Evaluation method of intelligent fusion terminal electric energy quality problem monitoring and treatment system

Abstract

The invention belongs to the technical field of electric energy quality monitoring and treatment systems, and particularly relates to an intelligent fusion terminal electric energy quality problem monitoring and treatment system evaluation method. Non-parametric kernel density estimation with the capacity of fitting a random variable distribution density function is adopted as a main framework, and reliability evaluation is conducted on the precision of the electric energy quality monitoring and management system of the transformer area intelligent fusion terminal. According to the method, the probability distribution function of the power quality data can be fitted according to historical data, so that the confidence interval of each power quality monitoring and treatment system index is established. And the interval is taken as a monitoring reliability judgment basis, and reliability evaluation is carried out on the precision of the electric energy quality monitoring management system under the continuous operation state of the terminal. The method can solve the problems that according to a traditional evaluation method, the influence of uncertainty factors and human factors on the evaluation result is large, the calculation complexity is high, and correlation analysis is not enough.

Description

Technical Field

[0001] The present invention belongs to the technical field of power quality monitoring and management systems, and specifically relates to an evaluation method for a smart fusion terminal power quality problem monitoring and management system. Background Art

[0002] The distribution network is a hub connecting the energy side and the user side, and is one of the important core links of the power system. With the emergence and large-scale access of diversified loads such as distributed power sources, electric vehicles and energy storage, the function and form of the distribution network are undergoing significant changes, gradually developing from a one-way flow to a two-way flow, showing an increasingly complex "multi-source" feature, which poses severe challenges to the safe operation and power supply quality of the distribution network. At the same time, the distribution network has many access points and a wide power supply range, flexible and diverse communication methods, large data volume, complex and diverse data models, lack of a common model semantic foundation, and poor scalability of the model architecture. In addition, with the deployment of a large number of sensing terminals in the distribution network, the access of distributed energy, and the change in the requirements for the operation and maintenance of distribution network equipment, new requirements have been put forward for the security protection, computing power, information model, massive terminal operation and maintenance, and advanced business applications of the distribution network terminals. The traditional distribution network data collection and information integration methods can no longer adapt to the new form of distribution network operation and management needs, and it is necessary to explore new global solutions from an overall perspective.

[0003] The existing methods for testing power quality data of intelligent fusion terminals in substations are mainly to test the monitoring accuracy of the equipment through the standard source method. As long as the accuracy requirements are met during the test, the equipment is considered qualified. However, this detection method ignores the problem of the power quality monitoring and management system of the terminal equipment in continuous operation. Therefore, there is currently no method or research to evaluate the power quality monitoring and management system of intelligent fusion terminals in substations. Summary of the invention

[0004] The purpose of the present invention is to provide an evaluation method for an intelligent integrated terminal power quality problem monitoring and management system to address the problems existing in the prior art. The evaluation method has high evaluation accuracy, is conducive to discovering hidden dangers that reduce terminal reliability, and improves the intelligence level of terminal detection.

[0005] The technical solution of the present invention is:

[0006] The evaluation method of the intelligent fusion terminal power quality problem monitoring and management system includes the following steps:

[0007] S1: Collect the power quality data of the power grid company to be evaluated within a certain period of time as sample data;

[0008] S2: preprocess the sample data;

[0009] S3: Use the preprocessed sample data to fit the nonparametric kernel density estimation model to obtain the data probability density function;

[0010] S4: Integrate the probability density function to obtain the cumulative probability function, and generate a confidence interval with a data confidence level of 99%;

[0011] S5: According to the confidence interval, determine the number M of abnormal data in the monitoring data of the substation intelligent fusion terminal. When the detected data is within the confidence interval, it means that the data is normal. If the detected data is no longer within the confidence interval, it means that the data is abnormal. If the substation intelligent fusion terminal does not detect the data, it means that the lost data exceeds the confidence interval, which is also abnormal data.

[0012] Specifically, the step S2 includes the following steps:

[0013] S2.1. Supplement missing data in sample data by interpolation;

[0014] S2.2, by formula Calculate the interpolation result L of the missing value n (x), where n is the number of data before and after the missing value, x is the subscript of the missing value, and x i is a non-missing value y i The subscript of x j is a non-missing value y j The subscript of ;

[0015] S2.3, by formula Calculate the normalized difference processed sample data Where μ is the mean of the sample data, σ is the standard deviation of the sample data, and x gk For the data in the g-th row and k-th column in the power quality data matrix, complete the preprocessing of the sample data.

[0016] Specifically, the step S3 includes the following steps:

[0017] S3.1. Divide the preprocessed sample data into different time series according to type. Each type of data is a matrix of X, X∈R N*T , R is the real number space, N is the number of nodes, and T is the total monitoring time;

[0018] S3.2, by formula Calculate the rate density function f(x), where K(·) is the kernel function, h is the bandwidth coefficient, n is the total number of samples, and x m is the mth sample;

[0019] S3.3. The mathematical model of humpback whales’ behavior of surrounding prey is established by the formula D = |C·X*(t)-X(t)| and X(t+1) = X*(t)-A·D, where D is the distance between the search agent and the target prey, t is the current iteration number, X*(t) is the current optimal solution position, X(t) is the current position vector, X(t+1) is the position vector at the t+1 iteration, A and C are coefficient vectors, and A = 2a·r a -a, C = 2·r c , r a With r c are all random vectors between [0,1], a represents a linear decrease from 2 to 0 during the iteration process, X(t+1) = D′·e bl ·cos(2πl)+X*(t), where D′=|X*(t)-X(t)|, b is a constant that defines the shape of the logarithmic spiral, l is a random number between [-1,1], and e is the base of the natural logarithm, e=2.71828183;

[0020] S3.4, by formula The mathematical model of position update is calculated, where p is a random number between [0,1];

[0021] S3.5, through the formula D = | C · X rand (t)-X(t)| and X(t+1)=X rand (t)-A·D establishes a prey search model, where X rand (t) is the search agent position vector randomly selected from the population;

[0022] S3.6. Set the number of iterations to T max , the number of whales is m, and the value of candidate bandwidth h is initialized. The value range of candidate bandwidth h is 0.25-1.1 times h ref ,

[0023] p is the model order, n is the total number of data;

[0024] S3.7, by formula Calculate the fitness LSCV(H), where L gk =(x g -x k ) T H -1 (x g -x k ), H = h 2 S, S is the sample covariance matrix, x g and x k is the sample data;

[0025] S3.8, if |A|≥1, go to step 3.4; if |A|<1, go to step S3.5;

[0026] S3.9. Repeat S3.7 to S3.8 until the number of iterations reaches the set number of iterations M.

[0027] Specifically, the step S4 includes the following steps:

[0028] S4.1. Output the position of the best individual whale and obtain the optimal bandwidth h;

[0029] S4.2, by formula Calculate the cumulative probability function F(x), where Υ(·) is the cumulative probability function of the Gaussian kernel function;

[0030] S4.3. Establish a confidence interval with a 99% confidence level for the data.

[0031]

[0032] Specifically, after step S5, the formula The reliability r of the monitoring data of the intelligent fusion terminal in the substation area is calculated, where N is the total number of test data.

[0033] The complex installation, high cost and large communication volume of special power quality monitoring devices have hindered the large-scale promotion of power quality disturbance monitoring and management. The traditional distribution transformer terminal has the functions of distribution transformer monitoring, data collection and communication networking, but the function expansion is not flexible and the power quality monitoring function lacks refinement. To this end, scholars have carried out a lot of research on the problem of power quality disturbance perception from the algorithm level. The idea is usually to extract features first and then use different classification and recognition methods to accurately identify disturbances. Common feature extraction methods include short-time Fourier transform, wavelet transform, S transform, empirical mode decomposition, etc. In terms of classification and recognition methods, similarity matching, neural network, support vector machine, Bayesian classification, fuzzy comprehensive evaluation, K-nearest neighbor graph, etc. are usually used. On this basis, based on the pattern recognition idea, the classifier is designed using the classification algorithm, and the disturbance category can be further identified according to the monitoring data. The main methods include neural network, support vector machine, principal component analysis simplification, fuzzy comprehensive evaluation, etc. With the complexity of power grid equipment and the regionalization of power consumption patterns, it is difficult to establish an accurate and universal mathematical and statistical model for complex time series data in the power system. At the same time, the information loss in the feature extraction process and the excessive complexity of the classification model have also made the defects of the existing methods increasingly prominent. In the era of big data, data-driven methods are highly favored. Their characteristics of not relying on physical models effectively make up for the shortcomings of mathematical models. Applying deep learning to power quality classification has become a trend. However, the above-mentioned perception methods still face the problem of not being able to balance the dynamic real-time and accuracy of the perception process with the efficiency of algorithm calculation. How to apply perception technology with low computational complexity and high accuracy in intelligent perception terminals is still a problem that needs to be solved urgently.

[0034] In terms of the governance of power quality disturbance problems, the current research at home and abroad focuses on the quantification of user governance needs and governance decision-making. In terms of the quantification of user governance needs, existing research mainly uses user power quality monitoring data, based on power consumption characteristics identification and parameter estimation, to build a multi-dimensional spatiotemporal data mining user governance demand analysis model; or by defining different demand quantification indicators, the power supply demand of power users is characterized from the perspective of power quality assessment, but existing research has failed to timely user models based on the actual operating status of users. In terms of micro-applications of power quality disturbance governance decision-making, at this stage, it is mainly based on the governance scheme of high-quality power technology and customized power equipment, considering the economy, reliability and serviceability, etc., to establish a comprehensive evaluation index system for governance effects, estimate the user power quality governance effect, and select the governance scheme with the highest satisfaction. However, this method has certain limitations due to the complexity of the entire process from determining the scheme to governance decision-making, the long business cycle, the difficulty of deploying governance services, the low utilization rate of power quality data, and the inability to present transparent governance decisions on the user side. Although some power grid companies in China have installed online power quality monitoring systems, collected massive user data through the business middle platform, and designed an overall power quality intelligent analysis system framework and application plan covering the perception layer, network layer, platform layer and application layer, there is still a lack of practical and feasible specific technical solutions and results on how to develop a power quality disturbance management service platform to present a transparent management solution.

[0035] The beneficial effects of the present invention are: it can accurately describe the probability density function of the data, realize the reliability analysis and evaluation of the power quality monitoring of the intelligent fusion terminal in the substation, has high evaluation accuracy, is conducive to discovering the hidden dangers that reduce the reliability of the terminal, and improves the intelligent level of terminal detection. It has high practical value in the engineering field. The non-parametric kernel density estimation with the ability to fit the distribution density function of random variables is used as the main framework to perform reliability evaluation on the power quality monitoring accuracy of the intelligent fusion terminal in the substation. This method can fit the probability distribution function of the power quality data according to historical data, so as to establish the confidence interval of each power quality monitoring indicator. And use this interval as the basis for judging the monitoring reliability, and perform reliability evaluation on the power quality monitoring accuracy under the continuous operation state of the terminal. It can overcome the problems of the traditional evaluation method that the uncertainty factors and human factors have a large impact on the evaluation results, the high calculation complexity, and the insufficient correlation analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of the method for evaluating the power quality monitoring and management system of the intelligent fusion terminal in the substation area of ​​the present invention. DETAILED DESCRIPTION

[0037] The technical solution of the present invention is described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0038] Example 1

[0039] like Figure 1 The flowchart of the intelligent fusion terminal power quality problem monitoring and management system evaluation method provided in this embodiment includes the following steps:

[0040] S1: Collect the power quality data of the power grid company to be evaluated within a certain period of time as sample data;

[0041] S2: preprocess the sample data;

[0042] S3: Use the preprocessed sample data to fit the nonparametric kernel density estimation model to obtain the data probability density function;

[0043] S4: Integrate the probability density function to obtain the cumulative probability function, and generate a confidence interval with a data confidence level of 99%;

[0044] S5: According to the confidence interval, determine the number M of abnormal data in the monitoring data of the substation intelligent fusion terminal. When the detected data is within the confidence interval, it means that the data is normal. If the detected data is no longer within the confidence interval, it means that the data is abnormal. If the substation intelligent fusion terminal does not detect the data, it means that the lost data exceeds the confidence interval, which is also abnormal data.

[0045] Then through the formula The reliability r of the monitoring data of the intelligent fusion terminal in the substation area is calculated, where N is the total number of test data.

[0046] This method can accurately describe the probability density function of the data, realize the reliability analysis and evaluation of the power quality monitoring of the intelligent fusion terminal in the substation, has high evaluation accuracy, is conducive to discovering hidden dangers that reduce the reliability of the terminal, and improves the intelligence level of terminal detection. It has high practical value in the engineering field.

[0047] Example 2

[0048] This embodiment aims at step S2 described in embodiment 1, and supplements the missing data by interpolation, and interpolation is performed by Lagrange interpolation method, including the following steps:

[0049] S2.1. Supplement the missing data in the sample data by interpolation, select 5 data before and after the missing value from the original sample data, and group the selected 10 data, that is, n = 10;

[0050] S2.2, by formula Calculate the interpolation result L of the missing value n (x), where n is the number of data before and after the missing value, x is the subscript of the missing value, and x i is a non-missing value yi The subscript of x j is a non-missing value y j All missing values ​​are inserted in turn until no value is missing;

[0051] S2.3, use Z-score to normalize the data, which can delete the data dimension and only retain the characteristic shape of the original data. The specific formula is Calculate the normalized difference processed sample data Where μ is the mean of the sample data, σ is the standard deviation of the sample data, and x gk For the data in the g-th row and k-th column in the power quality data matrix, complete the preprocessing of the sample data.

[0052] Example 3

[0053] This embodiment is directed to step S2 described in embodiment 1, and step S3 described in embodiment 1 includes the following steps:

[0054] S3.1. The preprocessed sample data is divided into different time series according to the type. The terminal will monitor the power quality data from each node. Each type of data is a matrix of X, X∈R N*T , R is the real number space, N is the number of nodes, and T is the total monitoring time;

[0055] S3.2, by formula Calculate the rate density function f(x), where K(·) is the kernel function, h is the bandwidth coefficient, n is the total number of samples, and x m For the mth sample, the probability density function based on the Gaussian kernel function can be written as:

[0056]

[0057] S3.3. The mathematical model of humpback whales’ behavior of surrounding prey is established by the formula D = |C·X*(t)-X(t)| and X(t+1) = X*(t)-A·D, where D is the distance between the search agent and the target prey, t is the current iteration number, X*(t) is the current optimal solution position, X(t) is the current position vector, X(t+1) is the position vector at the t+1 iteration, A and C are coefficient vectors, and A = 2a·r a -a, C = 2·r c , r a With r c are all random vectors between [0,1], a represents a linear decrease from 2 to 0 during the iteration process, X(t+1) = D′·e bl·cos(2πl)+X*(t), where D′=|X*(t)-X(t)|, b is a constant that defines the shape of the logarithmic spiral, l is a random number between [-1,1], and e is the base of the natural logarithm, e=2.71828183;

[0058] S3.4, by formula The mathematical model of position update is calculated, where p is a random number between [0,1];

[0059] S3.5, through the formula D = | C · X rand (t)-X(t)| and X(t+1)=X rand (t)-A·D establishes a prey search model, where X rand (t) is the search agent position vector randomly selected from the population;

[0060] S3.6. Set the number of iterations to T max , the number of whales is m, and the value of candidate bandwidth h is initialized. The value range of candidate bandwidth h is 0.25-1.1 times h ref ,

[0061] p is the model order, n is the total number of data;

[0062] S3.7, by formula Calculate the fitness LSCV(H), where L gk =(x g -x k ) T H -1 (x g -x k ), H = h 2 S, S is the sample covariance matrix, x g and x k is the sample data;

[0063] S3.8, if |A|≥1, go to step 3.4; if |A|<1, go to step S3.5;

[0064] S3.9. Repeat S3.7 to S3.8 until the number of iterations reaches the set number of iterations M.

[0065] Example 4

[0066] This embodiment is directed to step S4 described in embodiment 1, and step S4 includes the following steps:

[0067] S4.1. Output the position of the best individual whale and obtain the optimal bandwidth h;

[0068] S4.2, by formula Calculate the cumulative probability function F(x), where Υ(·) is the cumulative probability function of the Gaussian kernel function;

[0069] S4.3. Establish a confidence interval with a 99% confidence level for the data.

[0070]

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, ordinary technicians in the field should understand that the specific implementation methods of the present invention can still be modified or some technical features can be replaced by equivalents without departing from the spirit of the technical solution of the present invention, which should be included in the scope of the technical solution for protection of the present invention.

Claims

1. Evaluation method of intelligent fusion terminal power quality problem monitoring and management system, characterized by: The steps include: S1: Collect the power quality data of the power grid company to be evaluated within a certain period of time as sample data; S2: preprocess the sample data; S3: Use the preprocessed sample data to fit the nonparametric kernel density estimation model to obtain the data probability density function; S4: Integrate the probability density function to obtain the cumulative probability function, and generate a confidence interval with a data confidence level of 99%; S5: According to the confidence interval, determine the number M of abnormal data in the monitoring data of the substation intelligent fusion terminal. When the detected data is within the confidence interval, it means that the data is normal. If the detected data is no longer within the confidence interval, it means that the data is abnormal. If the substation intelligent fusion terminal does not detect the data, it means that the lost data exceeds the confidence interval, which is also abnormal data.

2. According to claim 1, the intelligent fusion terminal power quality problem monitoring and management system evaluation method is characterized in that: The step S2 comprises the following steps: S2.

1. Supplement missing data in sample data by interpolation; S2.2, by formula Calculate the interpolation result L of the missing value n (x), where n is the number of data before and after the missing value, x is the subscript of the missing value, and x i is a non-missing value y i The subscript of x j is a non-missing value y j The subscript of ; S2.3, by formula Calculate the normalized difference processed sample data Where μ is the mean of the sample data, σ is the standard deviation of the sample data, and x gk For the data in the g-th row and k-th column in the power quality data matrix, complete the preprocessing of the sample data.

3. According to claim 1, the intelligent fusion terminal power quality problem monitoring and management system evaluation method is characterized in that: The step S3 comprises the following steps: S3.

1. Divide the preprocessed sample data into different time series according to type. Each type of data is a matrix of X, X∈R N*T , R is the real number space, N is the number of nodes, and T is the total monitoring time; S3.2, by formula Calculate the rate density function f(x), where K(·) is the kernel function, h is the bandwidth coefficient, n is the total number of samples, and x m is the mth sample; S3.

3. The mathematical model of humpback whales’ behavior of surrounding prey is established by the formula D = |C·X*(t)-X(t)| and X(t+1) = X*(t)-A·D, where D is the distance between the search agent and the target prey, t is the current iteration number, X*(t) is the current optimal solution position, X(t) is the current position vector, X(t+1) is the position vector at the t+1 iteration, A and C are coefficient vectors, and A = 2a·r a -a, C = 2·r c , r a With r c are all random vectors between [0,1], a represents a linear decrease from 2 to 0 during the iteration process, X(t+1) = D′·e bl ·cos(2πl)+X*(t), where D′=|X*(t)-X(t)|, b is a constant that defines the shape of the logarithmic spiral, l is a random number between [-1,1], and e is the base of the natural logarithm, e=2.71828183; S3.4, by formula The mathematical model of position update is calculated, where p is a random number between [0,1]; S3.5, through the formula D = | C · X rand (t)-X(t)| and X(t+1)=X rand (t)-A·D establishes a prey search model, where X rand (t) is the search agent position vector randomly selected from the population; S3.

6. Set the number of iterations to T max , the number of whales is m, and the value of candidate bandwidth h is initialized. The value range of candidate bandwidth h is 0.25-1.1 times h ref , p is the model order, n is the total number of data; S3.7, by formula Calculate the fitness LSCV(H), where L gk =(x g -x k ) T H -1 (x g -x k ), H = h 2 S, S is the sample covariance matrix, x g and x k is the sample data; S3.8, if |A|≥1, go to step 3.4; if |A|<1, go to step S3.5; S3.

9. Repeat S3.7 to S3.8 until the number of iterations reaches the set number of iterations M.

4. According to claim 1, the intelligent fusion terminal power quality problem monitoring and management system evaluation method is characterized in that: The step S4 comprises the following steps: S4.

1. Output the position of the best individual whale and obtain the optimal bandwidth h; S4.

2. By formula Calculate the cumulative probability function F(x), where Υ(·) is the cumulative probability function of the Gaussian kernel function; S4.

3. Establish a confidence interval with a 99% confidence level for the data.

5. According to claim 1, the intelligent fusion terminal power quality problem monitoring and management system evaluation method is characterized in that: After step S5, the formula The reliability r of the monitoring data of the intelligent fusion terminal in the substation area is calculated, where N is the total number of test data.

Images