A method for extracting power quality fluctuation periods based on probability statistics

Abstract

This invention discloses a method for extracting power quality fluctuation periods based on probability statistics. By considering the probability distribution and statistical characteristics of power quality index data, the optimal period division scheme for power quality data is determined. Based on the differences between stable and fluctuating periods, a fluctuation threshold is established and fluctuating periods are extracted. A power quality index fluctuation event database is constructed to assist in the identification of abnormal power quality fluctuations. The proposed fluctuation period division method can be used in the field of responsibility division, which can assist in the division of responsibility for interference sources, and more quickly and accurately trace the source of interference to the user, which is conducive to the proactive prevention and early treatment of power quality disturbances.

Description

Technical Field

[0001] This invention belongs to the field of power system method technology, and specifically relates to a method for extracting power quality fluctuation periods based on probability statistics. Background Technology

[0002] On the one hand, the commissioning of nonlinear loads such as electric arc furnaces, electric locomotives, and large converter equipment is constantly increasing; on the other hand, the penetration rate of new energy sources, mainly photovoltaic and wind power, in the power grid is rapidly increasing. The diverse characteristics of power quality exacerbate the severity and complexity of power quality problems at the point of common coupling (PCC), increasing the difficulty of accurate assessment, targeted remediation, and responsibility allocation for power quality issues. Currently, the commonly used assessment standard for power quality is the static limit value of the national standard.

[0003] In the prior art, Chinese invention patent application number 201510528443.2 discloses a power quality monitoring method. This method involves a data acquisition terminal installed in a substation, synchronized with a GPS time synchronization server, to synchronously collect real-time electrical quantity data from monitoring points. The data acquisition terminal transmits the time-stamped real-time electrical quantity data to a cloud computing server via a high-speed wireless communication network. The cloud computing server processes the real-time electrical quantity data to obtain power quality data and non-power quality data, and saves the obtained data to a cloud database server. A cloud-based advanced application analysis server extracts the power quality data and non-power quality data, performing real-time display of the monitoring data, historical trend statistics, report generation and output, waveform playback display, and harmonic source tracing analysis to complete the monitoring, analysis, and management of power quality. This invention significantly reduces the system's operation and maintenance costs, fundamentally improving the scalability of the monitoring system.

[0004] While the static limits of national standards are effective for assessing power quality at the point of common coupling (PCC) during the current time period, they cannot dynamically evaluate deviations from normal power quality indicators or predict subsequent abnormal trends. Furthermore, the rapid increase in the types and number of industrial user loads on a single busbar, along with the widespread use of power electronic equipment, means that interference sources possess diverse emission characteristics. These range from short-term, irregular emission characteristics, such as those from electric vehicles charging, to long-term, periodic emission characteristics, such as those from photovoltaic systems and industrial production facilities. Without extracting the periods of power quality fluctuations, direct responsibility assignment can easily lead to missed or incorrect identifications. Therefore, before identifying power quality anomalies or assigning responsibility, it is necessary to extract the periods of power quality fluctuations and conduct a comprehensive evaluation of these fluctuations. This aids in identifying power quality anomalies and tracing the sources of power quality interference, facilitating proactive prevention and early management of power quality disturbances, and reducing losses to the power grid and users due to power quality problems or untimely remediation. Summary of the Invention

[0005] This invention provides a method for extracting power quality fluctuation periods based on probability statistics, aiming to solve the problem that existing power quality detection methods cannot quickly extract power quality fluctuation periods and cannot comprehensively evaluate the power quality fluctuation periods and fluctuation conditions.

[0006] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0007] A method for extracting power quality fluctuation periods based on probability statistics includes the following steps:

[0008] Step 1: Obtain power quality index data for the power supply bus and feeders of the power quality monitoring system deployed at the substation;

[0009] Step 2: Filter the power quality index data using a Savitzky-Golay filter;

[0010] Step 3: Divide the filtered power quality index data into time periods using a combined method of cumulative distribution function and Fisher's optimal segmentation;

[0011] Step 4: Use the standard deviation within each time period as the basis for judging the fluctuation period, and take the long-term low value of the power quality index data as the judgment threshold. When the standard deviation of a certain time period exceeds the long-term low value of the power quality index data, this time period is considered to be a fluctuation period.

[0012] Step 5: Merge consecutive fluctuation periods to form the final result. Each fluctuation period (denoted as) … ) and its time series (denoted as … );

[0013] Step 6: Construct a database of power quality index fluctuation events and record the fluctuation period and amplitude.

[0014] Furthermore, step 2 specifically includes the following steps:

[0015] Step 21: Select window length This window contains a set of power quality index datasets. Construct a polynomial ( To fit this set of data:

[0016] (1)

[0017] In the formula: Indicates the data used for fitting. Polynomial function, , This represents the maximum order of the polynomial fit. For the polynomial of the th Term coefficient;

[0018] Step 22: Calculate the sum of squared residuals between the fitted data points and the original data points. :

[0019] (2)

[0020] In the formula: To fit the sum of squared residuals between the fitted data points and the original data points, For power quality index dataset The data in For power quality index dataset The number of data points in the data;

[0021] Step 23: Use the least squares method to sum the squares of the residuals. Minimum, can be obtained right The partial derivative should be 0, that is...

[0022] (3)

[0023] Step 24: Solve formula (3) to obtain the Savitzky-Golay filter formula:

[0024] (4)

[0025] In the formula: The smoothing coefficient can be obtained using the least squares method.

[0026] Furthermore, step 3 specifically includes the following steps:

[0027] Step 31: Draw a cumulative distribution plot, whereby the cumulative distribution plot is the cumulative distribution function with respect to the random variable. any value in The curve of the random variable For the power quality index dataset in step 2 Values For power quality datasets The value in is denoted as The Used to describe random variables The probability distribution describes The value is not greater than The possibility of is defined as follows:

[0028] (5)

[0029] In the formula: Represents time series variables, Represents a probability function;

[0030] Step 32: Calculate the cumulative distribution function The inflection point of the curve is denoted as ,satisfy This point is the timing sequence. The fluctuating numerical points will be time-series The value of the middle is The moment As a moment of fluctuation, satisfying The set of fluctuation moments is denoted as ;

[0031] Step 33: With the goal of minimizing the Euclidean distance within each time period, construct the objective function of Fisher's optimal segmentation method, as shown in equation (6):

[0032] (6)

[0033] (7)

[0034] In the formula: for The harmonic ordered samples are divided into A time period division scheme, Indicates the first Intra-segment sample distance for each time period, For the first The first time segment of each period For the first The average harmonic data for each time period; for Power quality data at any given time;

[0035] Step 34: Set several different numbers of clusters, using the overlap between Fisher's time-segmentation points and the inflection points of the cumulative distribution map, and the proportion of segmented time periods, as the criteria for determining the optimal number of Fisher's clusters; select the clusters with the largest overlap and the largest proportion of segmented time periods as the optimal number of clusters; after determining the optimal number of clusters, determine the time-segmentation points using Fisher's optimal segmentation method, and denote the set of time-segmentation points as follows. ;

[0036] Step 35: Take the union of the set of time period segmentation points and the set of fluctuation moments as the final result of time period division, denoted as . .

[0037] Furthermore, step 4 specifically includes the following steps:

[0038] Step 41: Calculate the standard deviation for each time period, and then compare the standard deviation with the long-term low value of the power quality index data. The formula for the discrimination threshold is as follows:

[0039] (8)

[0040] In the formula: This refers to long-term low values ​​of power quality index data;

[0041] Step 42: Increase the standard deviation to over The time period is marked as a fluctuation period.

[0042] Furthermore, step 6 specifically includes the following steps:

[0043] Step 61: Match the fluctuation period with historical fluctuation periods for a given fluctuation sequence. , , ( , They are respectively , (dimensional vector), whose dynamic distance correlation coefficient is expressed as Its expression is shown in equation (9):

[0044] (9)

[0045] in: represent Sequence number elements and Sequence number The distance between elements is calculated using Euclidean distance. ;

[0046] The final calculation result is , The dynamic distance correlation coefficient represents the similarity between two fluctuating time series of different lengths. The calculation result is lower than This is considered a period of abnormal fluctuation. Based on engineering experience, the value here is 20;

[0047] Step 62: Compare the fluctuation amplitude during the abnormal fluctuation period with the national standard limit. If the fluctuation amplitude during the abnormal fluctuation period exceeds the national standard limit, send an abnormal fluctuation indicator signal.

[0048] Step 63: Archive the abnormal fluctuation events based on the abnormal fluctuation indicator signals, and record the fluctuation period and fluctuation amplitude.

[0049] Compared with the prior art, the present invention has the following technical effects:

[0050] 1. This invention determines the optimal time period division scheme for power quality data by considering the probability distribution and statistical characteristics of the data. Based on the differences between stable and fluctuating periods, it establishes a fluctuation threshold and extracts fluctuating periods, and constructs a power quality index fluctuation event database. This allows for the automatic identification of power quality anomalies at important nodes and lines with power quality monitoring systems. It can also be used in the field of responsibility allocation to assist in the allocation of responsibility for interference sources, enabling faster and more accurate tracing of interference source users, and facilitating proactive prevention and early management of power quality disturbances. Attached Figure Description

[0051] Figure 1 This is a flowchart of a method for extracting power quality fluctuation periods based on probability statistics, as described in this invention. Detailed Implementation

[0052] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0053] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below in conjunction with specific embodiments of the present application and with reference to the accompanying drawings.

[0054] like Figure 1 As shown, the method for extracting power quality fluctuation periods based on probability statistics according to the present invention includes the following steps:

[0055] Step 1: Obtain power quality index data for the power supply bus and feeders of the power quality monitoring system deployed at the substation;

[0056] The power quality index data include: 2nd-50th harmonic current, effective voltage value and phase angle, interharmonics, voltage deviation, voltage fluctuation and flicker, three-phase imbalance, voltage sag, and frequency deviation.

[0057] Step 2: Filter the power quality index data using a Savitzky-Golay filter;

[0058] Savitzky-Golay filters can perform multi-objective fitting on data of arbitrary window size in the time domain, which removes noise interference such as spikes while ensuring that the shape and length of the power quality signal remain unchanged, and the overall trend of power quality is more obvious after processing.

[0059] Step 3: Divide the filtered power quality index data into time periods using a combined method of cumulative distribution function and Fisher's optimal segmentation;

[0060] Step 4: Use the standard deviation within each time period as the basis for judging the fluctuation period, and take the long-term low value of the power quality index data as the judgment threshold. When the standard deviation of a certain time period exceeds the long-term low value of the power quality index data, this time period is considered to be a fluctuation period.

[0061] Step 5: Merge consecutive fluctuation periods to form the final result. Each fluctuation period (denoted as) … ) and its time series (denoted as … );

[0062] Step 6: Construct a database of power quality index fluctuation events and record the fluctuation period and amplitude.

[0063] Step 2 specifically includes the following steps:

[0064] Step 21: Select window length This window contains a set of power quality index datasets. Construct a polynomial ( To fit this set of data:

[0065] (1)

[0066] In the formula: Indicates the data used for fitting. Polynomial function, , This represents the maximum order of the polynomial fit. For the polynomial of the th Term coefficient;

[0067] Step 22: Calculate the sum of squared residuals between the fitted data points and the original data points. :

[0068] (2)

[0069] In the formula: To fit the sum of squared residuals between the fitted data points and the original data points, For power quality index dataset The data in For power quality index dataset The number of data points in the data;

[0070] Step 23: Use the least squares method to sum the squares of the residuals. Minimum, can be obtained right The partial derivative should be 0, that is...

[0071] (3)

[0072] Step 24: Solve formula (3) to obtain the Savitzky-Golay filter formula:

[0073] (4)

[0074] In the formula: The smoothing coefficient can be obtained using the least squares method.

[0075] Step 3 specifically includes the following steps:

[0076] Step 31: Draw a cumulative distribution plot, whereby the cumulative distribution plot is the cumulative distribution function with respect to the random variable. any value in The curve of the random variable For the power quality index dataset in step 2 Values For power quality datasets The value in is denoted as The Used to describe random variables The probability distribution describes The value is not greater than The possibility of is defined as follows:

[0077] (5)

[0078] In the formula: Represents time series variables, Represents a probability function;

[0079] Step 32: Calculate the cumulative distribution function The inflection point of the curve is denoted as ,satisfy This point is the timing sequence. The fluctuating numerical points will be time-series The value of the middle is The moment As a moment of fluctuation, satisfying The set of fluctuation moments is denoted as ;

[0080] Step 33: With the goal of minimizing the Euclidean distance within each time period, construct the objective function of Fisher's optimal segmentation method, as shown in equation (6):

[0081] (6)

[0082] (7)

[0083] In the formula: for The harmonic ordered samples are divided into A time period division scheme, Indicates the first Intra-segment sample distance for each time period, For the first The first time segment of each period For the first The average harmonic data for each time period; for Power quality data at any given time;

[0084] Step 34: Set several different numbers of clusters, using the overlap between Fisher's time-segmentation points and the inflection points of the cumulative distribution map, and the proportion of segmented time periods, as the criteria for determining the optimal number of Fisher's clusters; select the clusters with the largest overlap and the largest proportion of segmented time periods as the optimal number of clusters; after determining the optimal number of clusters, determine the time-segmentation points using Fisher's optimal segmentation method, and denote the set of time-segmentation points as follows. ;

[0085] Step 35: Take the union of the set of time period segmentation points and the set of fluctuation moments as the final result of time period division, denoted as . .

[0086] Step 4 specifically includes the following steps:

[0087] Step 41: Calculate the standard deviation for each time period, and then compare the standard deviation with the long-term low value of the power quality index data. The formula for the discrimination threshold is as follows:

[0088] (8)

[0089] In the formula: This refers to long-term low values ​​of power quality index data;

[0090] Step 42: Increase the standard deviation to over The time period is marked as a fluctuation period.

[0091] Step 6 specifically includes the following steps:

[0092] Step 61: Match the fluctuation period with historical fluctuation periods for a given fluctuation sequence. , , , They are respectively , A dimensional vector, whose dynamic distance correlation coefficient is expressed as: Its expression is shown in equation (9):

[0093] (9)

[0094] in: represent Sequence number elements and Sequence number The distance between elements is calculated using Euclidean distance. ;

[0095] The final calculation result is , The dynamic distance correlation coefficient represents the similarity between two fluctuating time series of different lengths. The calculation result is lower than This is considered a period of abnormal fluctuation. Based on engineering experience, the value here is 20;

[0096] Step 62: Compare the fluctuation amplitude during the abnormal fluctuation period with the national standard limit. If the fluctuation amplitude during the abnormal fluctuation period exceeds the national standard limit, send an abnormal fluctuation indicator signal.

[0097] Step 63: Archive the abnormal fluctuation events based on the abnormal fluctuation indicator signals, and record the fluctuation period and fluctuation amplitude.

[0098] The above description is only a preferred embodiment of the present invention. It should be noted that those skilled in the art can make several modifications and improvements without departing from the inventive concept of the present invention, and these all fall within the protection scope of the present invention.

Claims

1. A method for extracting power quality fluctuation periods based on probability statistics, characterized in that, Includes the following steps: Step 1: Obtain power quality index data for the power supply bus and feeders of the power quality monitoring system deployed at the substation; Step 2: Filter the power quality index data using a Savitzky-Golay filter; Step 3: Divide the filtered power quality index data into time periods using a combined method of cumulative distribution function and Fisher's optimal segmentation; Step 4: Use the standard deviation within each time period as the basis for judging the fluctuation period, and take the long-term low value of the power quality index data as the judgment threshold. When the standard deviation of a certain time period exceeds the long-term low value of the power quality index data, this time period is considered to be a fluctuation period. Step 5: Merge consecutive fluctuation periods to form the final result. Period of fluctuation … and its time series … ; Step 6: Construct a database of power quality index fluctuation events and record the fluctuation periods and amplitudes; Step 3 specifically includes the following steps: Step 31: Random Variables For the power quality index dataset in step 2 Draw a cumulative distribution plot, wherein the cumulative distribution plot is the cumulative distribution function with respect to the random variable. any value in The curve, taking values For power quality datasets The value in is denoted as The Used to describe random variables The probability distribution describes The value is not greater than The possibility of is defined as follows: （5） In the formula: Represents time series variables, Represents a probability function; Step 32: Calculate the cumulative distribution function The inflection point of the curve is denoted as ,satisfy This point is the random variable. The points where the values ​​fluctuate will be random variables. The value of the middle is The moment As a moment of fluctuation, satisfying The set of fluctuation moments is denoted as ; Step 33: With the goal of minimizing the Euclidean distance within each time period, construct the objective function of Fisher's optimal segmentation method, as shown in equation (6): （6） （7） In the formula: for The harmonic ordered samples are divided into A time period division scheme, Indicates the first Intra-segment sample distance for each time period, For the first The first time segment of each period For the first The average harmonic data for each time period; for Power quality data at any given time; Step 34: Set several different numbers of clusters, using the overlap between Fisher's time-segmentation points and the inflection points of the cumulative distribution map, and the proportion of segmented time periods, as the criteria for determining the optimal number of Fisher's clusters; select the clusters with the largest overlap and the largest proportion of segmented time periods as the optimal number of clusters; after determining the optimal number of clusters, determine the time-segmentation points using Fisher's optimal segmentation method, and denote the set of time-segmentation points as follows. ; Step 35: Take the union of the set of time period segmentation points and the set of fluctuation moments as the final result of time period division, denoted as . .

2. The method for extracting power quality fluctuation periods based on probability statistics according to claim 1, characterized in that, Step 2 specifically includes the following steps: Step 21: Select window length This window contains a set of power quality index datasets. Construct a polynomial To fit this set of data: （1） In the formula: Indicates the data used for fitting. Polynomial function, , This represents the maximum order of the polynomial fit. For the polynomial of the th Term coefficient; Step 22: Calculate the sum of squared residuals between the fitted data points and the original data points. : （2） In the formula: To fit the sum of squared residuals between the fitted data points and the original data points, For power quality index dataset The data in For power quality index dataset The number of data points in the data; Step 23: Use the least squares method to sum the squares of the residuals. Minimum, get right The partial derivative should be 0, that is... （3） Step 24: Solve formula (3) to obtain the Savitzky-Golay filter formula: （4） In the formula: The smoothing coefficient is obtained using the least squares method.

3. The method for extracting power quality fluctuation periods based on probability statistics according to claim 1, characterized in that, Step 4 specifically includes: Step 41: Calculate the standard deviation for each time period, and then compare the standard deviation with the long-term low value of the power quality index data. The formula for the discrimination threshold is as follows: （8） In the formula: This refers to long-term low values ​​of power quality index data; Step 42: Increase the standard deviation to over The time period is marked as a fluctuation period.

4. The method for extracting power quality fluctuation periods based on probability statistics according to claim 3, characterized in that, Step 6 specifically includes: Step 61: Match the fluctuation period with historical fluctuation periods for a given fluctuation sequence. , , , They are respectively , A dimensional vector, whose dynamic distance correlation coefficient is expressed as: Its expression is shown in equation (9): （9） in: represent Sequence number elements and Sequence number The distance between elements is calculated using Euclidean distance. ; The final calculation result is , The dynamic distance correlation coefficient represents the similarity between two fluctuating time series of different lengths. The calculation result is lower than This is considered a period of abnormal fluctuation. Based on engineering experience, the value here is 20; Step 62: Compare the fluctuation amplitude during the abnormal fluctuation period with the national standard limit. If the fluctuation amplitude during the abnormal fluctuation period exceeds the national standard limit, send an abnormal fluctuation indicator signal. Step 63: Archive the abnormal fluctuation events based on the abnormal fluctuation indicator signals, and record the fluctuation period and fluctuation amplitude.

Images