Power failure active monitoring method and system based on machine learning

Abstract

The invention discloses an active power failure monitoring method and system based on machine learning. The method comprises the steps of data acquisition, data processing, establishment of an active power failure monitoring model, adaptive optimization and active power failure monitoring. The invention relates to the technical field of power failure monitoring, in particular to a power failure active monitoring method and system based on machine learning, according to the scheme, noise data is removed by calculating a deviation item, constructing an entropy supplementing item and calculating noise intensity, and the data quality and the accuracy of a monitoring result are improved; the model is established by constructing a depth mapping function, designing a depth feature learning layer and constructing a loss function, so that the prediction performance, precision and stability of the model are enhanced; parameter search is carried out by designing an efficiency coordination factor, a direction transformation function and a noise reduction function, the search process is guided to evolve towards a more efficient direction, the search efficiency is improved, the diversity in the search process is increased, and jumping out of a local optimal solution and exploration of a global better solution are facilitated.

Description

Technical Field

[0001] The present invention relates to the technical field of power outage monitoring, and in particular to a method and system for active power outage monitoring based on machine learning. Background Art

[0002] The machine learning-based active power outage monitoring method and system is an advanced system that uses machine learning technology to predict power grid faults. It can improve the accuracy and efficiency of power outage monitoring and provide users with more reliable power services.

[0003] Traditional active power outage monitoring methods have problems such as insufficient data processing precision and large noise interference, which limits monitoring accuracy; traditional active power outage monitoring models have problems such as difficulty in mining potential features in data, and low prediction performance, accuracy and stability of the model; the optimization process of traditional active power outage monitoring models has problems such as low search efficiency, easy falling into local optimal solutions and insufficient flexibility in parameter adjustment. Summary of the invention

[0004] In view of the above situation, in order to overcome the defects of the prior art, the present invention provides a method and system for active power outage monitoring based on machine learning. In view of the problems of insufficient data processing accuracy and large noise interference in traditional active power outage monitoring methods, which lead to limited monitoring accuracy, this scheme removes noise data by neighborhood division, calculation of deviation terms, construction of entropy compensation terms, and calculation of noise intensity, thereby enhancing the ability to recognize data features, making data processing more comprehensive and accurate, and improving the quality of data and the accuracy of monitoring results. In view of the problems of difficulty in mining potential features in data and low prediction performance, accuracy and stability of traditional active power outage monitoring models, this scheme sets model labels, designs smoothing processing layers, constructs deep mapping functions, and designs deep feature learning. The active monitoring model for power outages is established by adopting the learning layer, dynamic deactivation processing, designing the compensation output layer and constructing the loss function, which can more deeply mine the potential features in the data, more comprehensively capture the key information in the data, and enhance the prediction performance, accuracy and stability of the model; In view of the problems of low search efficiency, easy to fall into the local optimal solution and insufficient flexibility of parameter adjustment in the optimization process of the traditional active monitoring model for power outages, this scheme guides the search process to evolve in a more efficient direction by designing efficiency coordination factors, designing direction transformation functions, designing noise reduction functions, dynamic evolution guided search, adaptive following search and dynamic search, avoiding blind search, thereby improving search efficiency, increasing the diversity of the search process, and helping to jump out of the local optimal solution and explore a better global solution.

[0005] The technical solution adopted by the present invention is as follows: a method for active power outage monitoring based on machine learning, the method comprising the following steps:

[0006] Step S1: data collection;

[0007] Step S2: data processing;

[0008] Step S3: Establishing a power outage active monitoring model;

[0009] Step S4: adaptive optimization;

[0010] Step S5: Active monitoring of power outages.

[0011] Furthermore, in step S1, the data collection is to collect historical power parameter data, equipment operation data, environmental data and power status data, the power parameter data includes current, voltage, power, frequency and phase angle; the equipment operation data includes the temperature of the transformer and switch cabinet, and the opening and closing status of the circuit breaker; the environmental data includes temperature, humidity, wind speed and precipitation; the power status data includes normal state and power outage state.

[0012] Furthermore, in step S2, the data processing specifically includes the following steps:

[0013] Step S21: Neighborhood division, constructing data feature vectors, and setting the m data points closest to each data point as the neighborhood data set of the data point;

[0014] Step S22: Calculate the deviation term, which is expressed as follows:

[0015] ;

[0016] Where i represents the index of the data vector, represents the i-th data vector, represents the deviation term of the i-th data vector, n represents the total number of data vectors, j represents the index of the data vector, represents the jth data vector, represents the mean data vector of all data vectors, Indicates the Mahalanobis distance, e represents a natural constant, m represents the total number of data vectors in the neighborhood data set, and k represents the index of the data vector in the neighborhood data set. represents the neighborhood data set of the i-th data vector, Indicates that it belongs to the symbol, represents the Euclidean distance between the i-th data vector and the k-th data vector;

[0017] Step S23: construct an entropy supplement term, which is expressed as follows:

[0018] ;

[0019] in, represents the complementary entropy term of the i-th data vector, d represents the dimension index of the data vector, and D represents the maximum dimension of the data vector. represents the value of the i-th data vector in the d-th dimension, Represents data value The probability of appearing in the dth dimension is, represents taking the logarithm, Indicates taking the absolute value, and Respectively represent the mean and variance of all data vectors in the dth dimension;

[0020] Step S24: Calculate the noise intensity, expressed as follows:

[0021] ;

[0022] in, represents the noise intensity of the ith data vector, represents the variance of the d-th dimension of the neighborhood data vector of the ith data vector, and represents the noise coordination weight;

[0023] Step S25: Data denoising, calculating the average noise intensity of all data vectors, setting the data vectors with noise intensity greater than 2 times the average noise intensity as noise vectors and removing them.

[0024] Furthermore, in step S3, the establishment of the active power outage monitoring model specifically includes the following steps:

[0025] Step S31: setting a model tag, and setting the power status data as model tag data;

[0026] Step S32: Design a smoothing layer, as shown below:

[0027] ;

[0028] in, represents the input value of the smoothing layer, Represents the output value of the smoothing layer, and Represent the weight and bias of the smoothing layer, represents the hyperbolic tangent function;

[0029] Step S33: construct a depth mapping function, which is expressed as follows:

[0030] ;

[0031] in, represents the input value of the depth mapping function, represents the depth mapping function, represents the logarithmic function with a natural constant as the base, Indicates square root;

[0032] Step S34: Design a deep feature learning layer, expressed as follows:

[0033] ;

[0034] in, represents the output value of the deep feature learning layer, represents the input value of the first layer of the deep feature learning layer, and denote the weights and biases of the first layer of the deep feature learning layer, respectively. represents the linear rectification function, and denote the weight and bias of the second layer of the deep feature learning layer, respectively. represents the normalized exponential function, and They represent the weights and biases of the third layer of the deep feature learning layer respectively;

[0035] Step S35: Dynamic deactivation processing, represented as follows:

[0036] ;

[0037] in, Represents the output value after dynamic deactivation processing, represents the input value before dynamic deactivation processing, represents the dynamic deactivation probability, Indicates is a Bernoulli distributed random variable with success probability, represents the insurance factor, represents the expansion factor, It means that the mean is 0 and the variance is Normally distributed random numbers;

[0038] Step S36: Design the compensation output layer, which is expressed as follows:

[0039] ;

[0040] in, represents the output value of the compensation output layer, and Represent the weight and bias of the compensated output layer, respectively. Represents the input value of the compensation output layer;

[0041] Step S37: construct a loss function, which is expressed as follows:

[0042] ;

[0043] in, represents the loss value of the model, g represents the index of the data used to train the model, and H represents the total number of data. represents the true value of the g-th data, represents the predicted value of the g-th data, and represents the loss weight, Represents the loss scale.

[0044] Furthermore, in step S4, the adaptive optimization specifically includes the following steps:

[0045] Step S41: Initialization, determining the optimization target, which includes: the weight, bias, dynamic deactivation probability, scaling factor and loss scale of the model, creating an optimization target space, setting the inverse of the loss value of the model as the parameter efficiency of the optimization parameter individual, randomly generating an initial search point cluster in the optimization target space, and calculating the parameter efficiency of the initial search point cluster;

[0046] Step S42: Design the performance coordination factor, which is expressed as follows:

[0047] ;

[0048] Where t represents the number of parameter searches, represents the efficiency coordination factor during the t-th parameter search, f represents the index of the number of searches, represents the average parameter effectiveness of all parameter search points during the f-th parameter search, represents the highest parameter efficiency value among all parameter search points during the fth parameter search. represents the difference between the average parameter performance of all parameter search points in the t-1th parameter search and the average parameter performance of all parameter search points in the t-2th parameter search;

[0049] Step S43: Design a direction conversion function, which is expressed as follows:

[0050] ;

[0051] in, represents the direction transformation function, Indicates the direction change value during the t-th parameter search. represents the sine function, Represents the angle between the current optimal parameter position and the worst parameter position, Represents a random number that follows a standard normal distribution;

[0052] Step S44: Design a noise reduction function, which is expressed as follows:

[0053] ;

[0054] in, represents the value of the noise reduction function, represents the value of the noise reduction function during the t-th parameter search, represents the standard deviation of parameter performance during the f-th parameter search, represents the noise reduction coefficient;

[0055] Step S45: Dynamic evolution guide search, expressed as follows:

[0056] ;

[0057] in, Indicates the parameter position obtained by dynamic evolution guided search during the t+1th parameter search, represents the parameter position obtained by dynamic evolution guided search during the tth parameter search, represents the symbolic function, represents the highest parameter efficiency value among all parameter search points during the tth parameter search, represents the average position of the initial parameter cluster;

[0058] Step S46: Adaptive follow-up search, expressed as follows:

[0059] ;

[0060] in, Indicates the parameter position obtained by adaptive following search during the t+1th parameter search, Indicates the parameter position obtained by adaptive following search during the tth parameter search, represents a random parameter point in the initial parameter cluster;

[0061] Step S47: Dynamic search, set twice the average parameter efficiency of the initial search point cluster as the efficiency threshold, first perform dynamic evolution guided search, then perform adaptive follow-up search, when the parameter efficiency is greater than the efficiency threshold, set it as the optimal model parameter; when the number of parameter searches is greater than the maximum number of parameter searches, search again; otherwise continue searching.

[0062] Furthermore, in step S5, the active power outage monitoring is performed by collecting real-time power parameter data, equipment operation data and environmental data, inputting the data into an active power outage monitoring model, and the model outputs the power status at that time. When the output power status is a power outage status, a power outage warning is issued.

[0063] The machine learning-based active power outage monitoring system provided by the present invention includes a data acquisition module, a data processing module, a power outage active monitoring model establishment module, an adaptive optimization module and a power outage active monitoring module;

[0064] The data acquisition module collects historical power parameter data, equipment operation data, environmental data and power status data, and sends the data to the data processing module;

[0065] The data processing module receives the data sent by the data acquisition module, removes the noise data by dividing the neighborhood, calculating the deviation term, constructing the entropy compensation term, and calculating the noise intensity, and sends the data to the power outage active monitoring model establishment module;

[0066] The power outage active monitoring model establishment module receives data sent by the data processing module, establishes a power outage active monitoring model, and sends the model data to the adaptive optimization module;

[0067] The adaptive optimization module receives data sent by the power outage active monitoring model establishment module, optimizes the parameters of the model, and sends the data to the power outage active monitoring module;

[0068] The power outage active monitoring module receives data sent by the adaptive optimization module, collects real-time power parameter data, equipment operation data and environmental data, and uses the power outage active monitoring model to monitor the power status in real time. When the output power status is a power outage state, a power outage warning is issued.

[0069] The beneficial effects achieved by the present invention using the above scheme are as follows:

[0070] (1) In order to address the problems of insufficient data processing accuracy and large noise interference in traditional active power outage monitoring methods, which limit monitoring accuracy, this scheme removes noise data by dividing the neighborhood, calculating deviation terms, constructing entropy compensation terms, and calculating noise intensity. This enhances the ability to identify data features, makes data processing more comprehensive and accurate, and improves the data quality and the accuracy of monitoring results.

[0071] (2) In order to solve the problems of difficulty in mining potential features in data and low prediction performance, accuracy and stability of traditional active power outage monitoring models, this scheme establishes an active power outage monitoring model by setting model labels, designing a smoothing processing layer, constructing a deep mapping function, designing a deep feature learning layer, dynamic deactivation processing, designing a compensation output layer and constructing a loss function. It can more deeply mine the potential features in the data, more comprehensively capture the key information in the data, and enhance the prediction performance, accuracy and stability of the model.

[0072] (3) In order to solve the problems of low search efficiency, easy to fall into local optimal solution and inflexible parameter adjustment in the optimization process of traditional power outage active monitoring model, this scheme guides the search process to evolve in a more efficient direction by designing efficiency coordination factors, direction transformation functions, noise reduction functions, dynamic evolution guided search, adaptive follow-up search and dynamic search, thus avoiding blind search, improving search efficiency, increasing diversity in the search process, and helping to escape from local optimal solution and explore global better solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0073] Figure 1 A schematic diagram of a method for active power outage monitoring based on machine learning provided by the present invention;

[0074] Figure 2 A schematic diagram of a power outage active monitoring system based on machine learning provided by the present invention;

[0075] Figure 3 A schematic diagram of data processing;

[0076] Figure 4 Schematic diagram of establishing an active monitoring model for power outages.

[0077] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention. DETAILED DESCRIPTION

[0078] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0079] In the description of the present invention, it should be understood that terms such as “upper”, “lower”, “front”, “back”, “left”, “right”, “top”, “bottom”, “inside” and “outside” indicating directions or positional relationships are based on the directions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific direction, be constructed and operated in a specific direction, and therefore should not be understood as limiting the present invention.

[0080] Example 1, see Figure 1 The present invention provides a method for active power outage monitoring based on machine learning, the method comprising the following steps:

[0081] Step S1: Data collection, collecting historical power parameter data, equipment operation data, environmental data and power status data, the power parameter data includes current, voltage, power, frequency and phase angle; the equipment operation data includes the temperature of the transformer and switch cabinet, the opening and closing status of the circuit breaker; the environmental data includes temperature, humidity, wind speed and precipitation; the power status data includes normal state and power outage state;

[0082] Step S2: Data processing, processing the data by neighborhood division, calculation of deviation term, construction of complementary entropy term, calculation of noise intensity and data denoising;

[0083] Step S3: Establishing a power outage active monitoring model, by setting model labels, designing a smoothing processing layer, constructing a deep mapping function, designing a deep feature learning layer, dynamic deactivation processing, designing a compensation output layer and constructing a loss function to establish a power outage active monitoring model;

[0084] Step S4: adaptive optimization, optimizing model parameters by designing efficiency coordination factors, designing direction transformation functions, designing noise reduction functions, dynamic evolution guided search, adaptive follow-up search and dynamic search;

[0085] Step S5: Active power outage monitoring, collecting real-time power parameter data, equipment operation data and environmental data, and using the active power outage monitoring model to monitor the power status in real time.

[0086] Example 2, see Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the data processing specifically includes the following steps:

[0087] Step S21: Neighborhood division, constructing data feature vectors, and setting the m data points closest to each data point as the neighborhood data set of the data point;

[0088] Step S22: Calculate the deviation term, which is expressed as follows:

[0089] ;

[0090] Where i represents the index of the data vector, represents the i-th data vector, represents the deviation term of the i-th data vector, n represents the total number of data vectors, j represents the index of the data vector, represents the jth data vector, represents the mean data vector of all data vectors, Indicates the Mahalanobis distance, e represents a natural constant, m represents the total number of data vectors in the neighborhood data set, and k represents the index of the data vector in the neighborhood data set. represents the neighborhood data set of the i-th data vector, Indicates that it belongs to the symbol, represents the Euclidean distance between the i-th data vector and the k-th data vector;

[0091] Step S23: construct an entropy supplement term, which is expressed as follows:

[0092] ;

[0093] in, represents the complementary entropy term of the i-th data vector, d represents the dimension index of the data vector, and D represents the maximum dimension of the data vector. represents the value of the i-th data vector in the d-th dimension, Represents data value The probability of appearing in the dth dimension is, represents taking the logarithm, Indicates taking the absolute value, and Respectively represent the mean and variance of all data vectors in the dth dimension;

[0094] Step S24: Calculate the noise intensity, expressed as follows:

[0095] ;

[0096] in, represents the noise intensity of the ith data vector, represents the variance of the d-th dimension of the neighborhood data vector of the ith data vector, and represents the noise coordination weight;

[0097] Step S25: Data denoising, calculating the average noise intensity of all data vectors, setting the data vectors with noise intensity greater than 2 times the average noise intensity as noise vectors and removing them.

[0098] By performing the above operations, in order to address the problems of insufficient data processing accuracy and large noise interference in traditional active power outage monitoring methods, which lead to limited monitoring accuracy, this solution removes noise data by dividing the neighborhood, calculating deviation terms, constructing entropy compensation terms, and calculating noise intensity, thereby enhancing the ability to recognize data features, making data processing more comprehensive and accurate, and improving the quality of data and the accuracy of monitoring results.

[0099] Example 3, see Figure 1 and Figure 4 This embodiment is based on the above embodiment, and the establishment of the active power outage monitoring model specifically includes the following steps:

[0100] Step S31: setting a model tag, and setting the power status data as model tag data;

[0101] Step S32: Design a smoothing layer, as shown below:

[0102] ;

[0103] in, represents the input value of the smoothing layer, Represents the output value of the smoothing layer, and Represent the weight and bias of the smoothing layer, represents the hyperbolic tangent function;

[0104] Step S33: construct a depth mapping function, which is expressed as follows:

[0105] ;

[0106] in, represents the input value of the depth mapping function, represents the depth mapping function, represents the logarithmic function with a natural constant as the base, Indicates square root;

[0107] Step S34: Design a deep feature learning layer, expressed as follows:

[0108] ;

[0109] in, represents the output value of the deep feature learning layer, represents the input value of the first layer of the deep feature learning layer, and denote the weights and biases of the first layer of the deep feature learning layer, respectively. represents the linear rectification function, and denote the weight and bias of the second layer of the deep feature learning layer, respectively. represents the normalized exponential function, and They represent the weights and biases of the third layer of the deep feature learning layer respectively;

[0110] Step S35: Dynamic deactivation processing, represented as follows:

[0111] ;

[0112] in, Represents the output value after dynamic deactivation processing, represents the input value before dynamic deactivation processing, represents the dynamic deactivation probability, Indicates is a Bernoulli distributed random variable with success probability, represents the insurance factor, represents the expansion factor, It means that the mean is 0 and the variance is Normally distributed random numbers;

[0113] Step S36: Design the compensation output layer, which is expressed as follows:

[0114] ;

[0115] in, represents the output value of the compensation output layer, and Represent the weight and bias of the compensated output layer, respectively. Represents the input value of the compensation output layer;

[0116] Step S37: construct a loss function, which is expressed as follows:

[0117] ;

[0118] in, represents the loss value of the model, g represents the index of the data used to train the model, and H represents the total number of data. represents the true value of the g-th data, represents the predicted value of the g-th data, and represents the loss weight, Represents the loss scale.

[0119] By performing the above operations, in order to address the problems of difficulty in mining potential features in data and low prediction performance, accuracy and stability of traditional active power outage monitoring models, this solution establishes an active power outage monitoring model by setting model labels, designing a smoothing processing layer, constructing a deep mapping function, designing a deep feature learning layer, dynamic deactivation processing, designing a compensation output layer and constructing a loss function. It can more deeply mine the potential features in the data, more comprehensively capture the key information in the data, and enhance the prediction performance, accuracy and stability of the model.

[0120] Example 4, see Figure 1 This embodiment is based on the above embodiment, and the adaptive optimization specifically includes the following steps:

[0121] Step S41: Initialization, determining the optimization target, which includes: the weight, bias, dynamic deactivation probability, scaling factor and loss scale of the model, creating an optimization target space, setting the inverse of the loss value of the model as the parameter efficiency of the optimization parameter individual, randomly generating an initial search point cluster in the optimization target space, and calculating the parameter efficiency of the initial search point cluster;

[0122] Step S42: Design the performance coordination factor, which is expressed as follows:

[0123] ;

[0124] Where t represents the number of parameter searches, represents the efficiency coordination factor during the t-th parameter search, f represents the index of the number of searches, represents the average parameter effectiveness of all parameter search points during the f-th parameter search, represents the highest parameter efficiency value among all parameter search points during the fth parameter search. represents the difference between the average parameter performance of all parameter search points in the t-1th parameter search and the average parameter performance of all parameter search points in the t-2th parameter search;

[0125] Step S43: Design a direction conversion function, which is expressed as follows:

[0126] ;

[0127] in, represents the direction transformation function, Indicates the direction change value during the t-th parameter search. represents the sine function, Represents the angle between the current optimal parameter position and the worst parameter position, Represents a random number that follows a standard normal distribution;

[0128] Step S44: Design a noise reduction function, which is expressed as follows:

[0129] ;

[0130] in, represents the value of the noise reduction function, represents the value of the noise reduction function during the t-th parameter search, represents the standard deviation of parameter performance during the f-th parameter search, represents the noise reduction coefficient;

[0131] Step S45: Dynamic evolution guide search, expressed as follows:

[0132] ;

[0133] in, Indicates the parameter position obtained by dynamic evolution guided search during the t+1th parameter search, represents the parameter position obtained by dynamic evolution guided search during the tth parameter search, represents the symbolic function, represents the highest parameter efficiency value among all parameter search points during the tth parameter search, represents the average position of the initial parameter cluster;

[0134] Step S46: Adaptive follow-up search, expressed as follows:

[0135] ;

[0136] in, Indicates the parameter position obtained by adaptive following search during the t+1th parameter search, Indicates the parameter position obtained by adaptive following search during the tth parameter search, represents a random parameter point in the initial parameter cluster;

[0137] Step S47: Dynamic search, set twice the average parameter efficiency of the initial search point cluster as the efficiency threshold, first perform dynamic evolution guided search, then perform adaptive follow-up search, when the parameter efficiency is greater than the efficiency threshold, set it as the optimal model parameter; when the number of parameter searches is greater than the maximum number of parameter searches, search again; otherwise continue searching.

[0138] By performing the above operations, in order to solve the problems of low search efficiency, easy to fall into local optimal solution and inflexible parameter adjustment in the optimization process of traditional power outage active monitoring model, this scheme guides the search process to evolve in a more efficient direction by designing efficiency coordination factors, direction transformation functions, noise reduction functions, dynamic evolution guided search, adaptive following search and dynamic search, avoiding blind search, thereby improving search efficiency, increasing diversity in the search process, and helping to jump out of local optimal solutions and explore better global solutions.

[0139] Example 5, see Figure 1 This embodiment is based on the above embodiment. The active power outage monitoring is achieved by collecting real-time power parameter data, equipment operation data and environmental data, inputting the data into the active power outage monitoring model, and the model outputs the power status at this time. When the output power status is a power outage status, a power outage warning is issued.

[0140] Example 6, see Figure 1 and Figure 2 , this embodiment is based on the above embodiment, and the machine learning-based active power outage monitoring system provided by the present invention includes a data acquisition module, a data processing module, a power outage active monitoring model establishment module, an adaptive optimization module and a power outage active monitoring module;

[0141] The data acquisition module collects historical power parameter data, equipment operation data, environmental data and power status data, and sends the data to the data processing module;

[0142] The data processing module receives the data sent by the data acquisition module, removes the noise data by dividing the neighborhood, calculating the deviation term, constructing the entropy compensation term, and calculating the noise intensity, and sends the data to the power outage active monitoring model establishment module;

[0143] The power outage active monitoring model establishment module receives data sent by the data processing module, establishes a power outage active monitoring model, and sends the model data to the adaptive optimization module;

[0144] The adaptive optimization module receives data sent by the power outage active monitoring model establishment module, optimizes the parameters of the model, and sends the data to the power outage active monitoring module;

[0145] The power outage active monitoring module receives data sent by the adaptive optimization module, collects real-time power parameter data, equipment operation data and environmental data, and uses the power outage active monitoring model to monitor the power status in real time. When the output power status is a power outage state, a power outage warning is issued.

[0146] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0147] While the embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that many changes, modifications, substitutions and alterations may be made to the embodiments without departing from the principles and spirit of the invention.

[0148] The present invention and its embodiments are described above, and such description is not restrictive. The drawings show only one embodiment of the present invention, and the actual structure is not limited thereto. In short, if ordinary technicians in the field are inspired by it, without departing from the purpose of the invention, they can design a structure and embodiment similar to the technical solution without creativity, which should belong to the protection scope of the present invention.

Claims

1. A method for active power outage monitoring based on machine learning, characterized in that: The method comprises the following steps: Step S1: Data collection, collecting historical power parameter data, equipment operation data, environmental data and power status data; Step S2: Data processing, processing the data by neighborhood division, calculation of deviation term, construction of complementary entropy term, calculation of noise intensity and data denoising; Among them, the calculation deviation term is expressed as follows: ； Where i represents the index of the data vector, represents the i-th data vector, represents the deviation term of the i-th data vector, n represents the total number of data vectors, j represents the index of the data vector, represents the jth data vector, represents the mean data vector of all data vectors, Indicates the Mahalanobis distance, e represents a natural constant, m represents the total number of data vectors in the neighborhood data set, and k represents the index of the data vector in the neighborhood data set. represents the neighborhood data set of the i-th data vector, Indicates that it belongs to the symbol, represents the Euclidean distance between the i-th data vector and the k-th data vector; Step S3: Establishing a power outage active monitoring model, by setting model labels, designing a smoothing processing layer, constructing a deep mapping function, designing a deep feature learning layer, dynamic deactivation processing, designing a compensation output layer and constructing a loss function to establish a power outage active monitoring model; Step S4: adaptive optimization, optimizing model parameters by designing efficiency coordination factors, designing direction transformation functions, designing noise reduction functions, dynamic evolution guided search, adaptive follow-up search and dynamic search; Step S5: Active power outage monitoring, collecting real-time power parameter data, equipment operation data and environmental data, and using the active power outage monitoring model to monitor the power status in real time.

2. The method for active power outage monitoring based on machine learning according to claim 1, characterized in that: In step S2, the data processing specifically includes the following steps: Step S21: Neighborhood division, constructing data feature vectors, and setting the m data points closest to each data point as the neighborhood data set of the data point; Step S22: Calculate the deviation term; Step S23: construct an entropy supplement term, which is expressed as follows: ； in, represents the complementary entropy term of the i-th data vector, d represents the dimension index of the data vector, and D represents the maximum dimension of the data vector. represents the value of the i-th data vector in the d-th dimension, Represents data value The probability of appearing in the dth dimension is, represents taking the logarithm, Indicates taking the absolute value, and Respectively represent the mean and variance of all data vectors in the dth dimension; Step S24: Calculate the noise intensity, expressed as follows: ； in, represents the noise intensity of the ith data vector, represents the variance of the d-th dimension of the neighborhood data vector of the ith data vector, and represents the noise coordination weight; Step S25: Data denoising, calculating the average noise intensity of all data vectors, setting the data vectors with noise intensity greater than 2 times the average noise intensity as noise vectors and removing them.

3. The method for active power outage monitoring based on machine learning according to claim 1, characterized in that: In step S3, the establishment of the active power outage monitoring model specifically includes the following steps: Step S31: setting a model tag, and setting the power status data as model tag data; Step S32: Design a smoothing layer, as shown below: ； in, represents the input value of the smoothing layer, Represents the output value of the smoothing layer, and Represent the weight and bias of the smoothing layer, represents the hyperbolic tangent function; Step S33: construct a depth mapping function, which is expressed as follows: ； in, Represents the input value of the depth mapping function, represents the depth mapping function, represents the logarithmic function with a natural constant as the base, Indicates square root; Step S34: Design a deep feature learning layer, expressed as follows: ； in, represents the output value of the deep feature learning layer, represents the input value of the first layer of the deep feature learning layer, and denote the weights and biases of the first layer of the deep feature learning layer, respectively. represents the linear rectification function, and denote the weight and bias of the second layer of the deep feature learning layer, respectively. represents the normalized exponential function, and They represent the weights and biases of the third layer of the deep feature learning layer respectively; Step S35: Dynamic deactivation processing, represented as follows: ； in, Represents the output value after dynamic deactivation processing, represents the input value before dynamic deactivation processing, represents the dynamic deactivation probability, Indicates is a Bernoulli distributed random variable with success probability, represents the insurance factor, represents the expansion factor, It means that the mean is 0 and the variance is Normally distributed random numbers; Step S36: Design the compensation output layer, which is expressed as follows: ； in, represents the output value of the compensation output layer, and denote the weight and bias of the compensated output layer, respectively. Represents the input value of the compensation output layer; Step S37: construct a loss function, which is expressed as follows: ； in, represents the loss value of the model, g represents the index of the data for training the model, and H represents the total number of data. represents the true value of the g-th data, represents the predicted value of the g-th data, and represents the loss weight, Represents the loss scale.

4. The method for active power outage monitoring based on machine learning according to claim 1, characterized in that: In step S4, the adaptive optimization specifically includes the following steps: Step S41: Initialization, determining the optimization target, which includes: the weight, bias, dynamic deactivation probability, scaling factor and loss scale of the model, creating an optimization target space, setting the inverse of the loss value of the model as the parameter efficiency of the optimization parameter individual, randomly generating an initial search point cluster in the optimization target space, and calculating the parameter efficiency of the initial search point cluster; Step S42: Design the performance coordination factor, which is expressed as follows: ； Where t represents the number of parameter searches, represents the efficiency coordination factor during the t-th parameter search, f represents the index of the number of searches, represents the average parameter effectiveness of all parameter search points during the f-th parameter search, represents the highest parameter efficiency value among all parameter search points during the fth parameter search. It represents the difference between the average parameter performance of all parameter search points in the t-1th parameter search and the average parameter performance of all parameter search points in the t-2th parameter search; Step S43: Design a direction conversion function, which is expressed as follows: ； in, represents the direction transformation function, Indicates the direction change value during the t-th parameter search. represents the sine function, Represents the angle between the current optimal parameter position and the worst parameter position, Represents a random number that follows a standard normal distribution; Step S44: Design a noise reduction function, which is expressed as follows: ； in, represents the value of the noise reduction function, represents the value of the noise reduction function during the t-th parameter search, represents the standard deviation of parameter performance during the f-th parameter search, represents the noise reduction coefficient; Step S45: Dynamic evolution guide search, expressed as follows: ； in, Indicates the parameter position obtained by dynamic evolution guided search during the t+1th parameter search, represents the parameter position obtained by dynamic evolution guided search during the tth parameter search, represents the symbolic function, represents the highest parameter efficiency value among all parameter search points during the tth parameter search, represents the average position of the initial parameter cluster; Step S46: Adaptive follow-up search, expressed as follows: ； in, Indicates the parameter position obtained by adaptive following search during the t+1th parameter search, Indicates the parameter position obtained by adaptive following search during the tth parameter search, represents a random parameter point in the initial parameter cluster; Step S47: Dynamic search, set twice the average parameter efficiency of the initial search point cluster as the efficiency threshold, first perform dynamic evolution guided search, then perform adaptive follow-up search, when the parameter efficiency is greater than the efficiency threshold, set it as the optimal model parameter; when the number of parameter searches is greater than the maximum number of parameter searches, search again; otherwise continue searching.

5. The method for active power outage monitoring based on machine learning according to claim 1, characterized in that: In step S1, the data collection is to collect historical power parameter data, equipment operation data, environmental data and power status data, the power parameter data includes current, voltage, power, frequency and phase angle; the equipment operation data includes the temperature of the transformer and switch cabinet, the opening and closing state of the circuit breaker; the environmental data includes temperature, humidity, wind speed and precipitation; the power status data includes normal state and power outage state; In step S5, the active power outage monitoring is performed by collecting real-time power parameter data, equipment operation data and environmental data, inputting the data into the active power outage monitoring model, and the model outputs the power status at this time. When the output power status is a power outage status, a power outage warning is issued.

6. A machine learning-based active power outage monitoring system, used to implement the machine learning-based active power outage monitoring method as described in any one of claims 1 to 5, characterized in that: It includes a data acquisition module, a data processing module, a power outage active monitoring model building module, an adaptive optimization module and a power outage active monitoring module; The data acquisition module collects historical power parameter data, equipment operation data, environmental data and power status data, and sends the data to the data processing module; The data processing module receives the data sent by the data acquisition module, removes the noise data by dividing the neighborhood, calculating the deviation term, constructing the entropy compensation term, and calculating the noise intensity, and sends the data to the power outage active monitoring model establishment module; The power outage active monitoring model establishment module receives data sent by the data processing module, establishes a power outage active monitoring model, and sends the model data to the adaptive optimization module; The adaptive optimization module receives data sent by the power outage active monitoring model establishment module, optimizes the parameters of the model, and sends the data to the power outage active monitoring module; The power outage active monitoring module receives data sent by the adaptive optimization module, collects real-time power parameter data, equipment operation data and environmental data, and uses the power outage active monitoring model to monitor the power status in real time. When the output power status is a power outage state, a power outage warning is issued.

Images