Distributed access unit with power quality anomaly monitoring and grid disconnection functions

Abstract

This invention relates to the field of power quality anomaly monitoring technology, and particularly to a distributed access unit with power quality anomaly monitoring and grid disconnection functions. The unit includes a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps: obtaining the optimal short window for voltage-frequency and the optimal long window for voltage-harmonics at the current moment; using the optimal short window and the optimal long window, obtaining the normalized mutual information of voltage-frequency and the normalized mutual information of voltage-harmonic distortion rate, respectively; and using the ratio to construct a power quality anomaly coefficient. This enables accurate differentiation between two types of anomalies: grid transient disturbances and photovoltaic (PV) self-degradation. This ensures accurate disconnection during grid disturbances, guaranteeing power supply continuity; and accurate grid disconnection during PV self-degradation, preventing fault escalation.

Description

Technical Field

[0001] This invention relates to the field of power quality anomaly monitoring technology, and in particular to a distributed access unit with power quality anomaly monitoring and grid disconnection functions. Background Technology

[0002] In scenarios where distributed access units monitor power quality at grid-connected points, islanding monitoring methods based on independent threshold combinations of voltage and frequency are commonly used to achieve grid-connected protection. Specifically, the distributed access unit independently collects and monitors the effective voltage and frequency at the grid-connected point. When either the effective voltage or frequency deviates from the rated range for a certain period, or when the number of fluctuations exceeds a preset threshold, it is determined to be in an islanding state. A tripping command is then sent to the dedicated photovoltaic circuit breaker to disconnect the photovoltaic system from the grid. The core advantage of this islanding monitoring method lies in its simple logic and low computational load. It eliminates the need for complex analysis of the relationships between electrical quantities, meeting basic anti-islanding and overvoltage protection requirements in areas with minimal grid disturbances and low photovoltaic penetration. Furthermore, it is easily implemented in resource-constrained embedded access units.

[0003] While islanding monitoring methods based on independent threshold combinations of voltage and frequency can achieve anti-islanding protection and overvoltage interruption with simple logic in areas with few grid disturbances and low photovoltaic penetration, in scenarios with high proportion of distributed photovoltaic access and frequent grid transient disturbances, this islanding monitoring method relies solely on independent over-limit judgments of voltage and frequency. It cannot distinguish whether the voltage over-limit at the grid connection point is caused by transient disturbances such as the start-up and shutdown of high-power equipment on the grid side, or by continuous degradation such as abnormal control of the photovoltaic inverter itself or hardware aging. This leads to frequent false interruptions of grid transient disturbances that do not require interruption, and the lack of perception of coupling characteristics such as harmonics results in the failure to detect the true degradation on the photovoltaic side.

[0004] Therefore, how to accurately distinguish between the abnormal root causes of grid transient disturbances and photovoltaic degradation itself, so as to reduce the misjudgment rate of grid disconnection based on power anomalies, has become an urgent problem to be solved. Summary of the Invention

[0005] In view of this, embodiments of the present invention provide a distributed access unit with power quality anomaly monitoring and grid disconnection functions to solve the problem of how to accurately distinguish between the abnormal root causes of grid transient disturbances and photovoltaic self-degradation, so as to reduce the misjudgment rate of grid disconnection based on power anomalies.

[0006] This invention provides a distributed access unit with power quality anomaly monitoring and grid disconnection functions, including a memory, a processor, and a computer program stored in the memory and running on the processor. When the processor executes the computer program, it performs the following steps:

[0007] Obtain the voltage sequence, frequency sequence, and harmonic distortion rate sequence within a preset time period up to the current moment;

[0008] In the process of calculating the mutual information between voltage and frequency sequences using a short window, the candidate length range of the short window is obtained. The peak search of the mutual information between the voltage and frequency sequences is performed using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length.

[0009] In the process of calculating the mutual information between voltage sequence and harmonic distortion rate sequence using a long window, the candidate length range of the long window is obtained. The mutual information between voltage sequence and harmonic distortion rate sequence is searched for stable changes using each candidate length within the candidate length range of the long window, and the optimal long window corresponding to the optimal candidate length is obtained.

[0010] The optimal short window is used to obtain voltage-frequency normalized mutual information between the voltage sequence and the frequency sequence, and the optimal long window is used to obtain voltage-harmonic distortion rate normalized mutual information between the voltage sequence and the harmonic distortion rate sequence. The power quality anomaly coefficient at the current moment is obtained based on the ratio between the voltage-frequency normalized mutual information and the voltage-harmonic distortion rate normalized mutual information.

[0011] Power quality anomaly coefficient at the current moment is used to realize power quality anomaly monitoring and grid disconnection.

[0012] Preferably, the step of performing peak search on the mutual information between the voltage sequence and the frequency sequence using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length includes:

[0013] For any candidate length within the candidate length range of the short window, the short window of any candidate length is used to slide in the voltage sequence and frequency sequence with a preset sliding step size to obtain multiple short window data. The voltage-frequency mutual information under each short window data is obtained, and the mean value of the voltage-frequency mutual information is obtained and recorded as the preferred value of the candidate length.

[0014] Based on the optimal value of each candidate length within the candidate length range of the short window, the candidate length corresponding to the maximum optimal value is selected as the optimal candidate length, and the length of the short window is set as the optimal candidate length to obtain the optimal short window.

[0015] Preferably, the step of using each candidate length within the candidate length range of the long window to perform a stable change search on the mutual information between the voltage sequence and the harmonic distortion rate sequence to obtain the optimal long window corresponding to the optimal candidate length includes:

[0016] For any candidate length within the candidate length range of the long window, the long window of any candidate length is used to slide in the voltage sequence and harmonic distortion rate sequence with a preset sliding step size to obtain multiple long window data. The voltage-frequency mutual information under each long window data is obtained respectively, and the coefficient of variation of the voltage-frequency mutual information is calculated and recorded as the preferred value of any candidate length.

[0017] Based on the optimal value of each candidate length within the candidate length range of the long window, the candidate length corresponding to the minimum optimal value is selected as the optimal candidate length, and the length of the long window is set as the optimal candidate length to obtain the optimal long window.

[0018] Preferably, the step of obtaining voltage-frequency normalized mutual information between the voltage sequence and the frequency sequence using the optimal short window includes:

[0019] Taking the current time as the last time in the optimal short window, the voltage subsequence and frequency subsequence are obtained in the voltage sequence and frequency sequence respectively using the optimal short window. The mutual information between the voltage subsequence and the frequency subsequence is calculated. The mutual information is geometrically normalized using the information entropy of the voltage subsequence and the information entropy of the frequency subsequence to obtain the voltage-frequency normalized mutual information.

[0020] Preferably, the step of obtaining voltage-harmonic distortion rate normalized mutual information between the voltage sequence and the harmonic distortion rate sequence using the optimal long window includes:

[0021] Taking the current time as the last time in the optimal long window, the voltage subsequence and harmonic distortion rate subsequence are obtained from the voltage sequence and harmonic distortion rate sequence respectively using the optimal long window. The mutual information between the voltage subsequence and the harmonic distortion rate subsequence is calculated. The mutual information is geometrically normalized using the information entropy of the voltage subsequence and the information entropy of the harmonic distortion rate subsequence to obtain the voltage-harmonic distortion rate normalized mutual information.

[0022] Preferably, obtaining the power quality anomaly coefficient at the current moment based on the ratio between voltage-frequency normalized mutual information and voltage-harmonic distortion rate normalized mutual information includes:

[0023] ;

[0024] in, This represents the power quality anomaly coefficient at the current moment. This represents voltage-frequency normalized mutual information. This represents the normalized mutual information of voltage-harmonic distortion rate. To represent extremely small real numbers, prevent the denominator from being 0.

[0025] Preferably, the step of using the power quality anomaly coefficient at the current moment to realize power quality anomaly monitoring and grid disconnection includes:

[0026] A preset fuzzy range is obtained. If the power quality anomaly coefficient at the current moment is greater than or equal to the upper limit of the fuzzy range, it is determined that the current limit violation is caused by a transient disturbance in the power grid, and the distributed access unit does not perform a grid disconnection operation. If the power quality anomaly coefficient at the current moment is less than the lower limit of the fuzzy range, it is determined that the current limit violation is caused by the degradation of the photovoltaic system itself, and the distributed access unit performs a grid disconnection operation. If the power quality anomaly coefficient at the current moment is within the fuzzy range, the distributed access unit first performs power reduction regulation, and if it remains within the fuzzy range, it forcibly performs a grid disconnection operation.

[0027] The beneficial effects of the embodiments of the present invention compared with the prior art are as follows:

[0028] In this invention, differentiated adaptive windows are designed, namely, an optimal short window and an optimal long window. The peak value of the optimal short window is used to track and match the instantaneous impact characteristics of grid transient disturbances, while the optimal long window is used to stably select and match the gradual characteristics of photovoltaic degradation, significantly improving the accuracy of anomaly detection. Simultaneously, based on the optimal short window and the optimal long window, normalized mutual information is obtained to quantify the dual-mode coupling strength of voltage-frequency and voltage-harmonics, replacing instantaneous value direction judgment. The ratio is used to construct a power quality anomaly coefficient, which can automatically offset the scale deviation caused by differences in operating conditions without the need to adapt to multiple thresholds. Finally, the power quality anomaly coefficient is used to realize power quality anomaly monitoring and grid disconnection, achieving accurate differentiation between the two types of anomaly root causes: grid transient disturbances and photovoltaic degradation. Ultimately, this achieves accurate disconnection during grid disturbances, ensuring power supply continuity, and accurate grid disconnection during photovoltaic degradation, preventing fault escalation. Attached Figure Description

[0029] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0030] Figure 1This is a flowchart of a method for power quality anomaly monitoring and grid disconnection for a distributed access unit, provided in Embodiment 1 of the present invention. Detailed Implementation

[0031] Embodiments of this disclosure are described in detail below, with examples of these embodiments illustrated in the accompanying drawings. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this disclosure, and should not be construed as limiting it.

[0032] It should be noted that the terms "first," "second," etc., used in this disclosure and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0033] To illustrate the technical solution of the present invention, specific embodiments are described below.

[0034] This invention provides a distributed access unit with power quality anomaly monitoring and grid disconnection functions, including a processor and a memory. The processor executes a computer program stored in the memory to implement a power quality anomaly monitoring and grid disconnection method for the distributed access unit, such as... Figure 1 As shown, the method includes the following steps:

[0035] Step S101: Obtain the voltage sequence, frequency sequence, and harmonic distortion rate sequence within a preset time period up to the current moment.

[0036] To address the inability of existing single-dimensional independent threshold monitoring methods to distinguish the root causes of voltage exceedances at grid-connected points, this invention proposes an anomaly root cause differentiation method based on normalized mutual information dual-modal comparison using long and short windows. First, during real-time monitoring of the electrical parameters of grid-connected points by the distributed access unit, three electrical quantities—voltage, frequency, and harmonic distortion rate—are sampled at a high sampling rate (e.g., 50Hz) to obtain corresponding time data sequences. After denoising, outlier removal, and normalization, two sets of time-aligned sequence pairs are formed, providing high-quality input data for subsequent dual-modal coupled comparative analysis. The specific acquisition method is as follows:

[0037] Using the current moment as the anomaly monitoring moment, the initial voltage sequence, initial frequency sequence, and initial harmonic distortion rate sequence within 30 minutes up to the current moment are obtained, ensuring that the sampling times of the three electrical quantities are aligned. Data preprocessing is performed on the initial voltage sequence, initial frequency sequence, and initial harmonic distortion rate sequence respectively: each sequence is sequentially subjected to DC component removal, low-pass filtering for noise reduction, and outlier removal to eliminate sampling noise and occasional glitches from interfering with subsequent mutual information calculations. Then, the maximum-minimum normalization method is used to normalize each sequence to eliminate dimensional influences, thus obtaining the voltage sequence, frequency sequence, and harmonic distortion rate sequence at the current moment.

[0038] After obtaining the voltage sequence, frequency sequence, and harmonic distortion rate sequence at the current moment, the voltage sequence and frequency sequence are combined into one sequence pair, and the voltage sequence and harmonic distortion rate sequence are combined into another sequence pair. These two sequence pairs are used to characterize two modes of electrical energy change, so as to realize dual-mode comparison later. It should be noted that the two sequence pairs share the same voltage sequence to ensure that the two subsequent coupling strengths are based on the same voltage reference.

[0039] Step S102: In the process of calculating the mutual information between the voltage sequence and the frequency sequence using a short window, the candidate length range of the short window is obtained. The peak search of the mutual information between the voltage sequence and the frequency sequence is performed using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length.

[0040] When voltage exceeds limits, the essential difference in electrical quantity coupling modes between grid transient disturbances and photovoltaic (PV) self-degradation is as follows: grid transient disturbances manifest as synchronous abrupt changes in voltage and frequency but unchanged harmonics, while PV self-degradation manifests as synchronous distortion of voltage and harmonics but stable frequency. However, directly judging the direction of rise and fall of instantaneous electrical quantities can lead to: blurred boundaries, making it difficult to set a unified judgment threshold for weak anomalies; frequent time lags in response, with frequency changes potentially lagging behind voltage by several cycles, making it easy to mismatch the instantaneous value direction within the time lag window; and sensor noise superimposed on frequency and harmonics, producing spurious fluctuations and causing misjudgments.

[0041] Therefore, in this embodiment of the invention, the normalized mutual information between the voltage sequence and the frequency sequence, as well as the normalized mutual information between the voltage sequence and the harmonic distortion rate sequence, is used to quantify the coupling strength in two dimensions, which is then used to achieve anomaly monitoring by constructing a dual-modal comparison. Mutual information (MI) measures the amount of information shared between two random variables (sequences), and its value range is MI≥0. A larger value indicates a stronger correlation, while MI=0 indicates that the two sequences are independent.

[0042] However, the dual-mode comparison faces an inherent contradiction: the strong voltage-frequency coupling of grid transient disturbances is instantaneous and requires a short window to capture the coupling peak; otherwise, the features will be diluted by including a large amount of weak coupling data before and after the disturbance. The voltage-harmonic coupling of photovoltaic degradation is continuous and gradual, requiring a longer window to ensure the statistical stability of the estimation; otherwise, insufficient samples will cause drastic fluctuations in the estimated value. Therefore, if a fixed window of uniform length is used to calculate the mutual information of the two sequence pairs separately, since the strong voltage-frequency coupling of grid transient disturbances is instantaneous and peaks only at the moment of the disturbance, if the window is too long, a large amount of weak coupling data before and after the disturbance will dilute the mutual information estimation, resulting in a low normalized mutual information between voltage and frequency. Conversely, the voltage-harmonic coupling of photovoltaic degradation is continuous and gradual, requiring sufficient samples to ensure the statistical stability of the mutual information estimation. If the window is too short, the estimated value will fluctuate drastically due to insufficient samples, leading to unreliable normalized mutual information between voltage and harmonic distortion rate. This inevitably causes the coupling strength estimate on at least one side to deviate from its true value.

[0043] Therefore, in this embodiment of the invention, a differentiated adaptive window optimization strategy based on peak tracking and stability selection is further introduced. For the voltage-frequency coupling side, the peak tracking strategy is used to search for the optimal window that maximizes the normalized mutual information within the short window candidate range, and the strongest coupling moment is selected. For the voltage-harmonic coupling side, the stability selection strategy is used to search for the optimal window that minimizes the estimated coefficient of variation within the long window candidate range, and the most stable coupling state is selected.

[0044] In the process of calculating the mutual information between the voltage and frequency sequences using a short window, considering that the strong voltage-frequency coupling of the power grid transient disturbance is pulse-like, the maximum value window is needed to capture this instantaneous feature. Therefore, within the candidate length range of the short window, each candidate length is traversed using a maximum value tracking strategy to search for the short window that maximizes the mutual information as the optimal short window.

[0045] Specifically, firstly, the candidate length range of the short window is obtained. In industry standards, the typical duration of a voltage spurt is 0.5 cycles to several seconds, so the lower limit of the candidate length range of the short window can be set to 1. According to the IEC 61000-4-7 standard, harmonic analysis of a 50Hz system recommends using 10 fundamental frequency cycles as the sampling window length, so the upper limit of the candidate length range of the short window can be set to 10, corresponding to a candidate length range of [1, 10]. Then, peak search is performed on the mutual information between the voltage sequence and the frequency sequence using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length.

[0046] For any candidate length within the candidate length range of the short window, the short window of any candidate length is slid across the voltage and frequency sequences with a preset sliding step size to obtain multiple short window data. The voltage-frequency mutual information under each short window data is obtained, and the mean value of the voltage-frequency mutual information is recorded as the optimal value of the candidate length. Similarly, the optimal value of each candidate length within the candidate length range of the short window is obtained. Then, based on the optimal value of each candidate length within the candidate length range of the short window, the candidate length corresponding to the maximum optimal value is selected as the optimal candidate length, and the length of the short window is set as the optimal candidate length to obtain the optimal short window, which is used to obtain mutual information between the voltage and frequency sequences.

[0047] For example: Using a short window of candidate length 1, the voltage sequence and frequency sequence are divided into multiple short window data with a sliding step size of 2. Each short window data includes a subsequence of the voltage sequence and a subsequence of the frequency sequence. The mutual information between voltage and frequency in each short window data is calculated and denoted as voltage-frequency mutual information. One short window data corresponds to one voltage-frequency mutual information. The mean of the voltage-frequency mutual information of all short window data is denoted as the optimal value of candidate length 1. Similarly, the optimal value of each candidate length within the candidate length range [1, 10] of the short window is calculated. The candidate length corresponding to the maximum optimal value is taken as the optimal length of the short window, thus obtaining the optimal short window.

[0048] Step S103: In the process of calculating the mutual information between the voltage sequence and the harmonic distortion rate sequence using a long window, the candidate length range of the long window is obtained. The mutual information between the voltage sequence and the harmonic distortion rate sequence is searched for stable changes using each candidate length within the candidate length range of the long window, and the optimal long window corresponding to the optimal candidate length is obtained.

[0049] Based on the above method for obtaining the optimal short window on the voltage-frequency coupling side, for the voltage-harmonic coupling side, in the process of calculating the mutual information between the voltage sequence and the harmonic distortion rate sequence using a long window, considering that the voltage-harmonic coupling of photovoltaic degradation is a continuous characteristic, taking the most stable window can ensure the reliability of the coupling estimation and avoid misjudging random fluctuations as real coupling due to improper window selection. Therefore, within the candidate length range of the long window, a stable selection strategy is used to traverse each candidate length and search for the optimal long window that minimizes the coefficient of variation of the estimated mutual information, so as to obtain its most stable coupling state.

[0050] Specifically, firstly, the candidate length range of the long window is obtained. According to industry standards, harmonic analysis recommends using 10 cycles, so the lower limit of the candidate length range can be set to 10, with no upper limit. This can be adjusted according to actual conditions. A smaller upper limit results in a more sensitive response; this can be adjusted when the power grid environment is complex and high response speed is required. Conversely, a larger upper limit results in a more stable response; this can be adjusted when the power grid environment is stable and high estimation stability is required. Preferably, in this embodiment, the upper limit is set to 50, corresponding to a candidate length range of [10, 50]. Then, using each candidate length within the long window's candidate length range, a stable change search is performed on the mutual information between the voltage sequence and the harmonic distortion rate sequence to obtain the optimal long window corresponding to the optimal candidate length.

[0051] For any candidate length within the candidate length range of the long window, the long window of any candidate length is slid across the voltage sequence and harmonic distortion rate sequence with a preset sliding step size to obtain multiple long window data. The voltage-frequency mutual information under each long window data is obtained, and the coefficient of variation of the voltage-frequency mutual information is calculated and recorded as the preferred value of the candidate length. Based on the preferred value of each candidate length within the candidate length range of the long window, the candidate length corresponding to the minimum preferred value is selected as the optimal candidate length, and the length of the long window is set as the optimal candidate length to obtain the optimal long window.

[0052] For example, using a long window with a candidate length of 10, the voltage sequence and harmonic distortion rate sequence are divided into multiple long window data points with a sliding step size of 2. Each long window data point includes a subsequence from the voltage sequence and a subsequence from the harmonic distortion rate sequence. The mutual information between the voltage and harmonic distortion rates in each long window data point is calculated and denoted as the voltage-harmonic distortion rate mutual information. One long window data point corresponds to one voltage-harmonic distortion rate mutual information point. The coefficient of variation of the voltage-harmonic distortion rate mutual information of all long window data points is calculated and denoted as the optimal value of the candidate length of 10. Similarly, the optimal value of each candidate length within the candidate length range [10, 50] of the long window is calculated. The candidate length corresponding to the minimum optimal value is taken as the optimal length of the long window, thus obtaining the optimal long window. The calculation of the coefficient of variation is a prior art technique.

[0053] It should be noted that the voltage-harmonic coupling of photovoltaics is a continuous and stable phenomenon. If the length of the long window is not properly selected or there are random fluctuations in the data, the coefficient of variation will increase. Selecting the window length based on minimizing the coefficient of variation can ensure that the selected coupling characteristics are stable and non-random, rather than false high values ​​caused by noise or transient interference at a certain location.

[0054] Step S104: Obtain voltage-frequency normalized mutual information between the voltage sequence and the frequency sequence using the optimal short window, and obtain voltage-harmonic distortion rate normalized mutual information between the voltage sequence and the harmonic distortion rate sequence using the optimal long window. Based on the ratio between the voltage-frequency normalized mutual information and the voltage-harmonic distortion rate normalized mutual information, obtain the power quality anomaly coefficient at the current moment.

[0055] The above-mentioned approach uses a differentiated strategy on both sides to match the physical timescale characteristics of two types of anomalies, and performs window adaptation that conforms to physical characteristics for the two coupling relationships. This results in the optimal short window for mutual information analysis of voltage and frequency sequences, and the optimal long window for mutual information analysis of voltage and harmonic distortion rate sequences. Furthermore, by matching a suitable window, better mutual information can be obtained, which improves the discriminative power of dual-mode comparison compared to the fixed window method, and can better realize anomaly monitoring and grid disconnection.

[0056] Specifically: Mutual information is used to measure the amount of information shared between two random variables (sequences). If X and Y are two discrete random variables, their joint probability density is... edge density is , mutual information The calculation formula is:

[0057]

[0058] Then, the geometric mean normalization method is used to normalize the mutual information, resulting in normalized mutual information. The calculation formula is:

[0059]

[0060] in, The information entropy of sequence X, The information entropy of sequence Y is represented.

[0061] Since the aforementioned mutual information and normalized mutual information are existing technologies, based on the calculation formulas for the aforementioned mutual information and normalized mutual information, the voltage-frequency normalized mutual information is obtained between the voltage sequence and the frequency sequence using an optimal short window: taking the current time as the last time in the optimal short window, the voltage subsequence and frequency subsequence are obtained in the voltage sequence and frequency sequence respectively using the optimal short window, and the mutual information between the voltage subsequence and the frequency subsequence is calculated. Using the information entropy of the voltage subsequence and the information entropy of the frequency subsequence, the mutual information is geometrically normalized to obtain the voltage-frequency normalized mutual information. .

[0062] Similarly, the normalized mutual information of voltage-harmonic distortion rate is obtained between the voltage sequence and the harmonic distortion rate sequence using an optimal long window: taking the current time as the last time in the optimal long window, the voltage subsequence and harmonic distortion rate subsequence are obtained in the voltage sequence and harmonic distortion rate sequence respectively using the optimal long window, and the mutual information between the voltage subsequence and the harmonic distortion rate subsequence is calculated. Using the information entropy of the voltage subsequence and the information entropy of the harmonic distortion rate subsequence, the mutual information is geometrically normalized to obtain the voltage-harmonic distortion rate normalized mutual information. This quantifies the degree of synchronization between harmonics and voltage changes.

[0063] It's important to note that mutual information captures how similar two sequences are, not how synchronous they become. Even if the frequency change lags the voltage by several cycles, as long as the waveforms are statistically dependent, mutual information can capture it, thus solving the problem of mismatches in instantaneous value direction judgment within the time-delay window. Normalized mutual information calculates the statistical similarity between two sequences based on the overall probability distribution of the waveforms, without relying on hard thresholds. It is tolerant of time delays and robust to noise, and can robustly capture the strength of coupling relationships at the waveform morphology level. Therefore, the necessity of normalizing mutual information is crucial: the information entropy of the three electrical quantities—voltage, frequency, and harmonics—is inherently different, leading to systematic deviations in the absolute values ​​of the two sets of mutual information. Normalization eliminates these differences in information entropy. Furthermore, compared to arithmetic mean normalization, geometric mean normalization is more sensitive to extreme differences in information entropy and better suited to the current scenario.

[0064] Furthermore, since the coupling modes of the two types of anomalies exhibit symmetrical differences—one being stronger than the other, and vice versa—the absolute value of the coupling strength in a single dimension is easily affected by operating conditions. However, the ratio of the two can amplify the difference and suppress common-mode fluctuations. Therefore, based on the ratio between the voltage-frequency normalized mutual information and the voltage-harmonic distortion rate normalized mutual information, the power quality anomaly coefficient at the current moment is obtained and used to determine the current anomaly situation.

[0065]

[0066] in, This represents the power quality anomaly coefficient at the current moment. This represents voltage-frequency normalized mutual information. This represents the normalized mutual information of voltage-harmonic distortion rate. To represent extremely small real numbers, prevent the denominator from being 0.

[0067] It should be noted that voltage-frequency normalized mutual information The degree of frequency synchronization during voltage changes was quantified, and the voltage-harmonic distortion rate normalized mutual information was calculated. The degree of harmonic synchronization during voltage changes was quantified because the coupling modes corresponding to photovoltaic (PV) degradation and grid transient disturbances exhibit a symmetry where one is strong and the other weak. Grid transient disturbances lead to a larger numerator and smaller denominator in the formula for calculating the power quality anomaly coefficient R, resulting in a larger R; conversely, PV degradation leads to a smaller numerator and larger denominator in the formula for calculating the power quality anomaly coefficient R, resulting in a smaller R. Therefore, the power quality anomaly coefficient R can measure whether the root cause of the anomaly at the current moment is PV degradation or grid transient disturbance.

[0068] Step S105: Power quality anomaly monitoring and grid disconnection are achieved using the power quality anomaly coefficient at the current moment.

[0069] In this embodiment of the invention, the abnormality root cause is determined by comparing two normalized mutual trust values, rather than by setting thresholds for each of the two normalized mutual trust values ​​separately. The advantage is that, in actual operating conditions, influenced by factors such as the short-circuit capacity at the grid connection point and the inverter model, the absolute values ​​of the normalized mutual trust values ​​for the same type of abnormality may be generally higher or lower in different transformer areas. Fixed thresholds are difficult to apply universally. However, by comparing the values, the two coupling strengths are compressed into a relative quantity, and the scale changes caused by the shared voltage entropy in the numerator and denominator are automatically offset, providing inherent robustness to differences in operating conditions. Therefore, using the current power quality anomaly coefficient to achieve power quality anomaly monitoring and grid disconnection can ensure accurate differentiation between two types of anomaly root causes: grid transient disturbances and photovoltaic (PV) self-degradation. Ultimately, this achieves differentiated grid disconnection control that prevents false triggering of grid transient disturbances and avoids missed triggering of PV self-degradation.

[0070] The general process of using the power quality anomaly coefficient at the current moment to achieve power quality anomaly monitoring and grid disconnection is as follows:

[0071] When the voltage-frequency coupling strength is significantly greater than the voltage-harmonic coupling strength, R is significantly greater than 1, and vice versa. Therefore, the theoretical ideal threshold value is 1. However, due to the influence of measurement noise and operating condition fluctuations, a fuzzy interval [0.5, 2) needs to be set to avoid critical jitter. When R is greater than or equal to 2, it is determined that the current limit violation is caused by a transient disturbance in the power grid. The distributed access unit does not perform grid disconnection operation, but only records the event timestamp, abnormal coefficient value, and disturbance characteristic summary for subsequent statistical analysis. When R is less than 0.5, it is determined that the current limit violation is caused by the degradation of the photovoltaic system itself. The distributed access unit immediately... The distributed access unit sends a tripping command to the photovoltaic circuit breaker via the RS-485 interface to perform grid disconnection. Simultaneously, it records the timestamp of the disconnection event, the anomaly coefficient value, and a snapshot of relevant electrical parameters, and updates the device status identifier through the local maintenance interface. When R is within [0.5, 2), the distributed access unit enters a conservative strategy—prioritizing power reduction control. It issues a power limiting command to the photovoltaic inverter via the RS-485 interface and continuously monitors the anomaly coefficient change trend of subsequent cycles. If R does not clearly deviate to one side within a preset extension time (e.g., 2 seconds), grid disconnection is performed according to the safety priority principle, triggering an alarm and reporting to the master station. Subsequently, the distributed access unit records metadata information for this monitoring and control through the local maintenance interface, including the event timestamp, R value, optimal short window and optimal long window, control decision type, and execution result. Once the communication channel is connected to the master station, the distributed access unit actively reports the judgment result and control record of this event. After confirmation by the master station, it can receive remote parameter adjustment or control commands.

[0072] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A distributed access unit with power quality anomaly monitoring and grid disconnection functions, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, characterized in that, When the processor executes the computer program, it performs the following steps: Obtain the voltage sequence, frequency sequence, and harmonic distortion rate sequence within a preset time period up to the current moment; In the process of calculating the mutual information between voltage and frequency sequences using a short window, the candidate length range of the short window is obtained. The peak search of the mutual information between the voltage and frequency sequences is performed using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length. In the process of calculating the mutual information between voltage sequence and harmonic distortion rate sequence using a long window, the candidate length range of the long window is obtained. The mutual information between voltage sequence and harmonic distortion rate sequence is searched for stable changes using each candidate length within the candidate length range of the long window, and the optimal long window corresponding to the optimal candidate length is obtained. The optimal short window is used to obtain voltage-frequency normalized mutual information between the voltage sequence and the frequency sequence, and the optimal long window is used to obtain voltage-harmonic distortion rate normalized mutual information between the voltage sequence and the harmonic distortion rate sequence. The power quality anomaly coefficient at the current moment is obtained based on the ratio between the voltage-frequency normalized mutual information and the voltage-harmonic distortion rate normalized mutual information. Power quality anomaly coefficient at the current moment is used to realize power quality anomaly monitoring and grid disconnection.

2. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The step of performing peak search on the mutual information between the voltage sequence and the frequency sequence using each candidate length within the candidate length range of the short window to obtain the optimal short window corresponding to the optimal candidate length includes: For any candidate length within the candidate length range of the short window, the short window of any candidate length is used to slide in the voltage sequence and frequency sequence with a preset sliding step size to obtain multiple short window data. The voltage-frequency mutual information under each short window data is obtained, and the mean value of the voltage-frequency mutual information is obtained and recorded as the preferred value of the candidate length. Based on the optimal value of each candidate length within the candidate length range of the short window, the candidate length corresponding to the maximum optimal value is selected as the optimal candidate length, and the length of the short window is set as the optimal candidate length to obtain the optimal short window.

3. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The step of using each candidate length within the candidate length range of the long window to perform a stable change search on the mutual information between the voltage sequence and the harmonic distortion rate sequence, and obtaining the optimal long window corresponding to the optimal candidate length, includes: For any candidate length within the candidate length range of the long window, the long window of any candidate length is used to slide in the voltage sequence and harmonic distortion rate sequence with a preset sliding step size to obtain multiple long window data. The voltage-frequency mutual information under each long window data is obtained respectively, and the coefficient of variation of the voltage-frequency mutual information is calculated and recorded as the preferred value of any candidate length. Based on the optimal value of each candidate length within the candidate length range of the long window, the candidate length corresponding to the minimum optimal value is selected as the optimal candidate length, and the length of the long window is set as the optimal candidate length to obtain the optimal long window.

4. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The step of obtaining voltage-frequency normalized mutual information between the voltage sequence and the frequency sequence using the optimal short window includes: Taking the current time as the last time in the optimal short window, the voltage subsequence and frequency subsequence are obtained in the voltage sequence and frequency sequence respectively using the optimal short window. The mutual information between the voltage subsequence and the frequency subsequence is calculated. The mutual information is geometrically normalized using the information entropy of the voltage subsequence and the information entropy of the frequency subsequence to obtain the voltage-frequency normalized mutual information.

5. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The step of obtaining voltage-harmonic distortion rate normalized mutual information between the voltage sequence and the harmonic distortion rate sequence using the optimal long window includes: Taking the current time as the last time in the optimal long window, the voltage subsequence and harmonic distortion rate subsequence are obtained from the voltage sequence and harmonic distortion rate sequence respectively using the optimal long window. The mutual information between the voltage subsequence and the harmonic distortion rate subsequence is calculated. The mutual information is geometrically normalized using the information entropy of the voltage subsequence and the information entropy of the harmonic distortion rate subsequence to obtain the voltage-harmonic distortion rate normalized mutual information.

6. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The method of obtaining the power quality anomaly coefficient at the current moment based on the ratio between voltage-frequency normalized mutual information and voltage-harmonic distortion rate normalized mutual information includes: ； in, This represents the power quality anomaly coefficient at the current moment. This represents voltage-frequency normalized mutual information. This represents the normalized mutual information of voltage-harmonic distortion rate. To represent extremely small real numbers, prevent the denominator from being 0.

7. The distributed access unit with power quality anomaly monitoring and grid disconnection functions according to claim 1, characterized in that, The method of using the power quality anomaly coefficient at the current moment to realize power quality anomaly monitoring and grid disconnection includes: A preset fuzzy range is obtained. If the power quality anomaly coefficient at the current moment is greater than or equal to the upper limit of the fuzzy range, it is determined that the current limit violation is caused by a transient disturbance in the power grid, and the distributed access unit does not perform a grid disconnection operation. If the power quality anomaly coefficient at the current moment is less than the lower limit of the fuzzy range, it is determined that the current limit violation is caused by the degradation of the photovoltaic system itself, and the distributed access unit performs a grid disconnection operation. If the power quality anomaly coefficient at the current moment is within the fuzzy range, the distributed access unit first performs power reduction regulation, and if it remains within the fuzzy range, it forcibly performs a grid disconnection operation.

Images