Machine learning based outage proactive monitoring method and system

Abstract

The application discloses a power outage active monitoring method and system based on machine learning, and the content comprises data acquisition, data processing, establishment of a power outage active monitoring model, adaptive optimization and power outage active monitoring. The application relates to the technical field of power outage monitoring, and particularly discloses a power outage active monitoring method and system based on machine learning. The scheme removes noise data by calculating a bias term, constructing a complementary entropy term and calculating noise intensity, thereby improving the quality of data and the accuracy of monitoring results. The scheme establishes a model by constructing a deep mapping function, designing a deep feature learning layer and constructing a loss function, thereby enhancing the prediction performance, precision and stability of the model. The scheme performs parameter search by designing an efficiency coordination factor, a direction transformation function and a noise reduction function, thereby guiding the search process to evolve in a more efficient direction, improving the search efficiency, increasing the diversity in the search process, helping to jump out of a local optimal solution and exploring a global optimal solution.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of power failure monitoring, in particular to a power failure active monitoring method and system based on machine learning. BACKGROUND

[0002] The power failure active monitoring method and system based on machine learning is an advanced system for power grid fault prediction by using machine learning technology, which can improve the accuracy and efficiency of power failure monitoring and provide more reliable power services for users.

[0003] The traditional power failure active monitoring method has the problems of insufficient data processing accuracy, large noise interference and limited monitoring accuracy; the traditional power failure active monitoring model has the problems of difficulty in mining potential features in data, low prediction performance, accuracy and stability of the model; and the optimization process of the traditional power failure active monitoring model has the problems of low search efficiency, easy falling into local optimal solution and inflexible parameter adjustment. SUMMARY

[0004] In view of the above problems, the power failure active monitoring method and system based on machine learning are provided to overcome the defects of the prior art, the power failure active monitoring method and system based on machine learning are provided, and the problems of insufficient data processing accuracy, large noise interference and limited monitoring accuracy of the traditional power failure active monitoring method are solved; the power failure active monitoring method and system based on machine learning are provided, and the problems of difficulty in mining potential features in data, low prediction performance, accuracy and stability of the model of the traditional power failure active monitoring model are solved; the power failure active monitoring method and system based on machine learning are provided, and the problems of low search efficiency, easy falling into local optimal solution and inflexible parameter adjustment in the optimization process of the traditional power failure active monitoring model are solved.

[0005] The technical solutions adopted by the application are as follows: the power failure active monitoring method based on machine learning comprises the following steps:

[0006] Step S1: data acquisition;

[0007] Step S2: data processing;

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

[0009] Step S4: adaptive optimization;

[0010] Step S5: power outage active monitoring.

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

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

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

[0014] Step S22: calculating bias term, represented as follows:

[0015] ;

[0016] Wherein, i represents the index of data vector, represents the i-th data vector, represents the bias term of the i-th data vector, n represents the total number of data vectors, j represents the index of data vector, represents the j-th data vector, represents the mean data vector of all data vectors, represents the Mahalanobis distance, e represents the natural constant, m represents the total number of data vectors in the neighborhood data set, k represents the index of data vector in the neighborhood data set, represents the neighborhood data set of the i-th data vector, represents the symbol belonging to, represents the Euclidean distance between the i-th data vector and the k-th data vector;

[0017] Step S23: constructing complementary entropy term, represented as follows:

[0018] ;

[0019] Wherein, represents the complement 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 the data value represents the probability of occurrence in the d-th dimension, represents taking the logarithm, represents taking the absolute value, and respectively represent the mean and variance of the d-th dimension of all data vectors;

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

[0021] ;

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

[0023] Step S25: Data denoising, calculate the noise intensity average of all data vectors, set the data vector with noise intensity greater than 2 times the noise intensity average as a noise vector and remove it.

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

[0025] Step S31: Set the model label, set the power state data as the model label data;

[0026] Step S32: Design a smoothing processing layer, represented as follows:

[0027] ;

[0028] wherein, represents the input value of the smoothing processing layer, represents the output value of the smoothing processing layer, and respectively represent the weight and bias of the smoothing processing layer, represents the hyperbolic tangent function;

[0029] Step S33: Construct a depth mapping function, represented as follows:

[0030] ;

[0031] wherein, represents the input value of the depth mapping function, denotes a depth mapping function, denotes a logarithm function with a natural constant as a base, denotes a square root;

[0032] Step S34: design a depth feature learning layer, denoted as follows:

[0033] ;

[0034] wherein, denotes an output value of the depth feature learning layer, denotes an input value of a first layer of the depth feature learning layer, and denote a weight and a bias of the first layer of the depth feature learning layer, respectively, denotes a linear rectification function, and denote a weight and a bias of a second layer of the depth feature learning layer, respectively, denotes a normalized exponential function, and denote a weight and a bias of a third layer of the depth feature learning layer, respectively;

[0035] Step S35: design a dynamic deactivation process, denoted as follows:

[0036] ;

[0037] wherein, denotes an output value after the dynamic deactivation process, denotes an input value before the dynamic deactivation process, denotes a dynamic deactivation probability, denotes a Bernoulli distribution random variable with as a success probability, denotes an insurance factor, denotes a scaling coefficient, denotes a random number subject to a normal distribution with a mean of 0 and a variance of ;

[0038] Step S36: design a compensation output layer, denoted as follows:

[0039] ;

[0040] wherein, denotes an output value of the compensation output layer, and denote a weight and a bias of the compensation output layer, respectively, denotes an input value of the compensation output layer;

[0041] Step S37: constructing a loss function, represented as follows:

[0042] ;

[0043] wherein, represents a loss value of the model, g represents an index of data for training the model, H represents a total number of data, represents a true value of the gth data, represents a predicted value of the gth data, and represents a loss weight, represents a loss scale.

[0044] Further, in step S4, the adaptive optimization specifically comprises the following steps:

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

[0046] Step S42: designing a performance coordination factor, represented as follows:

[0047] ;

[0048] wherein, t represents a parameter search number, represents a performance coordination factor at the tth parameter search, f represents an index of a search number, represents an average value of parameter performances of all parameter search points at the fth parameter search, represents a highest parameter performance value among all parameter search points at the fth parameter search, represents a difference value between an average value of parameter performances of all parameter search points at the t-1th parameter search and an average value of parameter performances of all parameter search points at the t-2th parameter search;

[0049] Step S43: designing a direction transformation function, represented as follows:

[0050] ;

[0051] wherein, represents a direction transformation function, represents a direction transformation value at the tth parameter search, represents a sine function, represents an included angle between a current optimal parameter position and a worst parameter position, represents a random number obeying a standard normal distribution;

[0052] Step S44: design a noise reduction function, denoted as follows:

[0053] ;

[0054] wherein, denotes a noise reduction function value, denotes a noise reduction function value in the tth parameter search, denotes a parameter performance standard deviation in the fth parameter search, denotes a noise reduction coefficient;

[0055] Step S45: dynamic evolution guided search, denoted as follows:

[0056] ;

[0057] wherein, denotes a parameter position obtained by dynamic evolution guided search in the t+1th parameter search, denotes a parameter position obtained by dynamic evolution guided search in the tth parameter search, denotes a sign function, denotes a highest parameter performance value in all parameter search points in the tth parameter search, denotes an average position of an initial parameter cluster;

[0058] Step S46: adaptive following search, denoted as follows:

[0059] ;

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

[0061] Step S47: dynamic search, set twice of an average value of parameter performance of an initial search point cluster as a performance threshold, first perform dynamic evolution guided search, then perform adaptive following search, when a parameter with a parameter performance greater than the performance threshold is searched, set it as an optimal model parameter; when a parameter search number is greater than a maximum parameter search number, re-perform search; otherwise, continue search.

[0062] Further, in step S5, the power failure active monitoring is performed by collecting real-time power parameter data, equipment operation data and environmental data, inputting the data into a power failure active monitoring model, outputting the power state at this time from the model, and issuing a power failure warning when the output power state is a power failure state.

[0063] The machine learning-based power failure active monitoring system provided by the application comprises a data collection module, a data processing module, a power failure active monitoring model establishment module, an adaptive optimization module and a power failure active monitoring module.

[0064] The data collection module collects historical power parameter data, equipment operation data, environmental data and power state data and sends the data to the data processing module.

[0065] The data processing module receives the data sent by the data collection module, removes noise data by neighborhood division, calculation of a bias term, construction of a complementary entropy term and calculation of noise intensity, and sends the data to the power failure active monitoring model establishment module.

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

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

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

[0069] The application has the following advantages:

[0070] (1) In view of the problem of insufficient data processing accuracy and large noise interference of the traditional power failure active monitoring method, which limits the monitoring accuracy, the present application removes noise data by neighborhood division, calculation of a bias term, construction of a complementary entropy term and calculation of noise intensity, enhances the recognition ability of data characteristics, makes the data processing more comprehensive and accurate, and improves the quality of data and the accuracy of monitoring results.

[0071] (2) In view of the problem that the traditional power outage active monitoring model cannot mine potential features in the data, and the prediction performance, precision and stability of the model are low, the model is established by setting the model label, designing the smoothing processing layer, constructing the deep mapping function, designing the deep feature learning layer, dynamic deactivation processing, designing the compensation output layer and constructing the loss function, so as to mine potential features in the data more deeply, capture key information in the data more comprehensively, and enhance the prediction performance, precision and stability of the model.

[0072] (3) In view of the problem that the search efficiency is low, the local optimal solution is easily trapped, and the parameter adjustment is not flexible in the optimization process of the traditional power outage active monitoring model, the search process is guided to evolve in a more efficient direction by designing the efficiency coordination factor, designing the direction transformation function, designing the noise reduction function, dynamic evolution guided search, adaptive following search and dynamic search, so as to improve the search efficiency, increase the diversity in the search process, help to jump out of the local optimal solution and explore the global optimal solution. BRIEF DESCRIPTION OF DRAWINGS

[0073] Figure 1 A schematic diagram of the power outage active monitoring method based on machine learning provided by the present application is shown in the figure.

[0074] Figure 2 A schematic diagram of the power outage active monitoring system based on machine learning provided by the present application is shown in the figure.

[0075] Figure 3 A schematic diagram of data processing is shown in the figure.

[0076] Figure 4 A schematic diagram of establishing the power outage active monitoring model is shown in the figure.

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

[0078] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0079] In the description of the present application, it should be understood that the terms "upper", "lower", "front", "back", "left", "right", "top", "bottom", "inner", "outer" and the like indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, and are only for the convenience of describing the present application and simplifying the description, and do not indicate or imply that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as a limitation on the present application.

[0080] Embodiment one, refer to Figure 1 The present application provides a power outage active monitoring method based on machine learning, which comprises the following steps:

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

[0082] Step S2: data processing, processing the data by neighborhood division, calculating bias term, constructing complementary entropy term, calculating noise intensity and data denoising;

[0083] Step S3: establishing a power outage active monitoring model, 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, designing an efficiency coordination factor, designing a direction transformation function, designing a noise reduction function, dynamic evolution guided search, adaptive following search and dynamic search to optimize model parameters;

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

[0086] Embodiment two, refer to Figure 1 and Figure 3 This embodiment is based on the above embodiment, and the data processing specifically comprises the following steps:

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

[0088] Step S22: calculating the bias term, which is expressed as follows:

[0089] ;

[0090] wherein i represents the index of the data vector, represents the i-th data vector, represents the bias term of the i-th data vector, n represents the total number of data vectors, and j represents the index of the data vector, represents the j-th data vector, represents the mean data vector of all data vectors, represents the Mahalanobis distance, e represents the natural constant, m represents the total number of data vectors of the neighborhood data set, and k represents the index of the data vector of the neighborhood data set, represents the neighborhood data set of the i-th data vector, represents belonging to the symbol, represents the Euclidean distance between the i-th data vector and the k-th data vector;

[0091] Step S23: Constructing the complementary entropy term, denoted as follows:

[0092] ;

[0093] wherein, 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 the data value the probability of occurrence in the d-th dimension, represents taking the logarithm, represents taking the absolute value, and respectively represent the mean and variance of all data vectors in the d-th dimension;

[0094] Step S24: Calculating the noise intensity, denoted as follows:

[0095] ;

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

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

[0098] By performing the above operation, for the problem of insufficient data processing accuracy, large noise interference and limited monitoring accuracy of the traditional power outage active monitoring method, the scheme removes noise data by neighborhood division, calculates the bias term, constructs the supplementary entropy term and calculates the noise intensity, enhances the recognition ability of data characteristics, makes the data processing more comprehensive and accurate, and improves the quality of data and the accuracy of monitoring results.

[0099] Embodiment three, refer to Figure 1 and Figure 4 , this embodiment is based on the above embodiment, and the establishment of the power outage active monitoring model specifically includes the following steps:

[0100] Step S31: setting model label, setting power state data as model label data;

[0101] Step S32: design a smoothing processing layer, which is represented as follows:

[0102] ;

[0103] Among them, represents the input value of the smoothing processing layer, represents the output value of the smoothing processing layer, and respectively represent the weight and bias of the smoothing processing layer, represents the hyperbolic tangent function;

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

[0105] ;

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

[0107] Step S34: design a deep feature learning layer, which is represented as follows:

[0108] ;

[0109] Among them, 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 respectively represent the weight and bias of the first layer of the deep feature learning layer, represents the linear rectification function, and respectively represent the weight and bias of the second layer of the deep feature learning layer, represents a normalized exponential function, and respectively represent the weight and bias of the third layer of the deep feature learning layer;

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

[0111] ;

[0112] wherein, represents the output value after the dynamic deactivation processing, represents the input value before the dynamic deactivation processing, represents the dynamic deactivation probability, represents a Bernoulli distribution random variable with as the success probability, represents an insurance factor, represents a scaling coefficient, represents a random number subject to a normal distribution with a mean of 0 and a variance of ;

[0113] Step S36: Designing a compensation output layer, represented as follows:

[0114] ;

[0115] wherein, represents the output value of the compensation output layer, and respectively represent the weight and bias of the compensation output layer, represents the input value of the compensation output layer;

[0116] Step S37: Constructing a loss function, represented as follows:

[0117] ;

[0118] wherein, represents the loss value of the model, g represents the index of the data for training the model, H represents the total number of data, represents the true value of the gth data, represents the predicted value of the gth data, and represent the loss weight, represents the loss scale.

[0119] Through the above operation, in order to solve the problem that it is difficult to mine potential features in the data for the traditional power outage active monitoring model, the prediction performance, accuracy and stability of the model are low, the power outage active monitoring model is established by setting the model label, designing the smoothing processing layer, constructing the deep mapping function, designing the deep feature learning layer, dynamic deactivation processing, designing the compensation output layer and constructing the loss function, the potential features in the data are mined more deeply, the key information in the data is captured more comprehensively, and the prediction performance, accuracy and stability of the model are enhanced.

[0120] Embodiment four, refer to Figure 1 , the embodiment is based on the above-mentioned embodiment, and the adaptive optimization specifically includes the following steps:

[0121] Step S41: initialization, determining an optimization target, the optimization target including: weights, biases, dynamic deactivation probability, stretching factor and loss scale of the model, creating an optimization target space, setting the reciprocal of the loss value of the model as the parameter performance of the optimization parameter individual, randomly generating an initial search point cluster in the optimization target space, and calculating the parameter performance of the initial search point cluster;

[0122] Step S42: design the performance coordination factor, as follows:

[0123] ;

[0124] Wherein, t represents the number of parameter searches, represents the performance coordination factor at the tth parameter search, f represents the index of the search number, represents the average value of the parameter performance of all parameter search points at the fth parameter search, represents the highest parameter performance value in all parameter search points at the fth parameter search, represents the difference between the average value of the parameter performance of all parameter search points at the t-1th parameter search and the average value of the parameter performance of all parameter search points at the t-2th parameter search;

[0125] Step S43: design the direction transformation function, as follows:

[0126] ;

[0127] Wherein, represents the direction transformation function, represents the direction transformation value at the tth parameter search, represents the sine function, represents the included angle between the current optimal parameter position and the worst parameter position, represents a random number subject to standard normal distribution;

[0128] Step S44: design the noise reduction function, denoted as follows:

[0129]

[0130] wherein, denotes the noise reduction function value, denotes the noise reduction function value at the tth parameter search, denotes the parameter performance standard deviation at the fth parameter search, denotes the noise reduction coefficient;

[0131] Step S45: dynamic evolution guided search, denoted as follows:

[0132]

[0133] wherein, denotes the parameter position obtained by dynamic evolution guided search at the t+1th parameter search, denotes the parameter position obtained by dynamic evolution guided search at the tth parameter search, denotes the sign function, denotes the highest parameter performance value among all parameter search points at the tth parameter search, denotes the average position of the initial parameter cluster;

[0134] Step S46: adaptive following search, denoted as follows:

[0135]

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

[0137] Step S47: dynamic search, set twice the average value of the parameter performance of the initial search point cluster as the performance threshold, first perform dynamic evolution guided search, then perform adaptive following search, when a parameter with a performance greater than the performance threshold is searched, set it as the optimal model parameter; when the number of parameter searches is greater than the maximum number of parameter searches, re-perform the search; otherwise, continue the search.

[0138] ​​​By performing the above operation, in order to solve the problems of low search efficiency, easy to fall into local optimal solution and not flexible parameter adjustment in the optimization process of the traditional power outage active monitoring model, the scheme guides the search process to evolve in a more efficient direction by designing the performance coordination factor, designing the direction transformation function, designing the noise reduction function, dynamic evolution guided search, adaptive following search and dynamic search, avoids blind search, thereby improving the search efficiency, increasing the diversity in the search process, helping to jump out of the local optimal solution and exploring the global optimal solution.

[0139] Embodiment five, refer to Figure 1 This embodiment is based on the above-mentioned embodiment, and the power outage active monitoring is performed by collecting real-time power parameter data, equipment operation data and environmental data, inputting the data into the power outage active monitoring model, and outputting the power state at this time. When the output power state is a power outage state, a power outage warning is issued.

[0140] Embodiment six, refer to Figure 1 and Figure 2 This embodiment is based on the above-mentioned embodiment, and the power outage active monitoring system based on machine learning provided by the application comprises a data acquisition module, a data processing module, a power outage active monitoring model establishing module, an adaptive optimization module and a power outage active monitoring module.

[0141] The data acquisition module acquires historical power parameter data, equipment operation data, environmental data and power state 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 noise data by neighborhood division, calculation of bias terms, construction of complementary entropy terms and calculation of noise intensity, and sends the data to the power outage active monitoring model establishing module.

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

[0144] The adaptive optimization module receives the data sent by the power outage active monitoring model establishing 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 the data sent by the adaptive optimization module, acquires real-time power parameter data, equipment operation data and environmental data, uses the power outage active monitoring model to monitor the power state in real time, and issues a power outage warning when the output power state is a power outage state.

[0146] It is to be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting; it is not intended to exclude myriad other embodiments of the present application that other inventors can develop based on the same general inventive concepts embodied by the described embodiments. That is, although the present application is described in terms of particular embodiments and implementations, it is to be understood that the terminology used is for the purpose of descriptive clarity and that it is intended to be limited only by the claims.

[0147] While the embodiments of the application have been shown and described, it is to be understood that the embodiments described are only by way of example and that modifications, changes, substitutions and variations can be made by those skilled in the art without departing from the spirit and scope of the application.

[0148] The above description of the present application has been presented for the purposes of illustration and description. It is not intended to be exhaustive or to limit the application to the precise form described. Many modifications, variations, alterations, and equivalents can be suggested to one skilled in the art in light of this disclosure and it is intended that the protected application embrace all such alternatives, modifications, and variations.

Claims

1. A machine learning-based active power outage monitoring method, characterized in that, The method includes the following steps: Step S1: Data acquisition, collecting historical power parameter data, equipment operation data, environmental data, and power status data; Step S2: Data processing, which involves processing the data through neighborhood partitioning, calculating the deviation term, constructing the entropy compensation term, calculating the noise intensity, and data denoising. Step S21: Neighborhood partitioning, constructing data feature vectors, and setting the m nearest data points to each data point as the neighborhood data set of the data point; Step S22: Calculate the deviation term, as shown below: ； Where i represents the index of the data vector, This represents the i-th data vector. Let represent the deviation term of the i-th data vector, n represent the total number of data vectors, and j represent the index of the data vector. This represents the j-th data vector. The data vector representing the mean of all data vectors. This indicates the use of Mahalanobis distance, where e represents the natural constant, m represents the total number of data vectors in the neighborhood dataset, and k represents the index of the data vector in the neighborhood dataset. This represents the neighborhood data set of the i-th data vector. Indicates that it belongs to the symbol. This represents the Euclidean distance between the i-th data vector and the k-th data vector; Step S23: Construct the entropy complement term, represented as follows: ； in, Let represent the entropy complement term of the i-th data vector, d represent the dimension index of the data vector, and D represent the maximum dimension of the data vector. This represents the value of the i-th data vector in the d-th dimension. Represents data value The probability of it appearing in the d-th dimension. Indicates taking the logarithm. This indicates taking the absolute value. and Let represent the mean and variance of all data vectors in the d-th dimension, respectively. Step S24: Calculate the noise intensity, as shown below: ； in, This represents the noise intensity of the i-th data vector. Let represent the variance of the d-th dimension of the neighboring data vectors of the i-th data vector. and Indicates noise coordination weights; Step S25: Data denoising, calculate the average noise intensity of all data vectors, and set the data vectors with noise intensity greater than twice the average noise intensity as noise vectors and remove them; Step S3: Establish an active power outage monitoring model by setting model labels, designing a smoothing layer, constructing a deep mapping function, designing a deep feature learning layer, implementing dynamic deactivation processing, designing a compensation output layer, and constructing a loss function. Step S4: Adaptive optimization, which optimizes model parameters by designing performance coordination factors, designing direction transformation functions, designing noise reduction functions, dynamic evolutionary guided search, adaptive follow-up search, and dynamic search; Step S5: Active power outage monitoring. Collect real-time power parameter data, equipment operation data, and environmental data, and use the active power outage monitoring model to monitor the power status in real time.

2. The machine learning-based active power outage monitoring method according to claim 1, characterized in that: In step S3, establishing the active power outage monitoring model specifically includes the following steps: Step S31: Set model labels, and set the power status data as model label data; Step S32: Design a smoothing layer, as shown below: ； in, This represents the input value of the smoothing layer. This represents the output value of the smoothing layer. and These represent the weights and biases of the smoothing layer, respectively. Represents the hyperbolic tangent function; Step S33: Construct the depth mapping function, as follows: ； in, This represents the input value of the depth mapping function. Represents the depth mapping function. This represents a logarithmic function with the natural constant as its base. To represent the square root; Step S34: Design a deep feature learning layer, as shown below: ； in, This represents the output value of the deep feature learning layer. This represents the input value of the first layer of the deep feature learning layer. and These represent the weights and biases of the first layer of the deep feature learning layer, respectively. Represents the linear rectified function. and These represent the weights and biases of the second layer in the deep feature learning layer, respectively. Represents the normalized exponential function, and These represent the weights and biases of the third layer in the deep feature learning layer, respectively. Step S35: Dynamic deactivation process, as shown below: ； in, This represents the output value after dynamic deactivation processing. This represents the input value before dynamic deactivation. Indicates the dynamic inactivation probability. Indicated by Let the probability of success be a Bernoulli distributed random variable. Indicates insurance factor, Indicates the scaling factor. This represents a sequence of elements that follow a mean of 0 and a variance of 0. Random numbers that follow a normal distribution; Step S36: Design the compensation output layer, as shown below: ； in, This indicates the output value of the compensation output layer. and These represent the weights and biases of the compensated output layer, respectively. This represents the input value of the compensation output layer; Step S37: Construct the loss function, as follows: ； in, Here, g 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 points. This represents the true value of the g-th data point. This represents the predicted value of the g-th data point. and Indicates the loss weight. Indicates the scale of loss.

3. The machine learning-based active power outage monitoring method according to claim 1, characterized in that: In step S4, the adaptive optimization specifically includes the following steps: Step S41: Initialization, determine the optimization objective, which includes: model weights, biases, dynamic inactivation probability, scaling factor and loss scale, create optimization objective space, set the reciprocal of the model's loss value as the parameter performance of individual optimization parameters, randomly generate an initial search point cluster in the optimization objective space, and calculate the parameter performance of the initial search point cluster. Step S42: Design the performance coordination factor, expressed as follows: ； Where t represents the number of parameter searches, Let f represent the performance coordination factor at the t-th parameter search, and let f represent the index of the search count. This represents the average parameter performance across all parameter search points during the f-th parameter search. This represents the highest parameter performance value among all parameter search points during the f-th parameter search. This represents the difference between the average parameter efficiency of all parameter search points during the (t-1)th parameter search and the average parameter efficiency of all parameter search points during the (t-2)th parameter search. Step S43: Design the direction transformation function, as follows: ； in, Represents the direction transformation function. This represents the direction change value during the t-th parameter search. Represents the sine function. This represents the angle between the current optimal parameter position and the worst parameter position. This represents a random number that follows a standard normal distribution. Step S44: Design the noise reduction function, as follows: ； in, This represents the noise reduction function value. This represents the noise reduction function value during the t-th parameter search. This represents the standard deviation of parameter performance during the f-th parameter search. Indicates the noise reduction coefficient; Step S45: Dynamic evolution-guided search, represented as follows: ； in, This indicates the position of the parameter obtained by the dynamic evolution-guided search during the (t+1)th parameter search. This indicates the position of the parameter obtained during the t-th parameter search using dynamic evolution-guided search. Represents a symbolic function. This represents the highest parameter performance value among all parameter search points at the t-th parameter search. This represents the average position of the initial parameter cluster; Step S46: Adaptive following search, represented as follows: ； in, This represents the parameter position obtained by adaptive following search during the (t+1)th parameter search. This represents the position of the parameter obtained by adaptive following search during the t-th parameter search. This 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 a parameter with efficiency greater than the efficiency threshold is found, set it as the optimal model parameter. If the number of parameter searches exceeds the maximum number of parameter searches, start the search again; otherwise, continue the search.

4. The machine learning-based active power outage monitoring method according to claim 1, characterized in that: In step S1, the data acquisition involves 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 transformers and switchgear, and the opening and closing status of circuit breakers. The environmental data includes temperature, humidity, wind speed, and precipitation. The power status data includes normal status and power outage status. In step S5, the active power outage monitoring involves 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 outputting the current power status. When the output power status is a power outage status, a power outage warning is issued.

5. 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 1-4, characterized in that: It includes a data acquisition module, a data processing module, a module for establishing an active power outage monitoring model, an adaptive optimization module, and an active power outage 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 data sent by the data acquisition module, removes noise data by dividing the neighborhood, calculating the deviation term, constructing the entropy compensation term, and calculating the noise intensity, and then sends the data to the power outage active monitoring model building module. The module for establishing an active power outage monitoring model receives data sent by the data processing module, establishes an active power outage 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 module, optimizes the model parameters, and sends the data back to the power outage active monitoring module. The active power outage 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 active power outage monitoring model to monitor the power status in real time. When the output power status is a power outage status, it issues a power outage warning.

Images