Risk situation prediction method for low-voltage power consumption area and related device

Abstract

The application provides a risk situation prediction method for a low-voltage power consumption area and a related device. The method comprises the following steps: obtaining historical data of a low-voltage power consumption area; grouping the historical data according to a first grouping rule to obtain a plurality of first training data groups; grouping the plurality of first training data groups according to a second grouping rule to obtain a plurality of first training subsets; preprocessing data in the plurality of first training subsets to obtain a target data set conforming to a data model input rule; dividing the target data set into a target training data set and a target test set; constructing a training model, training the training model according to the target training data set, and obtaining a trained risk situation prediction model; inputting real-time operation data and real-time environment data of a first time interval of the low-voltage power consumption area into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption area.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of Internet, and particularly relates to a risk situation prediction method for a low-voltage power consumption area and a related device. BACKGROUND

[0002] At present, in the risk prediction of a low-voltage power consumption area, the existing technology generally adopts offline analysis, that is, the data of the low-voltage power consumption area are collected, and the collected data are transferred to an offline platform, and the risk of the low-voltage power consumption area is analyzed on the offline platform.

[0003] It can be seen that the method of the existing technology cannot dynamically monitor and predict the real-time running state of the low-voltage power consumption area. SUMMARY

[0004] The application provides a risk situation prediction method for a low-voltage power consumption area and a related device, so as to realize real-time monitoring and prediction of the risk situation of the low-voltage power consumption area.

[0005] In a first aspect, the application provides a risk situation prediction method for a low-voltage power consumption area, comprising:

[0006] acquiring historical data of a low-voltage power consumption area; wherein the historical data comprises a plurality of running data, a plurality of environmental data and a plurality of risk data;

[0007] grouping the historical data according to a first grouping rule to obtain a plurality of first training data groups; the first grouping rule refers to dividing the running data, the environmental data and the risk data at the same time into a group; each first training data group comprises a plurality of input parameters and output parameters, wherein the input parameters comprise the running data and the environmental data, and the output parameters comprise the risk data;

[0008] grouping the plurality of first training data groups according to a second grouping rule to obtain a plurality of first training subsets; the second grouping rule refers to dividing the first training data groups in the same time interval into the same first training subset;

[0009] preprocessing the data in the plurality of first training subsets to obtain a target data set conforming to a data model input rule;

[0010] dividing the target data set into a target training data set and a target test set;

[0011] constructing a training model, training the training model according to the target training data set and the target test set, and obtaining a trained risk situation prediction model;

[0012] Input real-time operation data and real-time environment data of a first time interval of the low-voltage power consumption station area into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption station area; the second time interval is a next time interval adjacent to the first time interval.

[0013] With reference to the first aspect, in a possible implementation, the data in the plurality of first training subsets is preprocessed to obtain a target training data set conforming to data model input rules, including: performing data completion and normalization processing on the first training data groups in each first training subset respectively to obtain a plurality of second training subsets; each second training subset includes a plurality of second training data groups after data completion and normalization processing; performing feature extraction on the second training data groups in each second training subset to obtain the target training data set; the target training data set includes a plurality of third training subsets, and each third training subset includes a plurality of third training data groups obtained after feature extraction.

[0014] With reference to the first aspect, in a possible implementation, the data completion and normalization processing on the first training data groups in each first training subset to obtain the plurality of second training subsets includes: determining positions of data missing in each first training data group; performing data completion on the positions of data missing by an interpolation method; and scaling the data of each first training data group after data completion to a target range to obtain a plurality of second training data groups.

[0015] With reference to the first aspect, in a possible implementation, the feature extraction on the second training data groups in each second training subset to obtain the target training data set includes: calculating a current fluctuation rate, a voltage deviation rate, a maximum average load rate, and a maximum distribution transformer average three-phase unbalance degree of each second training subset; and determining a maximum environmental temperature and a maximum environmental humidity of each second training subset to obtain the target training data set.

[0016] With reference to the first aspect, in a possible implementation, the constructing a training model includes: constructing the training model based on a radial basis neural network; the training model includes an input layer, a processing layer, and an output layer; the input layer is configured to receive a third training subset in the target training data set; the processing layer is configured to map the input third training subset to a high-dimensional space to obtain a feature label matrix; and the output layer is configured to process the feature label matrix to obtain the risk type of the next time interval of the low-voltage power consumption station area.

[0017] With reference to the first aspect, in a possible implementation, the training of the to-be-trained model according to the target to-be-trained data set and the target test set to obtain the trained risk situation prediction model comprises: inputting each third training subset in the target to-be-trained data set into the to-be-trained model in sequence; processing, by the to-be-trained model, data in the input third training subset to obtain a first risk type; comparing the first risk type with a target risk type in the target test set; if the comparison is accurate, completing the training of the to-be-trained model to obtain the risk situation prediction model; if the comparison is not accurate, optimizing the to-be-trained model by using a Bayesian optimization method, continuing to train the optimized to-be-trained model by using the target to-be-trained data set, and repeating the above operations until the data result of the model is accurate, so as to obtain the risk situation prediction model.

[0018] With reference to the first aspect, in a possible implementation, the inputting of the real-time operation data and the real-time environment data of the low-voltage power consumption area into the risk situation prediction model to obtain a risk type of a next time interval of the low-voltage power consumption area comprises: collecting real-time operation data and real-time environment data of the low-voltage power consumption area in a first time window; the real-time operation data comprises an operating voltage, an operating current, an average load rate, and an average three-phase unbalance degree of a power distribution transformer; the real-time environment data comprises an environmental temperature and an environmental humidity; preprocessing the target data to obtain to-be-processed data that meets an input rule of the risk situation prediction model; determining, by using the risk situation prediction model, a target risk type of the low-voltage power consumption area in a second time window according to the to-be-processed data; the risk situation prediction model is trained based on a target training set for a plurality of continuous time windows, the time length of each time window is consistent, the plurality of time windows comprise the first time window and the second time window, and the second time window is a next time window of the first time window; and outputting target early warning information according to the target risk type.

[0019] In a second aspect, the present application provides a risk situation prediction device for a low-voltage power consumption area, comprising:

[0020] The acquisition unit is configured to acquire historical data of the low-voltage power consumption area; the historical data comprises a plurality of operation data, a plurality of environment data, and a plurality of risk data.

[0021] The grouping unit is configured to group the historical data according to a first grouping rule to obtain a plurality of first training data groups; the first grouping rule refers to grouping operation data, environment data, and risk data at the same time into a group; each first training data group comprises a plurality of input parameters and output parameters, wherein the input parameters comprise the operation data and the environment data, and the output parameters comprise the risk data.

[0022] The grouping unit is further configured to group the plurality of first training data sets according to a second grouping rule to obtain a plurality of first training subsets; the second grouping rule refers to dividing the first training data sets in a same time interval into a same first training subset;

[0023] The preprocessing unit is configured to preprocess data in the plurality of first training subsets to obtain a target data set conforming to a data model input rule, and divide the target data set into a target training data set and a target test set;

[0024] The model construction unit is configured to construct a training model, train the training model according to the target training data set and the target test set, and obtain a trained risk situation prediction model.

[0025] The prediction unit is configured to input real-time operation data and real-time environment data of a first time interval of the low-voltage power consumption area into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption area; the second time interval is a next time interval adjacent to the first time interval.

[0026] In a third aspect, the present application provides an electronic device, including a processor, a memory, a communication interface, and one or more programs, the one or more programs are stored in the memory and configured to be executed by the processor, and the program includes instructions for executing the steps of any one of the first aspect and the second aspect of the present application.

[0027] In a fourth aspect, the present application provides a computer readable storage medium, which stores a computer program for electronic data exchange, wherein the computer program causes a computer to execute some or all steps described in any one of the first aspect and the second aspect of the present application.

[0028] In a fifth aspect, the present application provides a computer program product, wherein the above computer program product includes a non-transitory computer readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute some or all steps described in any one of the first aspect and the second aspect of the present application. The computer program product can be a software installation package.

[0029] It can be seen that, in the present application, the historical data of the low-voltage power consumption area is acquired; then the historical data is grouped according to a first grouping rule to obtain a plurality of first training data groups; then the plurality of first training data groups are grouped according to a second grouping rule to obtain a plurality of first training subsets; then the data in the plurality of first training subsets is preprocessed to obtain a target data set conforming to a data model input rule; the target data set is divided into a target training data set and a target test set; then a training model is constructed, the training model is trained according to the target training data set, and a trained risk situation prediction model is obtained; finally, the real-time running data and real-time environmental data of the first time interval of the low-voltage power consumption area are input into the risk situation prediction model, and the risk type of the second time interval of the low-voltage power consumption area is obtained. In this way, the real-time running data and real-time environmental data of the low-voltage power consumption area are collected in real time, input into the trained risk situation prediction model, and the risk type of the next time interval of the low-voltage power consumption area is predicted, realizing real-time monitoring and prediction of the risk situation of the low-voltage power consumption area. BRIEF DESCRIPTION OF DRAWINGS

[0030] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0031] Figure 1 is a structural schematic diagram of a first risk situation prediction system provided by an embodiment of the present application;

[0032] Figure 2 is a structural schematic diagram of a second risk situation prediction system provided by an embodiment of the present application;

[0033] Figure 3 is a flow schematic diagram of a risk situation prediction method for a low-voltage power consumption area provided by an embodiment of the present application;

[0034] Figure 4 is a schematic diagram of a risk situation prediction model and a training process provided by an embodiment of the present application;

[0035] Figure 5 is a structural schematic diagram of a risk situation prediction device for a low-voltage power consumption area provided by an embodiment of the present application;

[0036] Figure 6 is a structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0037] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative effort belong to the scope of protection of the present application.

[0038] The terms "first", "second", etc. in the specification and claims of the present application and the above drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but can optionally include steps or units not listed, or can optionally include other steps or units inherent to the process, method, system, product or device.

[0039] Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the present application. The appearance of the phrase in various places in the specification does not necessarily all refer to the same embodiment, nor is it necessarily mutually exclusive of other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0040] The related terms involved in the present application will be introduced first.

[0041] The RBF model, i.e. the Radial Basis Function model, is a commonly used neural network model. The RBF network is a feedforward neural network composed of an input layer, a hidden layer and an output layer. The basic idea is to map the input data to a high-dimensional feature space, and then perform linear combination in this space to realize the non-linear modeling of the input data. The neurons in the hidden layer use radial basis functions as activation functions, and common radial basis functions include Gaussian functions, polynomial functions, etc. These functions take a point in the input space as the center, and decrease monotonically with the increase of the distance from the center, so as to form a localized response region in the input space.

[0042] The low-voltage electric power station area refers to the power supply area from the secondary side of the distribution transformer to the user electric meter box. It includes distribution transformers, low-voltage distribution cabinets, metering devices, low-voltage lines and user terminal equipment, etc., and is the last link of power supply, directly providing power for various users.

[0043] At present, the prior art generally adopts offline analysis in the risk prediction of low-voltage power consumption area, that is, the data of the low-voltage power consumption area is collected, and the collected data is transferred to an offline platform, and the risk of the low-voltage power consumption area is analyzed on the offline platform. It can be seen that the method of the prior art cannot dynamically monitor and predict the real-time running state of the low-voltage power consumption area.

[0044] To solve the above problems, the embodiments of the present application provide a risk situation prediction method for a low-voltage power consumption area. The risk situation prediction method for the low-voltage power consumption area can be applied to the scene of predicting the risk of the low-voltage power consumption area. The historical data of the low-voltage power consumption area can be obtained, and then the historical data is grouped according to a first grouping rule to obtain a plurality of first training data groups. Then, the plurality of first training data groups are grouped according to a second grouping rule to obtain a plurality of first training subsets. Then, the data in the plurality of first training subsets is preprocessed to obtain a target data set that meets the data model input rule. The target data set is divided into a target training data set and a target test set. Then, a training model is constructed, the training model is trained according to the target training data set, and a trained risk situation prediction model is obtained. Finally, the real-time running data and real-time environmental data of the low-voltage power consumption area in a first time interval are input into the risk situation prediction model, and the risk type of the low-voltage power consumption area in a second time interval is obtained. In this way, the real-time running data and real-time environmental data of the low-voltage power consumption area are collected in real time, input into the trained risk situation prediction model, and the risk type of the low-voltage power consumption area in the next time interval is predicted, realizing real-time monitoring and prediction of the risk situation of the low-voltage power consumption area. The present scheme can be applied to various scenes, including but not limited to the application scenes mentioned above.

[0045] The system architecture related to the embodiments of the present application will be introduced below.

[0046] Please refer to Figure 1The application provides a risk situation prediction system 20, which comprises a data acquisition module 21, a data processing module 22, a model training and optimization module 23, a model construction module 24 and a risk situation prediction module 25. The data acquisition module 21 is used for acquiring historical data of a low-voltage power consumption area, and the data preprocessing module is used for grouping the historical data according to a first grouping rule to obtain a plurality of first training data groups; the first grouping rule refers to that the operation data, the environmental data and the risk data at the same time are divided into a group; each first training data group comprises a plurality of input parameters and output parameters, wherein the input parameters comprise the operation data and the environmental data, and the output parameters comprise the risk data; the plurality of first training data groups are grouped according to a second grouping rule to obtain a plurality of first training subsets; the second grouping rule refers to that the first training data groups in the same time interval are divided into the same first training subset; the data processing module 22 is used for preprocessing the data in the plurality of first training subsets to obtain a target data set meeting a data model input rule; the target data set is divided into a target training data set and a target test set; the model construction module 24 is used for constructing a training model, the model training and optimization module 23 is used for training the training model according to the target training data set and the target test set to obtain a trained risk situation prediction model; and the risk situation prediction module 25 is used for inputting real-time operation data and real-time environmental data of a first time interval of the low-voltage power consumption area into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption area.

[0047] Further, please refer to Figure 2 The data processing module 22 comprises a data preprocessing module 221 and a feature extraction module 222, wherein the data preprocessing module 221 is used for performing data completion and normalization processing on each first training data group in each first training subset to obtain a plurality of second training subsets; each second training subset comprises a plurality of second training data groups after data completion and normalization processing; and the feature extraction module 222 is used for performing feature extraction on the second training data groups in each second training subset to obtain the target training data set; the target training data set comprises a plurality of third training subsets, and each third training subset comprises a plurality of third training data groups after feature extraction.

[0048] In this way, the real-time operation data and the real-time environmental data of the low-voltage power consumption area are collected in real time, input into the trained risk situation prediction model, the risk type of the next time interval of the low-voltage power consumption area is predicted, and the real-time monitoring and prediction of the risk situation of the low-voltage power consumption area are realized.

[0049] The specific method is described in detail below.

[0050] Referring to Figure 3 and Figure 4 The application also provides a risk situation prediction method for a low-voltage power consumption station area, which can be applied to a risk situation prediction system as shown in Figure 3 The risk situation prediction method for the low-voltage power consumption station area comprises the following steps.

[0051] Step S201: Obtain historical data of a low-voltage power consumption station area.

[0052] The historical data comprises multiple operation data, multiple environment data and multiple risk data.

[0053] Specifically, in this embodiment, the operation data, the environment data and the risk data of the low-voltage power consumption station area can be monitored in real time, and the monitored operation data, environment data and risk data can be stored. The low-voltage power consumption monitoring system can store historical data of the low-voltage power consumption station area for a certain length of time.

[0054] The historical operation data, the environment data and the risk data of the low-voltage power consumption station area are collected from the low-voltage power consumption monitoring system. The operation data comprises the operation voltage U0(t), the operation current I0(t), the average load rate P0(t) of the distribution transformer and the average three-phase unbalance degree S0(t) of the distribution transformer. The environment data comprises the environment temperature T0(t) and the environment humidity W0(t) of the low-voltage power consumption station area. The risk data comprises information data of whether a fault occurs at a corresponding time, i.e., the risk type label b(t) of the low-voltage power consumption station area at time t. When b(t) is 0, it indicates normal; when b(t) is 1, it indicates overload; when b(t) is 2, it indicates voltage out-of-limit; when b(t) is 3, it indicates short circuit; and when b(t) is 4, it indicates equipment failure.

[0055]

[0056] Step S202: Group the historical data according to a first grouping rule to obtain multiple first training data groups.

[0057] The first grouping rule refers to grouping the operation data, the environment data and the risk data at the same time into a group. Each first training data group comprises multiple input parameters and output parameters. The input parameters comprise the operation data and the environment data, and the output parameters comprise the risk data.

[0058] Specifically, when storing historical data, a time stamp is associated with the historical data to identify the time when the data is generated. The running data, the environmental data and the risk data are grouped based on the time stamp to obtain a first training data group corresponding to each time. Each first training data group is a training data, and the running data and the environmental data in each first training data group are used as input parameters during training, and the risk data is used as an output parameter during training, which is used to verify the training result.

[0059] In step S203, the plurality of first training data groups are grouped according to a second grouping rule to obtain a plurality of first training subsets.

[0060] The second grouping rule refers to dividing the first training data groups in the same time interval into the same first training subset.

[0061] Specifically, the total length of the historical data collected for participating in model training is T, and the time length T is divided into K windows (1, 2, …, k, …, K) according to the time length. The length of each window is n. That is, the time length T is divided into a plurality of time intervals, and each time scale in each time interval corresponds to a first training data group. Each time interval includes a plurality of first training data groups, forming a first training subset. Each first training subset can train the to-be-trained model once.

[0062] In step S204, the data in the plurality of first training subsets is preprocessed to obtain a target to-be-trained data set conforming to the data model input rule.

[0063] Specifically, before model training, the training data needs to be preprocessed to make the training data conform to the data input format of the to-be-trained model. At the same time, the collected data is cleaned, denoised and normalized to improve the data quality.

[0064] In one possible embodiment, the data in the plurality of first training subsets is preprocessed to obtain a target to-be-trained data set conforming to the data model input rule, including: performing data completion and normalization processing on each first training data group in each first training subset to obtain a plurality of second training subsets; each second training subset includes a plurality of second training data groups after data completion and normalization processing; performing feature extraction on the second training data groups in each second training subset to obtain the target to-be-trained data set; the target to-be-trained data set includes a plurality of third training subsets, and each third training subset includes a plurality of third training data groups obtained after feature extraction.

[0065] In a specific implementation, the collected historical data can be missing due to various reasons (e.g., data loss in the transmission process, or failure of acquisition due to failure, etc.), for example, a certain running data is missing in a certain first training data set, but the environment data and risk data are normally collected. In this embodiment, the missing data in each first training data set is completed by data completion, so that the data amount of each first training data set is consistent.

[0066] In addition, since the values of different data of each same data are different, in order to ensure the consistency of data, the data is normalized in this embodiment, and then the data quality is improved to improve the accuracy of model training.

[0067] In one possible embodiment, the data completion and normalization processing of each first training data set in each first training subset respectively to obtain a plurality of second training subsets comprises: determining the position of data missing in each first training data set; completing data missing at each data missing position by interpolation method; and scaling the data of each first training data set after data completion to a target range to obtain a plurality of second training data sets.

[0068] Specifically, in this embodiment, for the case of data missing, the interpolation method is used to complete the missing data. Taking a first training data set as an example, for the first training data set, the position of data missing is detected, and then interpolation is performed for those positions to complete the data. For example, it is detected that the voltage data at time t in the first training data set is missing, and the running voltage U0(t), the running current I0(t), the average load rate P0(t) of the distribution transformer, the average three-phase unbalance degree S0(t) of the distribution transformer, the environment temperature T0(t) and the environment humidity W0(t) at this time are calculated by the interpolation formula and filled into the first training set. Taking the running voltage as an example, the interpolation formula of the running voltage

[0069]

[0070] wherein, represents the running voltage at the previous time of time t, is a monitoring time interval, such as 1 minute, 5 minutes, 10 minutes, etc., which is not limited here.

[0071] Based on the interpolation formula of the running voltage , the running voltage at the previous time of time t is obtained. is substituted into the interpolation formula of the running voltage, and the running voltage at time t​​

[0072] It can be understood that if there is a missing value in the data of the operating current I0(t), the average load rate P0(t) of the transformer, the average three-phase unbalance degree S0(t) of the transformer, the ambient temperature T0(t) or the ambient humidity W0(t), the corresponding missing value can be calculated by the same processing.

[0073] Further, in the embodiment, the first training data set after data completion is normalized by data scaling. Specifically, each data in the first training data set is scaled to a target range, which is preferably between 0 and 1, so that the model can be easier to process data. The data scaling formula is: .

[0074] Taking the operating voltage in the first training data set as an example, the data scaling formula of the operating voltage is as follows:

[0075]

[0076] wherein, is the normalized (i.e. scaled) voltage data, represents the minimum value of the voltage data in the first training data set, represents the maximum value of the voltage data in the first training data set. The voltage in the target range is calculated by the data scaling formula of the voltage.

[0077] Based on the data scaling formula of the operating voltage , the operating voltage at time t is obtained, the minimum value of the voltage data in the first training data set , and the minimum value of the voltage data in the first training data set ; then the obtained data is substituted into the interpolation formula of the operating voltage, so that the normalized (i.e. scaled) voltage data is obtained.

[0078] It can be understood that the data scaling processing of the operating current I0(t), the average load rate P0(t) of the transformer, the average three-phase unbalance degree S0(t) of the transformer, the ambient temperature T0(t) or the ambient humidity W0(t) is consistent with the voltage , and the current I1(t), the average load rate P1(t) of the transformer, the average three-phase unbalance degree S1(t) of the transformer, the ambient temperature T1(t) and the ambient humidity W1(t) are obtained after normalization. The normalization process is consistent with the voltage , and will not be described here.

[0079] In a possible embodiment, the feature extraction on the training data groups in each second training subset to obtain the target training data set comprises: calculating a current fluctuation rate, a voltage deviation rate, a maximum average load rate and a maximum distribution transformer average three-phase unbalance degree of each second training subset; and determining a maximum environmental temperature and a maximum environmental humidity of each second training subset to obtain the target training data set.

[0080] Specifically, there are multiple first training data groups in each time interval, that is, there are multiple operating voltages , multiple operating currents I0(t), multiple distribution transformer average load rates P0(t), multiple distribution transformer average three-phase unbalance degrees S0(t), multiple environmental temperatures T0(t) and multiple environmental humidities W0(t).

[0081] Since it is a risk prediction for a time interval, the data needs to be processed to obtain optimized data, and then the corresponding training model can be trained.

[0082] The calculation process of each type of data is as follows:

[0083] (1) According to the multiple operating voltages, the corresponding voltage deviation rates are calculated, and the voltage deviation rate is used to reflect the stability of the deviation degree of the voltage from the rated value. The voltage deviation rate is calculated according to the following formula:

[0084]

[0085] wherein, is the standard deviation of the voltage, reflecting the dispersion degree of the voltage; is the average value of the voltage, reflecting the overall level of the voltage. Wherein,

[0086]

[0087]

[0088] is the current data after missing value completion and normalization, and n is a window length of the collected historical data.

[0089] (2) According to the multiple operating currents, the corresponding current fluctuation rates are calculated, and the current fluctuation rate is used to reflect the stability of the current. The current fluctuation rate is calculated according to the following formula:

[0090]

[0091] wherein, It is the standard deviation of the current value, reflecting the degree of dispersion of the current; It is the average value of the current, reflecting the overall level of the current. Among them,

[0092]

[0093]

[0094] The current data is filled with missing values ​​and normalized, where n is the length of a time interval of the collected historical data.

[0095] (3) Calculate the corresponding maximum average load factor based on the average load factor P0(t) of multiple distribution transformers. Maximum average load factor. The calculation formula is as follows:

[0096]

[0097] in, The maximum value is taken for the average load rate of the k-th time interval.

[0098] (4) Calculate the corresponding average three-phase unbalance of the transformer based on the average three-phase unbalance S0(t) of multiple transformers, and take the maximum value S. K Maximum average load factor S K The calculation formula is as follows:

[0099]

[0100] The maximum value is taken for the average three-phase imbalance of the distribution transformer in the k-th time interval.

[0101] (5) Calculate the corresponding ambient temperature based on multiple ambient temperatures T0(t) and take the maximum value T. K Maximum average load factor T K The calculation formula is as follows:

[0102]

[0103] The maximum value of the ambient temperature for the k-th time interval is taken.

[0104] (6) Calculate the corresponding environmental humidity based on multiple environmental humidity values ​​W0(t) and take the maximum value W. K Maximum average load factor W K The calculation formula is as follows:

[0105]

[0106] The maximum value of the ambient humidity of the kth time interval is taken.

[0107] It can be seen that, in the embodiment, the feature extraction is performed on the second training subset corresponding to each time interval to obtain data conforming to the format corresponding to the to-be-trained model.

[0108] In step S205, the target data set is divided into a target to-be-trained data set and a target test set.

[0109] In the specific implementation, the processed target data set is divided into a target to-be-trained data set and a target test set, and the division can be performed at a ratio of 7:3 or 8:2, but in some cases, other ratios can be selected as needed, which is not limited herein.

[0110] The determined current fluctuation rate, voltage deviation rate, maximum average load rate, maximum distribution transformer average three-phase unbalance degree, maximum ambient temperature, and maximum ambient humidity in the target test set are input into the to-be-trained model, and the corresponding risk label is also input into the to-be-trained model. A test set is generated, which is a feature label matrix X as follows:

[0111]

[0112] The first 6 columns are feature columns, and the 7th column is a label column; the first 6 columns are extracted features, and the last column b K represents the low-voltage power consumption area risk of the next time window, and is a label column. Thus, the construction of the target test set is completed.

[0113] In step S206, a to-be-trained model is constructed, the to-be-trained model is trained according to the target to-be-trained data set, and a trained risk situation prediction model is obtained.

[0114] Specifically, as shown in Figure 4 The construction of the to-be-trained model includes: constructing the to-be-trained model based on a radial basis function (RBF) neural network; the to-be-trained model includes an input layer, a processing layer, and an output layer; the input layer is configured to receive a third training subset in the target to-be-trained data set; the processing layer is configured to map the input third training subset to a high-dimensional space to obtain a feature label matrix; and the output layer is configured to process the feature label matrix to obtain the risk type of the next time interval of the low-voltage power consumption area.

[0115] In the specific implementation, as shown in Figure 4As shown, a training model is first constructed. To address the characteristics of low-voltage power station area risk prediction, the traditional SVM algorithm is improved by introducing a kernel function and adjusting the penalty parameter to enhance the model's prediction accuracy and generalization ability. The training model is constructed based on a radial basis function (RBF) neural network, which has a three-layer structure: an input layer, a processing layer (i.e., a hidden layer), and an output layer. In this embodiment, the input layer includes six input variables (current fluctuation rate, voltage deviation rate, maximum average load rate, maximum average three-phase imbalance of the distribution transformer, maximum ambient temperature, and maximum ambient humidity), and the output layer includes one output variable (low-voltage power station area risk type: 0, 1, 2, 3, 4). The input layer receives preprocessed and feature-extracted input data. The processing layer uses a radial basis function to map the input to a high-dimensional space, and the output layer uses a linear function to map the output of the processing layer to the target space, i.e., the output variable. The radial basis function serves as the activation function for the neurons in the processing layer; in this invention, a Gaussian function is used. The basis functions are radial basis functions. Therefore, the parameters of an RBF neural network include three parts: the center of the basis function, its width, and its weights (i.e., the linear weights from the hidden layer to the output layer). ).

[0116] After obtaining the target training dataset and target test set, the basis function centers of the RBF neural network are first determined. Specifically, the center positions of the basis functions in the input space are determined. This can be done by randomly selecting several sample points from the target training dataset as basis function centers; alternatively, algorithms such as K-means clustering can be used to cluster the training data, using the cluster centers as basis function centers. The number and distribution of basis function centers affect the network's approximation ability and generalization performance; generally, the more centers, the higher the approximation accuracy. Next, the width of the basis functions in the RBF neural network is determined. The width of the basis functions determines their scope and shape. The width can be determined based on the distance between the basis function centers, for example, by multiplying the average distance between all centers by an empirical coefficient. A larger width results in a smoother basis function, less sensitive to local changes in the input data, but may lead to reduced approximation accuracy; a smaller width results in stronger locality of the basis functions, more accurately modeling local features of the input data, but may make the network more sensitive to noise. Finally, the weights of the RBF neural network are determined. Specifically, weights are used to combine the outputs of basis functions to obtain the final output of the network. Initial weights can be set randomly, typically within a small range, such as [-0.1, 0.1] or [-1, 1]. They can also be initialized based on prior knowledge or experience; for example, if there is some understanding of the problem, the initial weights can be set according to the expected function characteristics.

[0117] Specifically, the processing procedure of the RBF neural network is as follows:

[0118] (1) The output of the basis function.

[0119] Calculate the distance: For each input sample x, calculate its distance to each basis function center , usually using the Euclidean distance .

[0120] Calculate the output by the basis function: Substitute the calculated distance into the selected radial basis function to get the output of each basis function for the input sample x. For example, use the Gaussian basis function , where is the width of the jth basis function. In this way, for each input sample, a set of basis function outputs is obtained, which constitutes a vector reflecting the characteristics of the input sample under the action of different basis functions.

[0121] (2) Train the network.

[0122] Define the loss function: Choose a suitable loss function to measure the difference between the network output and the target output. The commonly used loss function is mean square error (MSE), that is , where is the target output, is the actual output of the network, and n is the number of training samples. The smaller the value of the loss function, the better the approximation effect of the network.

[0123] Select optimization algorithm: Use optimization algorithms to adjust the weights of the RBF neural network to minimize the loss function. Common optimization algorithms include gradient descent, stochastic gradient descent, conjugate gradient, Levenberg-Marquardt algorithm, etc. Take the gradient descent method as an example, which calculates the gradient of the loss function with respect to the weights, and then updates the weights in the opposite direction of the gradient, that is , where is the weight at the kth iteration, and α is the learning rate, which controls the step size of weight update.

[0124] Iterative update of weights: Input the samples in the training set one by one into the network, and update the weights according to the weight update amount calculated by the optimization algorithm. This process will be repeated several times until the value of the loss function converges to a small value, or reaches the preset maximum number of iterations. In each iteration, the network will calculate the output of the network according to the current weights and basis function outputs, and compare it with the target output, then adjust the weights according to the feedback of the loss function, so that the network gradually learns the correct input-output mapping relationship.

[0125] (3) Test evaluation and adjustment.

[0126] Evaluate network performance: Test the trained RBF network using the test set, input the input samples in the test set into the network, get the output of the network, and compare it with the target output of the test set. Calculate some evaluation indicators such as mean square error (MSE), mean absolute error (MAE), determination coefficient (R2), etc. to measure the approximation performance of the network. These indicators can help us understand the performance of the network on unknown data, and evaluate whether the network is overfitting or underfitting.

[0127] Adjust network parameters: According to the test results, adjust and optimize the parameters of the network. If the approximation effect of the network is not ideal, for example, the mean square error is large, you can consider increasing the number of basis functions or adjusting the center and width of the basis functions, or adjusting the learning rate and other parameters in the training process, and then retrain and test until a satisfactory approximation effect is obtained.

[0128] Visualize the results: Visualize the approximation results of the network and the target function, draw the target function curve and the network approximation curve, and intuitively observe the approximation degree of the network to the function. Through visualization, we can more clearly see which areas the network approximates well and which areas have large errors, so as to improve them in a targeted manner.

[0129] In one possible embodiment, the training of the to-be-trained model according to the target to-be-trained data set and the target test set to obtain a trained risk situation prediction model comprises: inputting each third training subset in the target to-be-trained data set into the to-be-trained model in turn; processing the data in the input third training subset by the to-be-trained model to obtain a first risk type; comparing the first risk type with a target risk type in the target test set; if accurate, completing the training of the to-be-trained model to obtain the risk situation prediction model; if not accurate, optimizing the to-be-trained model by a Bayesian optimization method, continuing to train the optimized to-be-trained model by the target to-be-trained data set, and repeating the above operations until the data result of the model is accurate, thereby obtaining the risk situation prediction model.

[0130] Specifically, the second historical data is re-collected to generate a training set, the data in the training set is input into the to-be-trained model, the to-be-trained model generates a corresponding risk type based on the input data in the training set, and the risk type is compared with the corresponding risk type in the test set. If consistent, it is determined that the model training is completed, if not consistent, it indicates that the model has errors, the model parameters are adjusted, the data in the training set is re-trained, and the model training is completed until the output result of the model is accurate, thereby obtaining the risk situation prediction model.

[0131] It can be seen that, in the embodiment, the training result is verified by the test set to ensure that the model training is completed and improve the model training accuracy.

[0132] In step S207, the real-time operation data and the real-time environment data of the first time interval of the low-voltage power consumption station area are input into the risk situation prediction model to obtain the risk type of the second time interval of the low-voltage power consumption station area.

[0133] Specifically, the real-time operation data and the real-time environment data of the low-voltage power consumption station area are input into the risk situation prediction model to obtain the risk type of the next time interval of the low-voltage power consumption station area, including:

[0134] The real-time operation data and the real-time environment data of the low-voltage power consumption station area in a first time window are collected; the real-time operation data includes the operating voltage, current, average load rate and average three-phase unbalance degree of the distribution transformer; the real-time environment data includes the environmental temperature and humidity; the target data is preprocessed to obtain the to-be-processed data conforming to the input rule of the risk situation prediction model; the risk situation prediction model is used to determine the target risk type of the low-voltage power consumption station area in a second time window according to the to-be-processed data; wherein the risk situation prediction model is trained for a plurality of continuous time windows based on a target training set, the time length of each time window is consistent, the plurality of time windows include the first time window and the second time window, and the second time window is the next time window of the first time window; and target early warning information is output according to the target risk type.

[0135] In a specific implementation, after obtaining the trained risk situation prediction model, the risk situation prediction model can be used to predict the risk of the low-voltage power consumption station area.

[0136] Specifically, the real-time operation data and the real-time environment data of the low-voltage power consumption station area in a first time window are obtained, the first time window is a time interval corresponding to a moment when the low-voltage power consumption station area performs risk prediction, and the real-time operation data and the real-time environment data are related data generated in real time in the time interval.

[0137] After the real-time operation data and the real-time environment data are obtained, the real-time operation data and the real-time environment data are input into the risk situation prediction model, the risk situation prediction model generates the risk type of a next time window (i.e., a second time window) of the first time window according to the real-time operation data and the real-time environment data, and finally the system outputs target early warning information according to the target risk type. The user can perform corresponding processing according to the risk type, or the system can perform emergency processing.

[0138] It can be seen that, in the embodiment, the real-time running data and the real-time environment data of the low-voltage power consumption station area are collected in real time, input into the trained risk situation prediction model, and the risk type of the next time interval of the low-voltage power consumption station area is predicted, so that the real-time monitoring and prediction of the risk situation of the low-voltage power consumption station area are realized.

[0139] The above describes the scheme of the embodiments of the present application mainly from the perspective of the process of executing the method. It can be understood that, in order to implement the above functions, the mobile electronic device includes hardware structures and / or software modules corresponding to the execution of each function. Those skilled in the art should easily realize that, in combination with the unit and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driven hardware depends on the specific application and design constraints of the technical solution. The skilled person can use different methods to implement the described functions for each specific application, but such implementation should not be considered as beyond the scope of the present application.

[0140] The embodiments of the present application can divide the functional units of the electronic device according to the above method examples, for example, each functional unit can be divided according to each function, or two or more functions can be integrated in one processing unit. The integrated unit can be implemented in the form of hardware or software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, and is only a logical functional division. There can be another division method in actual implementation.

[0141] Please refer to Figure 5 The present application also provides a risk situation prediction device 30 for a low-voltage power consumption station area, comprising:

[0142] The acquisition unit 31 is configured to acquire historical data of the low-voltage power consumption station area; wherein the historical data includes a plurality of running data, a plurality of environment data and a plurality of risk data;

[0143] The grouping unit 32 is configured to group the historical data according to a first grouping rule to obtain a plurality of first training data groups; the first grouping rule refers to grouping the running data, the environment data and the risk data at the same time into a group; each first training data group includes a plurality of input parameters and output parameters, wherein the input parameters include the running data and the environment data, and the output parameters include the risk data;

[0144] The grouping unit 32 is further configured to group the plurality of first training data groups according to a second grouping rule to obtain a plurality of first training subsets; the second grouping rule refers to grouping the first training data groups in the same time interval into a first training subset;

[0145] The preprocessing unit 33 is configured to preprocess data in the plurality of first training subsets to obtain a target training data set conforming to a data model input rule.

[0146] The model construction unit 34 is configured to construct a training model, train the training model according to the target training data set, and obtain a trained risk situation prediction model.

[0147] The prediction unit 35 is configured to input real-time operation data and real-time environment data of a first time interval of the low-voltage power consumption area into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption area.

[0148] As can be seen, in the present application, historical data of a low-voltage power consumption area is acquired, the historical data is grouped according to a first grouping rule to obtain a plurality of first training data groups, the plurality of first training data groups are grouped according to a second grouping rule to obtain a plurality of first training subsets, data in the plurality of first training subsets is preprocessed to obtain a target training data set conforming to a data model input rule, a training model is constructed, the training model is trained according to the target training data set, and a trained risk situation prediction model is obtained, and finally, real-time operation data and real-time environment data of a first time interval of the low-voltage power consumption area are input into the risk situation prediction model to obtain a risk type of a second time interval of the low-voltage power consumption area. In this way, real-time operation data and real-time environment data of a low-voltage power consumption area are collected in real time, input into the trained risk situation prediction model, and a risk type of a next time interval of the low-voltage power consumption area is predicted, thereby realizing real-time monitoring and prediction of a risk situation of a low-voltage power consumption area.

[0149] In one possible implementation, the preprocessing unit 33 is specifically configured to perform data completion and normalization processing on each first training data group in each first training subset to obtain a plurality of second training subsets, each second training subset includes a plurality of second training data groups after data completion and normalization processing, feature extraction is performed on each training data group in each second training subset to obtain the target training data set, and the target training data set includes a plurality of third training subsets, each third training subset includes a plurality of third training data groups after feature extraction.

[0150] In a possible embodiment, the aspect of performing data completion and normalization processing on each first training data group in each first training subset respectively to obtain a plurality of second training subsets, the preprocessing unit 33 is specifically configured to: determine a position of data missing in each first training data group; perform data completion on the position of data missing by an interpolation method; and scale the data of each first training data group after data completion to a target range to obtain a plurality of second training data groups.

[0151] In a possible embodiment, the aspect of performing feature extraction on each training data group in each second training subset to obtain the target training data set, the preprocessing unit 33 is specifically configured to: calculate a current fluctuation rate, a voltage deviation rate, a maximum average load rate, and a maximum distribution transformer average three-phase unbalance degree of each second training subset; and determine a maximum environmental temperature and a maximum environmental humidity of each second training subset to obtain the target training data set.

[0152] In a possible embodiment, the aspect of constructing the training model, the model construction unit 34 is specifically configured to: construct the training model based on a radial basis neural network; the training model includes an input layer, a processing layer, and an output layer; the input layer is configured to receive a third training subset in the target training data set; the processing layer is configured to map the input third training subset to a high-dimensional space to obtain a feature label matrix; and the output layer is configured to process the feature label matrix to obtain the risk type of the next time interval of the low-voltage power consumption area.

[0153] In a possible embodiment, the aspect of training the training model according to the target training data set to obtain a trained risk situation prediction model, the model construction unit 34 is specifically configured to: input each third training subset in the target training data set into the training model in sequence; process the data in the input third training subset by the training model to obtain a first risk type; compare the first risk type with a target risk type in the third training subset; if the comparison is accurate, training of the training model is completed to obtain the risk situation prediction model; if the comparison is not accurate, the training model is optimized by a Bayesian optimization method, the optimized training model is trained by the target training data set, and the above operations are repeated until the data result of the model is accurate, and the risk situation prediction model is obtained.

[0154] In a possible embodiment, the aspect of inputting the real-time operation data and the real-time environment data of the low-voltage power consumption station area into the risk situation prediction model to obtain the risk type of the next time interval of the low-voltage power consumption station area, the prediction unit 35 is specifically configured to: collect real-time operation data and real-time environment data of the low-voltage power consumption station area in a first time window; the real-time operation data includes the operating voltage, current, average load rate and average three-phase unbalance degree of the power distribution transformer; the real-time environment data includes the ambient temperature and the ambient humidity; pre-process the target data to obtain to-be-processed data conforming to the input rule of the risk situation prediction model; determine the target risk type of the low-voltage power consumption station area in a second time window according to the to-be-processed data by using the risk situation prediction model; wherein the risk situation prediction model is obtained by training based on a target training set for a plurality of continuous time windows, the time length of each time window is consistent, the plurality of time windows include the first time window and the second time window, and the second time window is the next time window of the first time window; and output target early warning information according to the target risk type.

[0155] The above embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions described in the embodiments of the present application are generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another, for example, the computer instructions can be transmitted from one website, computer, server or data center to another website, computer, server or data center through a wired or wireless manner. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. The semiconductor medium can be a solid state disk.

[0156] The present application also provides an electronic device 10, such as Figure 6As shown, the electronic device 10 includes at least one processor 11, a display screen 12, and a memory 13, and can further include a communications interface 15 and a bus 14. The processor 11, the display screen 12, the memory 13, and the communications interface 15 can communicate with each other through the bus 14. The display screen 12 is configured to display a preset user guide interface in the initial setting mode. The communications interface 15 can transmit information. The processor 11 can invoke a logical instruction in the memory 13 to execute the method in the above-described embodiments.

[0157] Optionally, the electronic device 10 can be a mobile electronic device, or an electronic device or other device, which is not limited herein.

[0158] In addition, when the logical instruction in the memory 13 is implemented in the form of a software functional unit and sold or used as an independent product, the logical instruction can be stored in a computer readable storage medium.

[0159] The memory 13, as a computer readable storage medium, can be configured to store a software program, a computer executable program, such as a program instruction or a module corresponding to the method in the embodiments of the present disclosure. The processor 11 executes the function application and data processing by running the software program, the instruction or the module stored in the memory 13, that is, implements the method in the above-described embodiments.

[0160] The memory 13 can include a program storage area and a data storage area. The program storage area can store an operating system and at least one application required by a function. The data storage area can store data created according to the use of the electronic device 10, and the like. In addition, the memory 13 can include a high-speed random access memory, and can further include a nonvolatile memory. For example, the memory 13 can be a variety of media that can store program codes, such as a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and can be a transitory storage medium.

[0161] The embodiments of the present application further provide a computer storage medium, wherein the computer storage medium stores a computer program for electronic data exchange, and the computer program causes a computer to execute part or all of the steps of any method described in the above method embodiments, and the computer includes the electronic device.

[0162] The embodiment of the present application further provides a computer program product, which comprises a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to cause a computer to execute part or all of the steps of any method described in the above method embodiments. The computer program product can be a software installation package, and the computer comprises an electronic device.

[0163] It should be understood that, in various embodiments of the present application, the size of the sequence number of the above processes does not mean the order of execution, and the execution order of the processes should be determined according to its function and inherent logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0164] In several embodiments provided in the present application, it should be understood that the disclosed methods, devices and systems can be implemented in other ways. For example, the above-described device embodiments are only schematic; for example, the division of the units is only a logical function division, and actual implementation can have another division manner; for example, a plurality of units or components can be combined or integrated into another system, or some features can be omitted or not executed. In addition, the coupling or direct coupling or communication connection between the units shown or discussed can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or other forms.

[0165] The units described as separate components can or can not be physically separated, and the components shown as units can or can not be physical units, that is, they can be located in one place, or can be distributed on a plurality of network units. According to actual needs, part or all of the units can be selected to achieve the purpose of the present embodiment.

[0166] In addition, each functional unit in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically included separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware, or in the form of hardware plus software functional unit.

[0167] The integrated unit in the form of software function unit can be stored in a computer readable storage medium. The software function unit is stored in a storage medium, including a plurality of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute part of steps of the method according to various embodiments of the present application. The storage medium includes a U disk, a mobile hard disk, a magnetic disk, an optical disk, a volatile memory or a non-volatile memory. The non-volatile memory can be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM) or a flash memory. The volatile memory can be a random access memory (RAM) used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) can be used, such as static random access memory (SRAM), dynamic random access memory (DRAM), synchronous dynamic random access memory (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchlink dynamic random access memory (SLDRAM) and direct rambus random access memory (DR RAM). Various media that can store program codes.

[0168] Although the present application is disclosed as above, the present application is not limited thereto. Any person skilled in the art can easily conceive variations or substitutions without departing from the spirit and scope of the present application, and various modifications can be made, including combinations of different functions and implementation steps, including software and hardware implementations, which are all within the scope of the present application.

Claims

1. A method for predicting the risk situation of a low-voltage radio station area, characterized in that, include: Acquire historical data for the low-voltage radio station area; wherein, the historical data includes multiple operational data, multiple environmental data, and multiple risk data; The historical data is grouped according to a first grouping rule to obtain multiple first training data groups; the first grouping rule refers to grouping the running data, environmental data, and risk data at the same time into one group; each first training data group includes multiple input parameters and output parameters, wherein the input parameters include the running data and the environmental data, and the output parameters include the risk data; The multiple first training data groups are grouped according to the second grouping rule to obtain multiple first training subsets; the second grouping rule refers to dividing the first training data groups in the same time interval into the same first training subset. The data in the plurality of first training subsets are preprocessed to obtain a target dataset that conforms to the input rules of the data model. This includes: performing data completion and normalization on the first training data group in each of the first training subsets to obtain a plurality of second training subsets; each second training subset includes a plurality of second training data groups after data completion and normalization; and performing feature extraction on the second training data groups in each of the second training subsets to obtain a target training dataset; the target training dataset includes a plurality of third training subsets, and each third training subset includes a plurality of third training data groups obtained after feature extraction. The step of extracting features from the second training data group in each second training subset to obtain the target training dataset includes: calculating the current fluctuation rate, voltage deviation rate, maximum average load rate, and maximum average three-phase imbalance of the distribution transformer for each second training subset; and determining the maximum ambient temperature and maximum ambient humidity for each second training subset to obtain the target training dataset; wherein each time interval includes multiple operating voltages. Multiple operating currents I0(t), multiple average load rates of distribution transformers P0(t), multiple average three-phase unbalance of distribution transformers S0(t), multiple ambient temperatures T0(t), and multiple ambient humidity levels W0(t); Calculating the current fluctuation rate, voltage deviation rate, maximum average load rate, and maximum transformer average three-phase imbalance for each second training subset includes: calculating the multiple operating voltages. The standard deviation and mean, based on the multiple operating voltages The voltage deviation rate is calculated using the standard deviation and the average value. The voltage deviation rate is used to reflect the stability of the degree of deviation between the voltage and the rated value; and the standard deviation and average value of the plurality of operating currents I0(t) are calculated, and the current fluctuation rate is calculated based on the standard deviation and average value of the plurality of operating currents I0(t). The current fluctuation rate Used to stabilize the reaction current; and to calculate the corresponding maximum average load factor based on the average load factors P0(t) of multiple distribution transformers. Furthermore, the average three-phase unbalance of multiple transformers S0(t) is calculated, and the maximum value S is taken as the average three-phase unbalance of the transformer. K ; Determining the maximum ambient temperature and maximum ambient humidity of each second training subset to obtain the target training dataset includes: calculating the corresponding ambient temperature T based on multiple ambient temperatures T0(t) and taking the maximum value T. K Furthermore, the corresponding environmental humidity is calculated based on multiple environmental humidity values ​​W0(t), and the maximum value W is taken. K ; The target dataset is divided into a target training dataset and a target test set, wherein the target test set is based on a predetermined current fluctuation rate. Voltage deviation rate Maximum average load rate The average three-phase unbalance of the distribution transformer is taken as the maximum value S. K The ambient temperature is taken as the maximum value T. K The ambient humidity is taken as the maximum value W. K and corresponding risk labels The generated feature label matrix X; A model to be trained is constructed based on a radial basis function neural network. The model is trained using the target training dataset and the target test set to obtain a trained risk situation prediction model. A second set of historical data is re-collected to generate a training set. The data in the training set is input into the trained risk situation prediction model. The trained risk situation prediction model generates the corresponding risk type based on the data input from the training set. The risk type is compared with the corresponding risk type in the test set. If they match, the model training is considered complete. If they do not match, it indicates that the model has an error. The model parameters are then adjusted, and the model is retrained using the data in the training set until the model outputs an accurate result. The model training is then considered complete, and the risk situation prediction model is obtained. The real-time operating data and real-time environmental data of the first time interval of the low-voltage radio station area are input into the risk situation prediction model to obtain the risk type of the second time interval of the low-voltage radio station area; the second time interval is the next time interval adjacent to the first time interval.

2. The method according to claim 1, characterized in that, The process of performing data completion and normalization on the first training data group in each first training subset yields multiple second training subsets, including: Identify the locations of missing data in each first training data set; Data is filled in at each location where data is missing using interpolation. After the data is completed, the data of each first training data group is scaled to the target range to obtain multiple second training data groups.

3. The method according to claim 1 or 2, characterized in that, The construction of the model to be trained includes: The model to be trained is constructed based on a radial basis function neural network; The model to be trained includes an input layer, a processing layer, and an output layer; the input layer is used to receive a third training subset from the target training dataset. The processing layer is used to map the input third training subset to a high-dimensional space to obtain a feature label matrix; the output layer is used to process the feature label matrix to obtain the risk type of the low-voltage radio station area in the next time interval.

4. The method according to claim 1 or 2, characterized in that, The step of training the model to be trained based on the target training dataset and the target test set to obtain a trained risk situation prediction model includes: Each third training subset in the target training dataset is sequentially input into the model to be trained; The model to be trained processes the data in the input third training subset to obtain the first risk type; The first risk type is compared with the target risk types in the target test set. If the comparison is accurate, the training of the model to be trained is completed, and the risk situation prediction model is obtained. If the comparison is inaccurate, the model to be trained is optimized using Bayesian optimization. The optimized model to be trained is then trained using the target training dataset. This process is repeated until the model's data results are accurate, and the risk situation prediction model is obtained.

5. The method according to claim 1 or 2, characterized in that, The step of inputting real-time operational data and real-time environmental data of the low-voltage power station area into the risk situation prediction model for the first time interval to obtain the risk type of the low-voltage power station area for the second time interval includes: The system collects real-time operating data and real-time environmental data of the low-voltage power distribution area within a first time window. The real-time operating data includes the operating voltage, current, average load rate, and average three-phase imbalance of the distribution transformer. The real-time environmental data includes ambient temperature and ambient humidity. The target data is preprocessed to obtain data to be processed that conforms to the input rules of the risk situation prediction model; The risk situation prediction model is used to determine the target risk type of the low-voltage radio station area in the second time window based on the data to be processed; wherein, the risk situation prediction model is trained on a target training set for multiple consecutive time windows, each time window has the same duration, the multiple time windows include the first time window and the second time window, and the second time window is the next time window after the first time window; Output target early warning information based on the target risk type.

6. A risk situation prediction device for a low-voltage radio station area, characterized in that, include: The acquisition unit is used to acquire historical data of the low-voltage radio station area; wherein, the historical data includes multiple operational data, multiple environmental data, and multiple risk data; A grouping unit is used to group the historical data according to a first grouping rule to obtain multiple first training data groups; the first grouping rule refers to dividing the running data, environmental data, and risk data at the same time into a group; each first training data group includes multiple input parameters and output parameters, wherein the input parameters include the running data and the environmental data, and the output parameters include the risk data; The grouping unit is further configured to group the plurality of first training data groups according to a second grouping rule to obtain a plurality of first training subsets; the second grouping rule refers to dividing the first training data groups in the same time interval into the same first training subset; A preprocessing unit is used to preprocess the data in the plurality of first training subsets to obtain a target dataset that conforms to the input rules of the data model. Specifically, the preprocessing unit is used to: perform data completion and normalization processing on the first training data group in each first training subset to obtain a plurality of second training subsets; each second training subset includes a plurality of second training data groups after data completion and normalization processing; extract features from the training data groups in each second training subset to obtain a target training dataset, including: calculating the current fluctuation rate, voltage deviation rate, maximum average load rate, and maximum average three-phase imbalance of the distribution transformer for each second training subset; and determining the maximum ambient temperature and maximum ambient humidity for each second training subset to obtain the target training dataset; the target training dataset includes a plurality of third training subsets, each third training subset including a plurality of third training data groups obtained after feature extraction; wherein each time interval includes multiple operating voltages. The training dataset includes multiple operating currents I0(t), multiple average transformer load rates P0(t), multiple average transformer three-phase imbalances S0(t), multiple ambient temperatures T0(t), and multiple ambient humidity levels W0(t). The dataset is used to calculate the current fluctuation rate, voltage deviation rate, maximum average load rate, and maximum average transformer three-phase imbalance for each second training subset, including calculating the multiple operating voltages. The standard deviation and mean, based on the multiple operating voltages The voltage deviation rate is calculated using the standard deviation and the average value. The voltage deviation rate is used to reflect the stability of the degree of deviation between the voltage and the rated value; and the standard deviation and average value of the plurality of operating currents I0(t) are calculated, and the current fluctuation rate is calculated based on the standard deviation and average value of the plurality of operating currents I0(t). The current fluctuation rate Used to stabilize the reaction current; and to calculate the corresponding maximum average load factor based on the average load factors P0(t) of multiple distribution transformers. Furthermore, the average three-phase unbalance of multiple transformers S0(t) is calculated, and the maximum value S is taken as the average three-phase unbalance of the transformer. K Determine the maximum ambient temperature and maximum ambient humidity for each second training subset to obtain the target training dataset, including: calculating the corresponding ambient temperature T0(t) based on multiple ambient temperatures and taking the maximum value T. K Furthermore, the corresponding environmental humidity is calculated based on multiple environmental humidity values ​​W0(t), and the maximum value W is taken. K Furthermore, the target dataset is divided into a target training dataset and a target test set, wherein the target test set is based on a predetermined current fluctuation rate. Voltage deviation rate Maximum average load rate The average three-phase unbalance of the distribution transformer is taken as the maximum value S. K The ambient temperature is taken as the maximum value T. K The ambient humidity is taken as the maximum value W. K and corresponding risk labels The generated feature label matrix X; The model building unit is used to build a model to be trained. It trains the model to be trained using the target training dataset and the target test set to obtain a trained risk situation prediction model. It then re-collects second historical data to generate a training set, inputs the data from the training set into the trained risk situation prediction model, and generates corresponding risk types based on the input data in the training set. This risk type is compared with the corresponding risk type in the test set. If they match, the model training is considered complete; otherwise, if they do not match, it indicates an error in the model. The model parameters are then adjusted, and the model is retrained using the data in the training set until the output result is accurate. Once this is complete, the model training is considered complete, and the risk situation prediction model is obtained. The prediction unit is used to input the real-time operating data and real-time environmental data of the first time interval of the low-voltage radio station area into the risk situation prediction model to obtain the risk type of the second time interval of the low-voltage radio station area; the second time interval is the next time interval adjacent to the first time interval.

7. An electronic device, characterized in that, The method includes a processor, a memory, a communication interface, and one or more programs, said programs being stored in the memory and configured to be executed by the processor, said programs including instructions for performing the steps of the method as described in any one of claims 1-5.

8. A computer-readable storage medium, characterized in that, A computer program for storing electronic data interchange is provided, wherein the computer program causes a computer to execute instructions for the steps of the method as described in any one of claims 1-5.

Images