Power distribution network abnormity monitoring method and system oriented to Internet of Things, and related equipment

Abstract

The invention is suitable for the technical field of power distribution network abnormity monitoring, and provides a power distribution network abnormity monitoring method and system oriented to the Internet of Things, and related equipment, and the method comprises the steps: collecting the abnormal data of electrical parameters, weather data and geographic position information data from each monitoring point of a power distribution network, and carrying out the data preprocessing of the collected data; inputting the preprocessed data into a trained CNN-RNN model to evaluate whether the electrical parameters are influenced by weather factors or geographic position factors; when the electrical parameters are not influenced by weather factors or geographic position factors, determining target characteristic variables of the operation state of each monitoring point device of the power distribution network; judging whether the target feature vector meets an abnormal condition or not; and if yes, abnormal analysis is carried out on the electrical parameters by monitoring abnormal data of the operation state of each monitoring point device of the power distribution network, so that the comprehensiveness of abnormal monitoring is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution network anomaly monitoring, and in particular to a distribution network anomaly monitoring method, system and related equipment for the Internet of Things. Background Art

[0002] At present, various types of intelligent sensors are widely used in distribution network monitoring. For example, current sensors and voltage sensors that can monitor electrical parameters in real time can accurately measure basic parameters such as current and voltage of the distribution network. Environmental monitoring sensors such as temperature sensors and humidity sensors can monitor the temperature and humidity of the environment where the distribution network equipment is located, providing data support for the evaluation of the equipment's operating environment. Based on the data provided by sensors, many use big data technology and artificial intelligence technology to store and manage massive amounts of monitoring data.

[0003] In actual work, sensors are easily disturbed by environmental factors, resulting in reduced data collection accuracy. Although artificial intelligence algorithms have been applied to fault diagnosis, in complex distribution network operating environments, for example, when multiple faults occur simultaneously or the fault characteristics are not obvious, fault diagnosis algorithms may misjudge or miss faults, affecting the timeliness and accuracy of fault handling. In addition, the current monitoring of distribution network equipment is mainly focused on some key equipment and common fault types, and there are problems such as insufficient monitoring of some auxiliary equipment and potential fault modes.

[0004] In view of this, a distribution network anomaly monitoring method, system and related equipment for the Internet of Things are needed. Summary of the invention

[0005] The embodiments of the present application provide a distribution network anomaly monitoring method, system and related equipment for the Internet of Things, which are used to solve the problem that distribution network anomaly monitoring is not accurate and comprehensive enough.

[0006] A first aspect of an embodiment of the present application provides a distribution network abnormality monitoring method for the Internet of Things, comprising:

[0007] Collect abnormal data of electrical parameters, weather data and geographical location information data from each monitoring point of the distribution network, and perform data preprocessing on the collected data;

[0008] Input the preprocessed data into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors;

[0009] When the electrical parameters are not affected by weather factors or geographical location factors, the target characteristic variables of the operating status of the equipment at each monitoring point of the distribution network are determined;

[0010] Determining whether the target feature vector satisfies an abnormal condition;

[0011] If so, an abnormal analysis is performed on the electrical parameters by monitoring abnormal data of the operating status of equipment at each monitoring point of the distribution network.

[0012] Furthermore, the pre-processed data is input into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors, including:

[0013] Obtain historical abnormal data, weather data, and geographic location information of electrical parameters at each monitoring point in the distribution network within a preset time, and normalize the data

[0014] The normalized data is used as the first input sample and input into the convolutional neural network to obtain the corresponding feature map sequence;

[0015] Input the feature graph sequence as a second input sample into a recurrent neural network to perform learning and training on the model;

[0016] The binary cross entropy loss function and optimization algorithm are used to evaluate and optimize the model to obtain the trained CNN-RNN model.

[0017] Furthermore, the normalized data is used as the first input sample and input into the convolutional neural network to obtain the corresponding feature map sequence, including:

[0018] The normalized electrical parameter anomaly data, weather data, and geographic location information data are concatenated in the channel dimension to obtain an input tensor;

[0019] Use multiple convolution kernels to perform a convolution operation on the input tensor, apply an activation function to the output, and add a pooling layer to reduce the output data dimension to obtain a target feature map sequence.

[0020] Furthermore, the using multiple convolution kernels to perform a convolution operation on the input tensor includes:

[0021]

[0022] Where: y i,k,m is the output of each i-th sample, k-th time step, and m-th convolution kernel, w m,a,b,j is the convolution kernel weight, b m is the bias term.

[0023] Furthermore, the step of taking the feature graph sequence as a second input sample and inputting it into a recurrent neural network to perform learning and training on the model includes:

[0024] Initializing model parameters of a recurrent neural network, and inputting the feature graph sequence as a second output sample into the recurrent neural network;

[0025] LSTM unit calculations are performed separately, including forget gate, input gate, candidate cell state, cell state update, output gate, and hidden state update.

[0026] Furthermore, if so, the electrical parameters are analyzed abnormally by monitoring abnormal data of the operating status of equipment at each monitoring point of the distribution network, including:

[0027] Constructing a random forest model, and determining sample data of the model according to the target characteristic variables and electrical parameter abnormal data;

[0028] Calculate the contribution of each feature of each decision tree in the random forest model to reducing the Gini coefficient when the node is split, and obtain the importance score of the target feature variable;

[0029] The electrical parameters are analyzed for abnormality according to the order of importance scores of the target feature variables.

[0030] Furthermore, the contribution of each feature of each decision tree in the random forest model to reducing the Gini coefficient when the node is split is calculated to obtain the importance score of the target feature variable, including:

[0031]

[0032] Where: V i is the importance score of the target feature variable, k is the number of decision trees, t is the number of decision tree nodes, V ti is the target feature variable X i Contribution to reducing the Gini coefficient in the decision tree, N t is a set of nodes, G j is the Gini coefficient, G jL and G jR is the Gini coefficient of the two child nodes after splitting.

[0033] Furthermore, the determining whether the target feature vector satisfies an abnormal condition includes:

[0034] If not, then determine the abnormal characteristic variables of the monitoring data transmission of each monitoring point of the distribution network, and perform abnormal analysis on the electrical parameters according to the abnormal characteristic variables.

[0035] A second aspect of the embodiment of the present application provides a distribution network anomaly monitoring system for the Internet of Things, including:

[0036] The data acquisition and preprocessing unit is used to collect abnormal data of electrical parameters, weather data and geographical location information data from various monitoring points of the distribution network, and perform data preprocessing on the collected data;

[0037] A CNN-RNN model evaluation unit, used for inputting the preprocessed data into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors;

[0038] A target characteristic variable determination unit, used for determining the target characteristic variable of the operating state of the equipment at each monitoring point of the distribution network when the electrical parameter is not affected by weather factors or geographical location factors;

[0039] A target feature vector judgment unit, used to judge whether the target feature vector meets an abnormal condition;

[0040] The electrical parameter abnormality analysis unit is used to perform abnormality analysis on the electrical parameters by monitoring abnormal data of the operating status of equipment at each monitoring point of the distribution network when the target feature vector meets the abnormal condition.

[0041] A third aspect of an embodiment of the present application provides a computer device, including a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, wherein when the processor executes the computer-readable instructions, the steps of the distribution network anomaly monitoring method for the Internet of Things as described in any one of the above are implemented.

[0042] It can be seen from the above technical solutions that the embodiments of the present application have the following advantages:

[0043] The present invention first analyzes whether the abnormal electrical parameters at each monitoring point are affected by characteristic weather factors or special geographical location factors based on the collected electrical parameter abnormal data, weather data and geographical location information data. When it is determined that it is not affected by weather or geographical location factors, the device operation status data associated with the electrical parameter is determined, and it is judged whether the associated device operation status data meets the set abnormal conditions. If so, the electrical parameters are analyzed abnormally according to the abnormal data of the monitored device operation status data. The present invention analyzes the influencing factors of electrical anomalies in a hierarchical manner, first excluding the influence of weather and geographical location factors, and then based on the correlation between the device operation status and the electrical parameters, the device operation status data is fully monitored to obtain the cause of the abnormal electrical parameters, and the corresponding abnormal solutions can be obtained.

[0044] Other advantages, objectives, and features of the present invention will be set forth in part in the following description, and in part will be apparent to those skilled in the art based on an examination of the following or may be taught from the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 The present invention is a flowchart of an embodiment of a method for monitoring abnormal distribution network for the Internet of Things. DETAILED DESCRIPTION

[0046] The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein, for example. In addition, the terms "including" and "corresponding to" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0047] Embodiment 1

[0048] The implementation method in this embodiment can be implemented in the system, can be implemented in the server, and can also be implemented in the terminal, and there is no specific limitation. The following will introduce the distribution network abnormality monitoring method for the Internet of Things in this application from the perspective of system implementation. Figure 1 , the method provided in the embodiment of the present application comprises the following steps:

[0049] S11. Collect abnormal data of electrical parameters, weather data and geographic location information data from each monitoring point of the distribution network, and perform data preprocessing on the collected data;

[0050] In this embodiment, the portion of the distribution network to be monitored is first identified, such as a specific substation, line or equipment. The electrical parameters that need to be collected are determined, and common ones include voltage, current, power factor, active power, reactive power, etc. The abnormal data here is set with a threshold value based on actual needs. For example, a voltage exceeding ±10% of the rated value is considered abnormal, and a current exceeding a certain proportion of the rated current of the equipment. The monitoring equipment collects electrical parameter data at a certain frequency, and the collection frequency can be determined based on the frequency of occurrence of abnormal situations and the requirements for real-time performance.

[0051] The weather data here refers to the data collected through channels such as meteorological departments, meteorological monitoring stations, and third-party meteorological data providers, including factors related to the operation of the distribution network, such as temperature, humidity, wind speed and direction, rainfall, and special weather phenomena (lightning, haze, sandstorms). Geographic location information data uses the global positioning system (GPS) or other positioning technologies to obtain the longitude and latitude coordinates of the monitoring point; or divides the area covered by the distribution network into several areas and assigns a unique number to each area. For longitude and latitude coordinates, the GPS device can be used to obtain accurate coordinate information when the monitoring equipment is installed, and it can be transmitted to the data center together with the electrical parameter data. For the area number, the scope and number of each area can be pre-determined during the planning and management of the distribution network, and the area number where the monitoring point is located can be recorded during data collection.

[0052] After the specific data is collected, the data is preprocessed, including data cleaning, data normalization and other operations. The specific preprocessing operations need to be set according to the actual situation and are not limited here.

[0053] S12. Input the preprocessed data into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors;

[0054] 1. Obtain historical abnormal data, weather data, and geographic location information of electrical parameters at each monitoring point in the distribution network within a preset time, and normalize the data

[0055] 2. The normalized data is used as the first input sample and input into the convolutional neural network to obtain the corresponding feature map sequence;

[0056] Specifically, we collect abnormal data of electrical parameters of the distribution network. Assume that there are n abnormal data samples. Each sample contains time series data of electrical parameters (voltage, current, power, etc.), recorded as Where t is the length of the time series, d e It is the electrical parameter dimension.

[0057] Collect the corresponding weather data (temperature, humidity, wind speed, weather type, etc.), expressed as d w Is the weather data dimension.

[0058] Collect geographic location information. If the geographic location is divided into m regions, it can be represented by one-hot encoding. Then the geographic location data can be represented as L∈R n×m If you use continuous values ​​such as longitude and latitude to represent geographic location, you can convert it into a suitable numeric vector form.

[0059] For electrical parameter data, for each dimension j (1≤j≤d e ) to normalize. Calculate the mean and standard deviation. Perform similar normalization on the weather data to obtain X ′ w For geographic location data, if it is in one-hot encoding form, normalization is not required; if it is in continuous value form, appropriate normalization (such as standardization) can be performed according to its data characteristics.

[0060] Build a CNN model that combines weather and geographic location:

[0061] (1) The normalized electrical parameter anomaly data, weather data, and geographic location information data are concatenated in the channel dimension to obtain an input tensor;

[0062] (2) Use multiple convolution kernels to perform a convolution operation on the input tensor, apply an activation function to the output, and add a pooling layer to reduce the output data dimension to obtain a target feature map sequence.

[0063] Input layer: The normalized electrical parameter data X ′ e 、Weather dataX ′ w And the geographic location data L are spliced ​​in the channel dimension to obtain a new input tensor

[0064] Convolution layer: Use multiple convolution kernels to perform convolution operations. Assume that the size of the convolution kernel is k×k and the number of convolution kernels is c. For the input data X i Each time step slice of The convolution operation formula is:

[0065]

[0066] Where: y i,k,m is the output of each i-th sample, k-th time step, and m-th convolution kernel, w m,a,b,j is the convolution kernel weight, b m is the bias term.

[0067] Activation function: Apply an activation function to the convolutional layer output, such as z i,k,m =max(0,y i,k,m ).

[0068] Pooling layer: Add a pooling layer to reduce the data dimension, using the maximum pooling operation, the pooling window size is s×s, and the step size is p. For each channel m, the maximum pooling formula is: u i,k,m =max{z i,k×p+a,m}.

[0069] 3. The feature map sequence is used as the second input sample and input into the recurrent neural network to learn and train the model;

[0070] The feature map sequence output by CNN is assumed to be U∈R n×t×c , as the input of RNN. Here we take the long short-term memory network (LSTM) as an example, and its calculation process is as follows:

[0071] (1) Initializing model parameters of a recurrent neural network, and inputting the feature graph sequence as a second output sample into the recurrent neural network;

[0072] (2) Perform LSTM unit calculations separately, including forget gate, input gate, candidate cell state, cell state update, output gate, and hidden state update.

[0073] For each sample i, initialize the cell state C i,0 and the hidden state h i,0 is a zero vector or a random vector. Then the LSTM unit calculations are performed separately, including the forget gate, input gate, candidate cell state, cell state update, output gates, and hidden state update.

[0074] 4. Use the binary cross entropy loss function and optimization algorithm to evaluate and optimize the model to obtain the trained CNN-RNN model.

[0075] Since we need to determine whether the anomaly is affected by weather and geographical location, we can regard it as a binary classification problem, that is, affected is 1 and unaffected is 0. Set a label vector Y∈{0,1} n×t , indicating whether each sample is affected by weather and geographical location at each time step. The binary cross entropy loss function is used for calculation:

[0076]

[0077] in: The model predicts the probability of whether sample i is affected at time step k.

[0078] Initialize all the model's weights (including CNN and RNN weights) and biases to random values. At the same time, initialize the parameters of the Adam optimizer: learning rate, exponential decay rate of first-order moment estimate, exponential decay rate of second-order moment estimate, and a small constant. In each training iteration, calculate the gradient of the loss function with respect to the model parameters. After training is complete, use the test set to evaluate the model. Evaluation indicators such as accuracy, recall, and F1-score can be used.

[0079] Based on the predicted probability output by the model, the above analysis is performed to determine whether the abnormal samples of each electrical parameter are affected by weather and geographical location. For samples predicted to be affected, the degree of influence of weather and geographical location factors is further analyzed. This can be determined by analyzing the distribution of model prediction results under different weather conditions and different geographical locations, and comparing them with actual abnormal data. For example, if in a certain area, the model predicts that the probability of being affected is high under multiple weather conditions, it means that geographical location may be the main influencing factor; if under a specific weather type, the predicted probability of being affected in multiple areas increases, it means that the weather factor has a greater impact.

[0080] S13. When the electrical parameters are not affected by weather factors or geographical location factors, the target characteristic variables of the operating status of the equipment at each monitoring point of the distribution network are determined;

[0081] After the above steps, it is calculated that the electrical parameters are not affected by weather factors or geographical location factors. Then, through the correlation analysis algorithm and the Pearson correlation coefficient, the related variables with abnormal operating status of each monitoring device in the distribution network and electrical parameters are screened, and the screened related variables are determined as the target characteristic variables.

[0082] S14. Determine whether the target feature vector meets the abnormal condition;

[0083] The abnormal condition here refers to determining whether these variables are abnormal among the variables determined above that are correlated with the electrical parameter abnormality. The method of judging whether the target feature vector is abnormal is not limited here. The above-mentioned CNN-RNN model method can be used for abnormal screening, or other methods can be used, which are not limited here.

[0084] S15. If yes, then an abnormal analysis of the electrical parameters is performed by monitoring the abnormal data of the operating status of the equipment at each monitoring point of the distribution network.

[0085] When it is determined that the selected related abnormal characteristic variables are also abnormal, the abnormal data of the equipment operation status at each monitoring point of the distribution network are used to perform abnormal analysis on the electrical parameters. The details are as follows:

[0086] 1. Build a random forest model and determine the sample data of the model based on the target characteristic variables and electrical parameter abnormal data;

[0087] Assume that the screened abnormal data of equipment operation status includes n features (such as equipment temperature, vibration amplitude, current, etc.), denoted as X = {X1, X2, ..., X n}, there are m groups of sample data. The electrical parameter abnormal data is the target variable, such as the voltage abnormal value, recorded as Y = {Y1, Y2, ..., Y m}.

[0088] 2. Calculate the contribution of each feature of each decision tree in the random forest model to reducing the Gini coefficient when the node is split, and obtain the importance score of the target feature variable;

[0089] Determine the number of decision trees through cross-validation or experience. Too few decision trees may lead to underfitting of the model, while too many may lead to overfitting and increase computational costs. Generally, you can start with a smaller number (such as 10), gradually increase it and observe the model performance. Such as the maximum depth of each decision tree, the minimum number of samples required for each node to split, etc. These parameters can be adjusted according to the characteristics of the data and the complexity of the problem. From the sample data set determined above, samples are extracted with replacement to construct each decision tree. The training process for each decision tree is as follows:

[0090] When each node splits, a subset is randomly selected from the n features, and then the best split features and split points are selected in this subset to minimize the impurity of the child nodes after the split. Impurity can be measured using the Gini coefficient. According to the selected split features and split points, a decision tree is recursively constructed until the stopping condition is met. For example, the maximum depth is reached, the number of node samples is less than a certain threshold, etc. Repeat the above process to construct k decision trees and form a random forest model.

[0091] In a random forest, for each decision tree, calculate the contribution of each feature to reducing the Gini coefficient when the node is split:

[0092]

[0093] Where: V i is the importance score of the target feature variable, k is the number of decision trees, t is the number of decision tree nodes, V ti is the target feature variable X i Contribution to reducing the Gini coefficient in the decision tree, N t is a set of nodes, G j is the Gini coefficient, G jL and G jR is the Gini coefficient of the two child nodes after splitting.

[0094] 3. Perform abnormal analysis on electrical parameters based on the ranking of importance scores of target feature variables.

[0095] According to the calculated feature importance score V i (i=1,2,…,n), sort the equipment operation status features. The higher the importance score of the feature, the greater the impact on the electrical parameter abnormality.

[0096] Finally, combined with the physical principles and actual conditions of equipment operation, the actual impact represented by each feature importance score is explained. For example, if the importance score of the device temperature feature is high, it means that changes in device temperature have a greater impact on electrical parameter anomalies, perhaps because excessive temperature causes changes in device resistance, which in turn affects electrical parameters.

[0097] S16. If not, determine the abnormal characteristic variables of the monitoring data transmission of each monitoring point in the distribution network, and perform abnormal analysis on the electrical parameters based on the abnormal characteristic variables.

[0098] In this embodiment, when it is determined that the screened device operation status data has no correlation with the electrical parameter anomaly, it is necessary to further consider whether it is associated with the data transmission anomaly. When it is determined that the electrical parameter anomaly is associated with the data transmission, the electrical parameter is analyzed for anomaly according to the data transmission anomaly characteristic variable. Specifically, 1. By analyzing the communication parameter monitoring, such as transmission delay, packet loss rate, and model strength and quality. When the electrical parameter anomaly is detected, the communication parameter data for a period of time before and after the moment, such as 5 minutes before the anomaly to 5 minutes after the anomaly occurs, is recorded at the same time. This includes the changes in transmission delay, packet loss rate, signal strength, etc., forming a time series data set for subsequent detailed analysis. 2. Analyze the impact of transmission delay on electrical parameter anomaly, draw a time series graph of transmission delay and electrical parameter anomaly values, and observe the time correlation between the delay peak and the occurrence of electrical parameter anomaly. If the delay peak always appears before the electrical parameter anomaly, or the delay change has an obvious synchronization relationship with the electrical parameter abnormal change, then the transmission delay may be an important factor leading to the judgment of electrical parameter anomaly. 3. Analyze the impact of packet loss on electrical parameter anomaly, and count the incidence of electrical parameter anomaly in the time period with high packet loss rate. If the rate of electrical parameter anomalies increases significantly during a period when the packet loss rate exceeds a certain threshold (such as 10%), then packet loss is likely to be one of the causes of the electrical parameter anomalies. In addition, the electrical parameter data before and after packet loss can be restored and simulated. For example, the lost data points can be restored through interpolation methods, and then the electrical parameters are re-analyzed to see if they are still abnormal, so as to determine the extent of the impact of packet loss on the abnormality judgment.

[0099] By analyzing abnormal conditions such as transmission delay, packet loss rate, signal strength and quality in data communication transmission, it is possible to accurately determine whether they lead to abnormal electrical parameters, effectively avoid misjudgment of electrical parameters due to communication problems, and improve the accuracy and reliability of the evaluation of the operating status of the electrical system.

[0100] Embodiment 2

[0101] An embodiment of a distribution network abnormality monitoring system for the Internet of Things in the present invention includes the following steps:

[0102] The data acquisition and preprocessing unit is used to collect abnormal data of electrical parameters, weather data and geographical location information data from various monitoring points of the distribution network, and perform data preprocessing on the collected data;

[0103] A CNN-RNN model evaluation unit is used to input the preprocessed data into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or geographical location factors;

[0104] A target characteristic variable determination unit, used to determine the target characteristic variables of the operating status of the equipment at each monitoring point of the distribution network when the electrical parameters are not affected by weather factors or geographical location factors;

[0105] A target feature vector judgment unit, used to judge whether the target feature vector meets the abnormal condition;

[0106] The electrical parameter anomaly analysis unit is used to perform an anomaly analysis on the electrical parameters by monitoring the abnormal data of the operating status of the equipment at each monitoring point of the distribution network when the target feature vector meets the abnormal condition.

[0107] For the specific definition of the system, please refer to the definition of the method above, which will not be repeated here. Each module in the above system can be implemented in whole or in part by software, hardware and a combination thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

[0108] Embodiment 3

[0109] The present invention provides a computer device, comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor. When the processor executes the computer-readable instructions, the steps of the above method are implemented.

[0110] Those of ordinary skill in the art will appreciate that the units of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0111] In the embodiments provided by the present invention, it should be understood that the division of units is only a logical function division, and there may be other division methods in actual implementation, such as multiple units can be combined into one unit, one unit can be split into multiple units, or some features can be ignored. In addition, each functional unit in each embodiment of the present invention can be integrated into a processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units.

[0112] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, read-only memory (ROM, Read-0nlyMemory), random access memory (RAM, RandomAccessMemory), mobile hard disk, magnetic disk or optical disk, etc., which can store program code.

[0113] It can be understood that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or replace some or all of the technical features therein by equivalents. These modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present invention, and they should all be included in the scope of the claims and specification of the present invention.

Claims

1. A distribution network abnormality monitoring method for the Internet of Things, characterized in that: include: Collect abnormal data of electrical parameters, weather data and geographic location information data from each monitoring point of the distribution network, and perform data preprocessing on the collected data; The preprocessed data is input into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors; When the electrical parameters are not affected by weather factors or geographical location factors, the target characteristic variables of the operating status of the equipment at each monitoring point of the distribution network are determined; Determining whether the target feature vector satisfies an abnormal condition; If so, an abnormal analysis is performed on the electrical parameters by monitoring the abnormal data of the operating status of the equipment at each monitoring point of the distribution network.

2. The method for monitoring abnormal distribution network for the Internet of Things according to claim 1, characterized in that: The pre-processed data is input into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors, including: Obtain historical abnormal data, weather data, and geographic location information of electrical parameters at each monitoring point in the distribution network within a preset time, and normalize the data The normalized data is used as the first input sample and input into the convolutional neural network to obtain the corresponding feature map sequence; Input the feature graph sequence as a second input sample into a recurrent neural network to perform learning and training on the model; The binary cross entropy loss function and optimization algorithm are used to evaluate and optimize the model to obtain the trained CNN-RNN model.

3. The method for monitoring abnormal distribution network for the Internet of Things according to claim 2, characterized in that: The normalized data is used as the first input sample and input into the convolutional neural network to obtain the corresponding feature map sequence, including: The normalized electrical parameter anomaly data, weather data, and geographic location information data are concatenated in the channel dimension to obtain an input tensor; Use multiple convolution kernels to perform a convolution operation on the input tensor, apply an activation function to the output, and add a pooling layer to reduce the output data dimension to obtain a target feature map sequence.

4. The method for monitoring abnormal distribution network for the Internet of Things according to claim 3 is characterized in that: The using multiple convolution kernels to perform a convolution operation on the input tensor includes: Where: y i,k,m is the output of each i-th sample, k-th time step, and m-th convolution kernel, w m,a,b,j is the convolution kernel weight, b m is the bias term.

5. The method for monitoring abnormality of distribution network for the Internet of Things according to claim 2, characterized in that: The step of taking the feature graph sequence as a second input sample and inputting it into a recurrent neural network to perform learning and training on the model includes: Initializing model parameters of a recurrent neural network, and inputting the feature graph sequence into the recurrent neural network as a second output sample; LSTM unit calculations are performed separately, including forget gate, input gate, candidate cell state, cell state update, output gate, and hidden state update.

6. The method for monitoring abnormality of distribution network for the Internet of Things according to claim 1, characterized in that: If so, an abnormal analysis of the electrical parameters is performed by monitoring the abnormal data of the operating status of the equipment at each monitoring point of the distribution network, including: Constructing a random forest model, and determining sample data of the model according to the target characteristic variables and electrical parameter abnormal data; Calculate the contribution of each feature of each decision tree in the random forest model to reducing the Gini coefficient when the node is split, and obtain the importance score of the target feature variable; The electrical parameters are analyzed for abnormality according to the order of importance scores of the target feature variables.

7. The method for monitoring abnormality of distribution network for the Internet of Things according to claim 6, characterized in that: The step of respectively calculating the contribution of each feature of each decision tree in the random forest model to reducing the Gini coefficient when the node is split, and obtaining the importance score of the target feature variable, includes: Where: V i is the importance score of the target feature variable, k is the number of decision trees, t is the number of decision tree nodes, V ti is the target feature variable X i Contribution to reducing the Gini coefficient in the decision tree, N t is a set of nodes, G j is the Gini coefficient, G jL and G jR is the Gini coefficient of the two child nodes after splitting.

8. The method for monitoring abnormality of distribution network for the Internet of Things according to claim 1, characterized in that: The determining whether the target feature vector satisfies an abnormal condition includes: If not, then determine the abnormal characteristic variables of the monitoring data transmission of each monitoring point of the distribution network, and perform abnormal analysis on the electrical parameters according to the abnormal characteristic variables.

9. A distribution network abnormality monitoring system for the Internet of Things, characterized in that: include: The data acquisition and preprocessing unit is used to collect abnormal data of electrical parameters, weather data and geographical location information data from various monitoring points of the distribution network, and perform data preprocessing on the collected data; A CNN-RNN model evaluation unit, used for inputting the preprocessed data into the trained CNN-RNN model to evaluate whether the electrical parameters are affected by weather factors or by geographical location factors; A target characteristic variable determination unit, used for determining the target characteristic variable of the operating state of the equipment at each monitoring point of the distribution network when the electrical parameter is not affected by weather factors or geographical location factors; A target feature vector judgment unit, used to judge whether the target feature vector meets an abnormal condition; The electrical parameter abnormality analysis unit is used to perform abnormality analysis on the electrical parameters by monitoring abnormal data of the operating status of equipment at each monitoring point of the distribution network when the target feature vector meets the abnormal condition.

10. A computer device comprising a memory, a processor, and computer-readable instructions stored in the memory and executable on the processor, characterized in that: When the processor executes the computer-readable instructions, the steps of the distribution network anomaly monitoring method for the Internet of Things as described in any one of claims 1 to 8 are implemented.

Images