scale.

4. The method according to claim 3, characterized in that, The process involves determining the fusion coefficients corresponding to each time scale based on the current operating state of the target power grid, and calculating fusion discriminant values ​​corresponding to multiple time scales based on the fusion coefficients and the fusion feature values ​​corresponding to each time scale, including: Based on the operating status of the target power grid, the fusion coefficient corresponding to each time scale is calculated; wherein, the operating status includes the current voltage change rate, photovoltaic power change rate, and islanding type identification factor of the target power grid; Based on the fusion coefficient, the fusion feature values ​​corresponding to each time scale, and the preset cross-time scale correction term are weighted and fused to obtain the fusion discrimination value; wherein, the preset cross-time scale correction term includes a preset cross-time scale compensation term, a preset stability adjustment term, and a preset error correction term.

5. The method according to claim 4, characterized in that, The process of determining islands based on the fused discriminant value, the multi-dimensional feature parameters corresponding to each time scale, and a preset non-island feature database, to obtain island determination results, includes: The target time scale for similarity matching is determined based on the operating status of the target power grid; Similarity matching is performed based on the multi-dimensional feature parameters corresponding to the target time scale and the preset non-isolated feature database to obtain a similarity matching value; When the similarity matching value is greater than or equal to the second preset threshold, it is determined that the current target power grid has a non-islanding disturbance; When the similarity matching value is less than the second preset threshold, and the fusion discrimination value and the island state data meet the first preset discrimination condition, it is determined that the grid connection point has an island effect; wherein, the first preset discrimination condition includes: the fusion discrimination value is greater than or equal to the sum of the third preset threshold and the preset discrimination correction term, the duration of the island effect is greater than the fourth preset threshold, and the difference between the island effect discrimination time and the triggering time is less than the fifth preset threshold; When the similarity matching value is less than a second preset threshold and the fusion discrimination value meets a second preset discrimination condition, the target power grid is determined to be operating normally; wherein, the second preset discrimination condition includes: the fusion discrimination value is less than the difference between the third preset threshold and the preset discrimination correction term; When the similarity matching value is less than a second preset threshold and the fusion discrimination value meets a third preset discrimination condition, the target power grid is determined to be in an uncertain state; wherein, the third preset discrimination condition includes: the fusion discrimination value is greater than or equal to the difference between the third preset threshold and the preset discrimination correction term, and is less than the sum of the third preset threshold and the preset discrimination correction term.

6. The method according to claim 5, characterized in that, After obtaining the island determination result, the method further includes: When the islanding determination result indicates that the target power grid has a non-islanding disturbance, a blocking protection action is performed; When the islanding determination result indicates that the target power grid has an islanding effect, a protection action is performed; When the islanding determination result indicates that the target power grid is in an uncertain state, the monitoring time is extended and the determination is performed again.

7. An island detection device, characterized in that, include: The acquisition module is used to acquire electrical signals at the target power grid connection point based on a preset frequency; wherein, the electrical signals include three-phase current signals, three-phase voltage signals, zero-sequence current signals, and zero-sequence voltage signals; The first processing module is used to perform signal processing and multi-timescale feature extraction based on the electrical signal to obtain multi-dimensional feature parameters corresponding to each timescale; wherein, the multi-timescale includes millisecond-level timescales and second-level timescales; The second processing module is used to perform weighted fusion calculation on the multi-dimensional feature parameters corresponding to each time scale based on the preset timeliness weight corresponding to the time scale, so as to obtain the fusion feature value corresponding to each time scale. The third processing module is used to determine the fusion coefficient corresponding to each time scale based on the current operating state of the target power grid, and to calculate the fusion discrimination value corresponding to multiple time scales based on the fusion coefficient and the fusion feature value corresponding to each time scale. The fourth processing module is used to determine islands based on the fusion discrimination value, the multi-dimensional feature parameters corresponding to each time scale, and the preset non-island feature database, and to obtain the island determination result.

8. An electronic device, characterized in that, include: Memory, processor; The memory stores computer-executed instructions; The processor executes computer execution instructions stored in the memory, causing the processor to perform the method as described in any one of claims 1-6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, which, when executed by a processor, are used to implement the method as described in any one of claims 1-6.

10. A computer program product, characterized in that, Includes a computer program that, when executed by a processor, implements the method described in any one of claims 1-6.

Images