Wireless sensing power distribution network intelligent online monitoring method and system

Abstract

The application provides a wireless sensing power distribution network intelligent online monitoring method and system, relates to the technical field of power distribution network monitoring, and comprises the following steps: deploying a wireless sensing module on a target power distribution network, wherein the wireless sensing module supports a low-power wide-area network protocol; extracting a monitoring data sequence set, identifying a health state by using a health state evaluator, and determining a first resource allocation coefficient set according to an identification result; calling a historical health state identification result sequence set, identifying health degradation, and determining a second resource allocation coefficient set; performing resource allocation optimization, obtaining an updated wireless sensing module, and using the updated wireless sensing module to perform intelligent online monitoring on the target power distribution network. The application solves the technical problem that, in the prior art, resource allocation is often based on fixed standards and preset schemes, cannot flexibly cope with the actual needs of different monitoring areas, and further leads to unreasonable resource allocation and affects the monitoring effect.

Description

Technical Field

[0001] This invention relates to the field of power distribution network monitoring technology, and specifically to a method and system for intelligent online monitoring of power distribution networks using wireless sensing. Background Technology

[0002] As a crucial component of the power system, the distribution network is responsible for distributing electrical energy from substations to end users. With increasing electricity demand and the continuous expansion of the power grid, the monitoring and management of the distribution network have become increasingly complex. In existing distribution network monitoring technologies, resource allocation is often based on fixed standards and preset schemes, lacking the ability to adjust resources according to real-time data and historical trends. For example, equipment in some areas may require more resources due to increased load or accumulated faults, while other areas may require fewer resources due to healthy equipment. The fixed resource allocation model of existing technologies makes it impossible to flexibly respond to the actual needs of different monitoring areas, leading to unreasonable resource allocation, failure to maximize monitoring effectiveness, and preventing the distribution network from achieving optimal management under limited resources. Summary of the Invention

[0003] This application provides a wireless sensing-based intelligent online monitoring method and system for distribution networks, aiming to solve the technical problem that in existing distribution network monitoring technologies, resource allocation is often based on fixed standards and preset schemes, which cannot flexibly meet the actual needs of different monitoring areas, thus leading to unreasonable resource allocation and affecting the monitoring effect.

[0004] The first aspect of this application discloses a method for intelligent online monitoring of a distribution network using wireless sensing. The method includes: deploying a wireless sensing module on a target distribution network, wherein the wireless sensing module supports a low-power wide-area network protocol; extracting a set of monitoring data sequences from the wireless sensing module within a preset monitoring window; using a pre-built health status evaluator to identify the health status of the monitoring data sequence set; and determining a first set of resource allocation coefficients based on the identification results; retrieving a set of historical health status identification result sequences from the wireless sensing module within a historical monitoring window; identifying health decay in the historical health status identification result sequence set; and determining a second set of resource allocation coefficients; optimizing resource allocation for the wireless sensing module based on the first set of resource allocation coefficients and the second set of resource allocation coefficients to obtain an updated wireless sensing module; and using the updated wireless sensing module to perform intelligent online monitoring of the target distribution network.

[0005] In a second aspect, the application discloses a wireless-sensing power distribution network intelligent online monitoring system, which is used for the wireless-sensing power distribution network intelligent online monitoring method, and comprises a deployment module, a health state identification module, a health degradation identification module and an intelligent online monitoring module.

[0006] The one or more technical solutions provided in the application have at least the following beneficial effects:

[0007] By deploying the wireless sensing module supporting the low-power wide-area network protocol, long-time and low-energy operation of the target power distribution network monitoring is achieved, which means that the monitoring equipment of the target power distribution network can continuously operate in a wide area without frequent battery replacement or power supply, thereby improving the sustainability and maintenance efficiency of the system; the health state of the target power distribution network is evaluated in real time by the health state evaluator, and the first resource allocation coefficient set is generated based on the evaluation result, which enables the system to dynamically allocate resources according to the current health state of the target power distribution network, so as to ensure that more resources are provided to the area with poor health state, thereby improving the efficiency of monitoring and maintenance; the historical health state identification result sequence set is called and health degradation identification is performed, so that the trend of the health state of the target power distribution network is predicted, that is, not only the current health state is concerned, but also potential faults or areas with gradually deteriorating health state are identified by analyzing the historical degradation, and the second resource allocation coefficient set generated based on the analysis can provide more resources to the area with degraded or about-to-degraded health state, so as to prevent faults in advance; the first resource allocation coefficient set and the second resource allocation coefficient set are combined to perform resource allocation optimization, and the updated wireless sensing module is finally obtained, which ensures that the key part of the target power distribution network is reasonably and effectively configured with resources, and the target power distribution network is more efficiently and accurately monitored by the updated wireless sensing module, thereby improving the operation efficiency of the target power distribution network and ensuring the stability and reliability of the system.

[0008] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0009] Figure 1 This is a schematic flowchart of a wireless sensing-based intelligent online monitoring method for power distribution networks provided in an embodiment of this application.

[0010] Figure 2 A schematic diagram of the structure of a wireless sensing-based intelligent online monitoring system for power distribution networks provided in this application embodiment.

[0011] Figure labeling: Deployment module 10, health status recognition module 20, health decline recognition module 30, intelligent online monitoring module 40. Detailed Implementation

[0012] This application provides a wireless sensing-based intelligent online monitoring method and system for power distribution networks. It solves the technical problem that in existing power distribution network monitoring, resource allocation is often based on fixed standards and preset schemes, which cannot flexibly meet the actual needs of different monitoring areas, resulting in unreasonable resource allocation and affecting the monitoring effect.

[0013] After introducing the basic principles of this application, various non-limiting embodiments of this application will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit this application.

[0014] Example 1, as Figure 1 As shown in the figure, this application provides a method for intelligent online monitoring of power distribution networks using wireless sensing. The method includes:

[0015] A wireless sensing module is deployed on the target power distribution network, wherein the wireless sensing module supports low-power wide-area network protocols.

[0016] Key nodes of the target distribution network, such as substations, switching equipment, and load centers, are identified. These nodes are the core components of the distribution network. Wireless sensors are deployed at the identified key nodes to form wireless sensing modules, which are used to collect relevant monitoring data of the target distribution network. The wireless sensing modules support low-power wide-area network protocols, which means that these wireless sensors can operate for a long time and can transmit data through low-power wireless networks. This design makes them suitable for distribution networks with a wide monitoring range and a large number of devices, without the need for frequent battery replacements or charging.

[0017] extracting a monitoring data sequence set of the wireless sensing module in a preset monitoring window, performing health state identification on the monitoring data sequence set by using a pre-constructed health state evaluator, and determining a first resource allocation coefficient set according to an identification result.

[0018] The preset monitoring window is a pre-set data collection period. The wireless sensing module collects data according to the preset monitoring window, and obtains a monitoring data sequence set including at least voltage amplitude, voltage fluctuation frequency, current amplitude, overload current, active power, reactive power and load data. The health state evaluator is a pre-trained model, which can analyze the monitoring data sequence set and judge the operation state of the power grid, and output an identification result of the health state. According to the identification result of the health state evaluator, the first resource allocation coefficient set is calculated. The first resource allocation coefficient set is adjusted according to the health state of the power grid, and determines how to optimize resource allocation, so as to ensure that the most critical or most concerned part obtains more resources and monitoring.

[0019] retrieving a historical health state identification result sequence set of the wireless sensing module in a historical monitoring window, performing health attenuation identification on the historical health state identification result sequence set, and determining a second resource allocation coefficient set.

[0020] The historical health state data recorded by the previous wireless sensing module is retrieved, and the data comes from the historical monitoring window, that is, the monitoring results in a certain period of time in the past. The historical health state identification result sequence set refers to the results after the health state evaluation in the past, including the output results of various health state identification analyses performed on the target distribution network in the historical monitoring window.

[0021] The health attenuation identification is an analysis on the historical health state identification result sequence set to identify the change trend of the system health state, that is, to analyze the health state change in the historical data, so as to identify the degradation mode or aging trend of the distribution network equipment or module. Specifically, according to the historical health state identification result sequence set, a health attenuation coefficient set is calculated. The health attenuation coefficient set reflects the health degradation degree of the target distribution network. The ratio of each health attenuation coefficient to the sum of the entire health attenuation coefficient set is calculated to obtain the second resource allocation coefficient set.

[0022] Based on the first resource allocation coefficient set and the second resource allocation coefficient set, the resource allocation optimization of the wireless sensing module is performed to obtain an updated wireless sensing module, and the intelligent online monitoring of the target distribution network is performed by using the updated wireless sensing module.

[0023] The first resource allocation coefficient set and the second resource allocation coefficient set are weighted to combine their effects and ensure that resources are allocated reasonably based on the current health status and historical health degradation. By weighting mapping the two coefficient sets, a comprehensive resource allocation coefficient set is generated. The selection of the weighting coefficient can be adjusted according to actual needs to ensure optimal decision-making for resource allocation under the current health status and the influence of historical health degradation.

[0024] After obtaining the target resource allocation coefficient set, resource allocation optimization is performed on the wireless sensing module based on the set. This optimization process is based on the calculated target resource allocation coefficient to ensure that the resources (such as computing power, battery capacity, bandwidth, etc.) allocated to each sensor, module, or device maximize the real-time monitoring needs of the power distribution network. After resource allocation optimization, the configuration or parameters of the wireless sensing module are adjusted according to the optimization results, including adjustments to the module hardware, configuration updates, or the introduction of new functions to ensure that it can effectively support updated monitoring tasks. The updated wireless sensing module is used to perform intelligent online monitoring of the target power distribution network. At this time, not only can resource allocation be adjusted based on real-time data feedback, but also more accurate and long-term monitoring and prediction functions can be provided through the results of historical health status and degradation identification, ensuring stable operation of the power distribution network.

[0025] Further, the wireless sensing module is deployed for the target power distribution network, wherein the wireless sensing module supports a low-power wide-area network protocol, including:

[0026] A set of key nodes of the target power distribution network is obtained, and wireless sensors are arranged for the key nodes to obtain a set of wireless sensors. The set of key nodes is sampled using the set of wireless sensors to obtain a set of trial sampling data. Data quality analysis is performed on the set of trial sampling data, and wireless sensors that do not meet the requirements of the data quality analysis are identified to obtain a set of abnormal wireless sensors. Sampling enhancers are associated with the set of abnormal wireless sensors, and the set of abnormal wireless sensors and a set of normal wireless sensors in the set of wireless sensors are aggregated to obtain the wireless sensing module.

[0027] Modeling the target power distribution network, analyzing the topology of the target power distribution network, and identifying various key nodes, including but not limited to substations, distribution lines, switchgear, load centers, monitoring points, etc., which are the key nodes for power system monitoring and control. After determining the key nodes, wireless sensors are deployed for each key node. The wireless sensors are responsible for collecting real-time data from the target power distribution network, monitoring current, voltage, power, temperature, and other types of information. When deploying, the energy efficiency, signal transmission stability, and adaptability of the wireless sensors need to be considered. After the deployment of wireless sensors, a set of wireless sensors is obtained, which will be the data source of the entire monitoring system, ensuring that relevant data can be collected from each key node in real time.

[0028] Trial sampling is a process of preliminary testing and data collection for the working performance of each wireless sensor, including collecting a certain amount of data, such as current, voltage, power, etc., from each key node where a wireless sensor is deployed. The collected trial sampling data form a trial sampling data set.

[0029] Data quality analysis is performed on the trial sampling data set, including checking for missing or missing values, identifying and marking cases where wireless sensors do not collect data at certain times, checking if the data is within the expected range, such as voltage, current, etc., if it is outside the expected range, it needs to be marked, detecting outliers in the data through statistical methods or machine learning algorithms, such as sudden changes, excessive fluctuations, etc., which may be caused by sensor failure or external interference, checking for excessive noise or interference in the data, which may affect the accuracy of subsequent analysis. After data quality analysis, identify the part of the wireless sensor that has problems in collecting data, and mark it as an abnormal wireless sensor set.

[0030] For the identified abnormal wireless sensor set, associate a sampling enhancer to enhance its sampling capability. The purpose of the sampling enhancer is to improve the performance of the abnormal wireless sensor set, reduce noise and inaccuracy in the data, and ensure the quality and reliability of the monitoring data. The normal wireless sensor set is a set of sensors that have passed data quality analysis and have not been identified as abnormal. The monitoring data provided by these sensors is considered accurate and reliable. The enhanced abnormal wireless sensor set and the normal wireless sensor set are combined to form a wireless sensor module, which is used for comprehensive intelligent online monitoring of the target power distribution network.

[0031] Further, the abnormal wireless sensor set is associated with a sampling enhancer, including:

[0032] A sampling enhancer is obtained, wherein the sampling enhancer includes a depthwise separable convolutional branch and a low-resolution self-attention branch; the depthwise separable convolutional branch includes channel-wise convolutional layers and pointwise convolutional layers for capturing local details; the low-resolution self-attention branch performs self-attention calculation by downsampling the input data to a low-resolution space, upsampling the calculation result, and performing residual concatenation between the upsampled result and the input data to output sample-enhanced data for global analysis.

[0033] The goal of a sampling enhancer is to remove noise and improve the accuracy and reliability of data by processing data from anomalous sensors, so that even poor-quality sensors can provide useful data. Sampling enhancers include depthwise separable convolution branches and low-resolution self-attention branches.

[0034] The depthwise separable convolutional branch includes channel-wise convolutional layers and pointwise convolutional layers. The input is assumed to be a multi-channel feature map of size H×W×C. in Where H and W are the height and width respectively, and C in This refers to the number of input channels. Each channel-wise convolutional layer uses a K×K convolutional kernel (typically 3×3 or 5×5) to perform convolution operations on each input channel, resulting in an output feature map of size H×W×C. in The pointwise convolutional layer uses a 1×1 convolutional kernel to convolve all channels, generating a new output feature map of size H×W×C. out C out This is the number of output channels.

[0035] The low-resolution self-attention branch enhances global relationships in data through a self-attention model. Specifically, it downsamples the input data, reducing its resolution to decrease computational complexity while capturing global information without being distracted by local details. Self-attention computation is performed in this low-resolution space, allowing the model to focus on different parts of the input data and capture global information. The self-attention model calculates relationships between data points, enhancing long-range dependencies. After the self-attention computation, the results are upsampled to restore the resolution to the same as the original input data. This process helps preserve global information without losing details. Finally, the upsampled self-attention results are residually concatenated with the input data to ensure that the original input information is preserved during the enhancement process. The self-attention results effectively supplement and enhance the input data. This low-resolution self-attention branch enhances the ability to capture global features, especially when identifying global trends or long-term changes. Through the self-attention mechanism, it focuses more on key parts of the data, improving global analysis capabilities.

[0036] Further, the extraction of the monitoring data sequence set of the wireless sensing module in the preset monitoring window includes:

[0037] Obtaining a preset monitoring index set, wherein the preset monitoring index set at least includes voltage amplitude, voltage fluctuation frequency, current amplitude, overload current, active power, reactive power and load data; monitoring the target power distribution network according to the preset monitoring index set through the wireless sensing module in the preset monitoring window to obtain the monitoring data sequence set.

[0038] The preset monitoring index set is a set of multiple power system key parameters, which are used to monitor and evaluate the running state of the power distribution network in real time. Through these monitoring indexes, it can be judged whether the target power distribution network is running normally, and an alarm can be sent in time when an abnormality occurs.

[0039] Among them, the voltage amplitude represents the size of the voltage in the power grid, and too high or too low voltage amplitude may cause equipment damage or system instability; voltage fluctuation is a common phenomenon in power grid operation, and monitoring voltage fluctuation frequency helps to judge whether the power grid exists the risk of voltage instability; current amplitude represents the current intensity in the power distribution network, which is used to monitor the system load and current overload; when the current exceeds the rated carrying capacity of the equipment, the power grid may be overloaded, and monitoring the overload current helps to prevent power equipment damage caused by overload; active power is the actual transmission of electric energy in the power distribution network, and monitoring active power can evaluate the power supply capacity and efficiency of the power grid; reactive power is related to the electromagnetic field generation of the power system, although it does not directly participate in doing work, but the monitoring of reactive power helps the stability of the system and the quality of power transmission; load data indicates the power consumption of each part of the power distribution network, which is used to judge the load balancing and timely find overload, fault and other problems.

[0040] The preset monitoring window is a pre-set time range for collecting data, which can be adjusted according to actual needs, such as several minutes, several hours, or longer, depending on the frequency of monitoring and data processing capacity. In the preset monitoring window, all the required monitoring data is collected by the wireless sensing module according to the preset monitoring index set to obtain the monitoring data sequence set.

[0041] Further, the health state of the monitoring data sequence set is identified by using a pre-constructed health state evaluator, and a first resource allocation coefficient set is determined according to the identification result, including:

[0042] perform semantic conversion on the monitoring data sequence set respectively to obtain a monitoring state semantic information sequence set; a first monitoring state semantic information sequence is randomly extracted from the monitoring state semantic information sequence set; a first monitoring state semantic information in the first monitoring state semantic information sequence is subjected to health identification by using the health state evaluator to obtain a first health state coefficient; the health state evaluator is used to perform health state identification on the monitoring state semantic information sequence set respectively to obtain a health state coefficient sequence set; a mean value of each health state coefficient sequence in the health state coefficient sequence set is calculated to obtain a health state coefficient mean value set, the health state coefficient mean value set is taken as an identification result, and a first resource allocation coefficient set is determined based on the identification result.

[0043] The monitoring data sequence set is numerical data, and these raw data need to be converted into descriptive sentences with practical significance in order to more easily understand the state of the power grid. The goal of semantic conversion is to convert these numerical data into descriptive sentences by combining with the power database, which can clearly convey the operating state of the power grid, such as whether the voltage exceeds the standard range, whether the current is overloaded, etc. The power database contains some pre-defined standard values and ranges, such as standard voltage range, current load range, etc. By comparing the monitoring data sequence set with the corresponding standard values and ranges, the monitoring data is converted into descriptive sentences such as "voltage exceeds standard voltage range", "voltage is too high", "voltage is too low", etc. Through semantic conversion, a monitoring state semantic information sequence set is obtained, and each element in this set corresponds to a descriptive sentence of a monitoring data point, which represents the operating state of the power grid at different time periods.

[0044] A first monitoring state semantic information sequence is randomly extracted from the monitoring state semantic information sequence set as the current analysis object, ensuring comprehensive analysis of different situations.

[0045] The first monitoring state semantic information in the extracted first monitoring state semantic information sequence is subjected to health identification by the health state evaluator, which is a pre-trained model that can judge the health state of the distribution network based on monitoring data and historical experience. The process of health identification judges whether the target distribution network is in a normal state or has a potential fault by evaluating the descriptive information of each monitoring state. According to predetermined standards, each descriptive sentence is matched and analyzed to generate a first health state coefficient. The first health state coefficient reflects the health status corresponding to the monitoring state and is a numerical value representing the health status of the equipment or system. If the monitoring state indicates that the power grid is operating normally, the health state coefficient is higher; if the state indicates a problem, such as high voltage, current overload, etc., the health state coefficient is lower, indicating poor health status.

[0046] The health state evaluator acts on all monitoring state semantic information in the monitoring state semantic information sequence set for health state recognition. For each monitoring state semantic information, the health state evaluator makes a judgment according to the nature of the statement to generate a health state coefficient sequence set containing health state coefficients of all monitoring states.

[0047] The mean value of each health state coefficient sequence in the health state coefficient sequence set is calculated. Each health state coefficient sequence corresponds to a monitoring state in a certain period of time and is composed of multiple health state coefficients. The average value of each health state coefficient sequence is calculated to reflect the overall health level of the power distribution network in that period of time. The health state coefficient mean value set is obtained by calculating the mean value of each health state coefficient sequence.

[0048] For each mean value in the health state coefficient mean value set, the ratio between the mean value and the sum of all mean values is calculated. Through this ratio, the resource demand corresponding to each monitoring state can be determined. The difference between each ratio and 1 is calculated to obtain the resource allocation coefficient of the ratio. This resource allocation coefficient is used to adjust the allocation of resources. For example, if the ratio corresponding to a certain health state coefficient mean value is large, i.e., the health state is good, the resource allocation coefficient of that node will be low, and vice versa. The node with a poor health state will obtain more resources. Finally, based on the calculated difference, a resource allocation coefficient is determined for each node. These resource allocation coefficients will determine how to optimize resource allocation in the target power distribution network to ensure that problem areas are given priority resources.

[0049] Further, the health state evaluator is used to perform health recognition on the first monitoring state semantic information in the first monitoring state semantic information sequence to obtain a first health state coefficient, which includes:

[0050] The health state evaluator is used to perform double-stage analysis on the first monitoring state semantic information in the first monitoring state semantic information sequence to obtain a first monitoring state semantic memory vector. The health state evaluator is used again to perform double-stage analysis on the second monitoring state semantic information in the first monitoring state semantic information sequence in combination with the first monitoring state semantic memory vector to obtain a second monitoring state semantic memory vector. In this way, a first target monitoring state semantic memory vector is obtained. The health state evaluator is called to perform health state recognition on the first target monitoring state semantic memory vector to determine a first health state coefficient.

[0051] The two-stage analysis is a multi-step data analysis technique. When processing the monitoring state semantic information, the health state evaluator enhances the understanding and memory of the data through a two-stage analysis process. The first monitoring state semantic information refers to the first descriptive sentence in the first monitoring state semantic information sequence, which represents a real-time monitoring state of the power distribution network. The health state evaluator first performs two-stage analysis on the first monitoring state semantic information to obtain a first monitoring state semantic memory vector. The first monitoring state semantic memory vector is a vector generated by analyzing the current state and context information, representing the memory of the state. This memory vector contains the deep features of the current state, enabling the model to combine historical data and current data to gain a deeper understanding of the state.

[0052] After completing the analysis of the first monitoring state semantic information and obtaining the first monitoring state semantic memory vector, the analysis of the second monitoring state semantic information (such as the voltage value or current value at the next monitoring time) is continued. The key of this step is to combine the first monitoring state semantic memory vector and use two-stage analysis to analyze the second monitoring state semantic information. This analysis method not only involves the second monitoring state semantic information itself, but also considers the historical information (i.e., the memory vector) of the first monitoring state semantic information to enhance the accuracy of the analysis and help the model understand each independent monitoring state and their relationships from a global perspective.

[0053] The two-stage analysis of subsequent monitoring state semantic information is continued, and the monitoring state semantic memory vector of each monitoring state semantic information is generated in turn to generate a first target monitoring state semantic memory vector. This vector is a deep understanding of the entire first monitoring state semantic information sequence, integrating the influence of historical states and current states.

[0054] The health state recognition network layer in the health state evaluator is called to perform the final health state evaluation. The health state recognition network layer is a trained deep learning network that can judge the health state of the target power distribution network based on the input semantic memory vector. It evaluates the current health condition of the power grid through the recognition network and finally obtains a first health state coefficient. This coefficient reflects the overall health state of the target power distribution network within a given monitoring period. If the health state is good, the first health state coefficient is high; if the power grid has abnormalities such as high voltage or current overload, the first health state coefficient is low.

[0055] Further, the two-stage analysis of the first monitoring state semantic information in the first monitoring state semantic information sequence using the health state evaluator obtains a first monitoring state semantic memory vector, including:

[0056] The first-order keyword matching branch of the health state evaluator is called to extract keywords from the first monitoring state semantic information, obtaining a first keyword set. The first keyword set is extracted from the sentence in which it is located in the first monitoring state semantic information, obtaining a first key sentence set. The first keyword set and the first key sentence set are added to the initially empty vector, obtaining a first monitoring state semantic memory vector.

[0057] The first-order keyword matching branch is a module in the health state evaluator that extracts keywords from the monitoring state semantic information. The first-order keyword matching branch stores keywords related to abnormal health data, which represent specific abnormal states such as high voltage and current overload. Keyword extraction is used to extract keywords related to these abnormal states from the first monitoring state semantic information sequence. By recognizing these keywords, the monitoring state semantic information can be effectively matched with known health abnormal patterns, thus more accurately determining the health state of the power grid. During keyword extraction, the first monitoring state semantic information is matched with stored abnormal health keywords to identify keywords related to abnormal states. For example, if the first monitoring state semantic information contains "voltage" and "overhigh", these words will be extracted to form the first keyword set.

[0058] Second-order extraction is used to further extract sentences related to keywords from the first monitoring state semantic information. Second-order extraction not only focuses on keywords, but also analyzes sentences containing these keywords to understand more detailed context and meaning. Based on the extracted first keyword set, sentences containing these keywords are extracted, aiming to identify complete context rather than isolated keywords. For example, the sentence "voltage is too high, exceeding the safety range" contains not only "voltage" and "overhigh", but also "exceeding the safety range", which is crucial for subsequent health assessment. After second-order extraction, the first key sentence set is obtained, which contains all relevant sentences.

[0059] The first keyword set and the first key sentence set represent keywords extracted from the monitoring state and the content of the related sentences, respectively. By combining the two, a more comprehensive indication of the current monitoring state can be obtained. The first monitoring state semantic memory vector is a multi-dimensional vector representing the first monitoring state semantic information, which not only contains the surface description of the monitoring state (keywords), but also includes the deep meaning of the state (key sentences). In this way, a memory of the monitoring state is established, serving as a basis for subsequent health state evaluation.

[0060] Further, the historical health state recognition result sequence set of the wireless sensing module in the historical monitoring window is called, health attenuation recognition is performed on the historical health state recognition result sequence set, and a second resource allocation coefficient set is determined, including:

[0061] A health attenuation coefficient set of the historical health state recognition result sequence set is calculated respectively; each health attenuation coefficient in the health attenuation coefficient set is divided by the sum of the health attenuation coefficient set to obtain a second resource allocation coefficient set.

[0062] The health attenuation coefficient reflects the health state degradation degree of the power distribution network in a historical monitoring time, and represents the change of the health state of the power distribution network from the past to the present. The change of the health attenuation coefficient reflects whether the equipment of the power distribution network has an aging or gradual failure trend. For the historical health state recognition result sequence, a corresponding health attenuation coefficient set is calculated. The health attenuation coefficient is calculated by comparing the health state change in the historical health state recognition result. For example, if the health state of the power distribution network is good, that is, the health coefficient is high, and a significant decrease occurs after a period of time, the attenuation coefficient of the time period will be larger. In the calculation process, the fluctuation of the historical health state, such as the decrease amplitude of the health coefficient, is used to determine the health attenuation coefficient. The larger the value of the health attenuation coefficient, the worse the health state of the equipment, and the more serious the degradation.

[0063] The sum of all health attenuation coefficients in the health attenuation coefficient set is calculated to measure the overall degree of power grid health degradation in all historical monitoring windows. Each health attenuation coefficient is divided by the sum of the health attenuation coefficient set. In this way, each health attenuation coefficient can be converted into a ratio. This ratio reflects the proportion of the degradation degree of each monitoring window in the entire historical monitoring period. These ratios constitute a second resource allocation coefficient set, which is used for subsequent resource allocation optimization.

[0064] Further, the method comprises:

[0065] The first resource allocation coefficient set and the second resource allocation coefficient set are mapped and weighted to determine a target resource allocation coefficient set; and resource allocation optimization is performed on the wireless sensing module based on the target resource allocation coefficient set to obtain an updated wireless sensing module.

[0066] The first resource allocation coefficient set and the second resource allocation coefficient set are weighted mapped, the purpose of the weighted mapping is to combine the current power grid health condition and the historical degradation trend to optimize the resource allocation, the weighted mapping process is adjusted based on the relative importance of the two, ensuring that resources are preferentially allocated to those parts that both have current health problems and have degradation trends, through the weighted mapping, a comprehensive target resource allocation coefficient set is obtained, which reflects the resource demand of each power distribution network node or wireless sensor after combining the current health status and the historical degradation trend, a high value of the target resource allocation coefficient set indicates that the health status of the node or wireless sensor is poor, and more resources should be preferentially allocated; a low value indicates that the part is in a better state and has less resource demand.

[0067] According to the target resource allocation coefficient set, the resource allocation of the wireless sensing module is optimized, that is, the allocation of resources is adjusted according to the resource demand of each node to ensure that the monitoring task can be preferentially concentrated on the nodes with poor health or faster degradation, through the resource allocation optimization, an updated wireless sensing module is generated, and the updated wireless sensing module can more efficiently monitor the target power distribution network, especially in the case of health problems or degradation signs of the power grid, more resources can be preferentially mobilized for monitoring and maintenance.

[0068] In summary, the wireless sensing power distribution network intelligent online monitoring method provided by the embodiments of the present application has the following technical effects:

[0069] By deploying wireless sensing modules supporting low-power wide-area network protocols, long-term and low-energy operation of target power distribution network monitoring is achieved, which means that the monitoring equipment of the target power distribution network can continuously operate in a wide area without frequent battery replacement or power supply, improving the sustainability and maintenance efficiency of the system; the health state evaluator identifies the health state of the wireless sensing module within the preset monitoring window, which can assess the health status of the target power distribution network in real time, and generate a first resource allocation coefficient set based on the assessment result, which enables the system to dynamically allocate resources based on the current health status of the target power distribution network, ensuring that more resources are allocated to areas with poor health status, thereby improving the efficiency of monitoring and maintenance; by retrieving a historical health state identification result sequence set and performing health decay identification, the trend of the health status of the target power distribution network is predicted, that is, not only the current health status is concerned, but also potential faults or gradually deteriorating areas are identified by analyzing historical degradation, and a second resource allocation coefficient set generated based on this analysis can provide more resource support to areas with degraded or about to degrade health status, preventing faults in advance; by combining the first and second resource allocation coefficient sets, resource allocation optimization is performed to obtain updated wireless sensing modules, which ensures that the key parts of the target power distribution network are reasonably and effectively allocated resources, and the target power distribution network is more efficiently and accurately monitored online by updating the wireless sensing modules, thereby improving the operation efficiency of the target power distribution network and ensuring the stability and reliability of the system.

[0070] In the second embodiment, based on the same inventive concept as the wireless-sensing-based intelligent online monitoring method of the power distribution network in the foregoing embodiments, as shown in the accompanying drawings, the embodiments of the present application provide a wireless-sensing-based intelligent online monitoring system for a power distribution network, which comprises: Figure 2

[0071] The deployment module 10 is configured to deploy wireless sensing modules on the target power distribution network, wherein the wireless sensing modules support low-power wide-area network protocols; the health state identification module 20 is configured to extract a monitoring data sequence set of the wireless sensing modules within a preset monitoring window, identify the health state of the monitoring data sequence set using a pre-constructed health state evaluator, and determine a first resource allocation coefficient set according to the identification result; the health decay identification module 30 is configured to retrieve a historical health state identification result sequence set of the wireless sensing modules within a historical monitoring window, identify the health decay of the historical health state identification result sequence set, and determine a second resource allocation coefficient set; the intelligent online monitoring module 40 is configured to optimize the resource allocation of the wireless sensing modules based on the first and second resource allocation coefficient sets, obtain updated wireless sensing modules, and perform intelligent online monitoring on the target power distribution network using the updated wireless sensing modules.​

[0072] Further, the deployment module 10 is configured to perform the following steps:

[0073] Obtaining a set of key nodes of the target power distribution network, respectively deploying wireless sensors on the key nodes to obtain a set of wireless sensors; respectively sampling the set of key nodes by using the set of wireless sensors to obtain a set of trial sampling data; performing data quality analysis on the set of trial sampling data, and identifying wireless sensors whose data quality analysis results do not meet requirements to obtain a set of abnormal wireless sensors; respectively associating the set of abnormal wireless sensors with sampling enhancers, and collecting the set of abnormal wireless sensors and a set of normal wireless sensors in the set of wireless sensors to obtain the wireless sensor module.

[0074] Further, the deployment module 10 is configured to perform the following steps:

[0075] Obtaining a sampling enhancer, wherein the sampling enhancer includes a depth separable convolution branch and a low-resolution self-attention branch; the depth separable convolution branch includes a channel-wise convolution layer and a point-wise convolution layer, and is configured to capture local details; the low-resolution self-attention branch is configured to perform self-attention calculation on input data by down-sampling the input data to a low-resolution space, up-sample the calculation result, and perform residual connection on the up-sampled result and the input data to output sampling enhancement data for global analysis.

[0076] Further, the health state identification module 20 is configured to perform the following steps:

[0077] Obtaining a set of preset monitoring indicators, wherein the set of preset monitoring indicators includes at least voltage amplitude, voltage fluctuation frequency, current amplitude, overload current, active power, reactive power, and load data; monitoring the target power distribution network according to the set of preset monitoring indicators in a preset monitoring window by using the wireless sensor module to obtain a set of monitoring data sequences.

[0078] Further, the health state identification module 20 is configured to perform the following steps:

[0079] respectively, to obtain a set of monitoring state semantic information sequences; a first monitoring state semantic information sequence is randomly extracted from the set of monitoring state semantic information sequences; a first monitoring state semantic information in the first monitoring state semantic information sequence is subjected to health identification by using the health state evaluator, to obtain a first health state coefficient; the set of monitoring state semantic information sequences is subjected to health state identification by using the health state evaluator respectively, to obtain a set of health state coefficient sequences; a mean value of each health state coefficient sequence in the set of health state coefficient sequences is calculated, to obtain a set of health state coefficient mean values, which is taken as an identification result, and based on which a first set of resource allocation coefficients is determined.

[0080] Further, the health state identification module 20 is configured to perform the following operation steps:

[0081] The first monitoring state semantic information in the first monitoring state semantic information sequence is subjected to two-stage analysis by using the health state evaluator, to obtain a first monitoring state semantic memory vector; the second monitoring state semantic information in the first monitoring state semantic information sequence is subjected to two-stage analysis by using the health state evaluator again in combination with the first monitoring state semantic memory vector, to obtain a second monitoring state semantic memory vector; in this way, a first target monitoring state semantic memory vector is obtained; the first target monitoring state semantic memory vector is subjected to health state identification by calling the health state identification network layer in the health state evaluator, to determine a first health state coefficient.

[0082] Further, the health state identification module 20 is configured to perform the following operation steps:

[0083] The first monitoring state semantic information in the first monitoring state semantic information sequence is subjected to keyword extraction by calling the first-order keyword matching branch in the health state evaluator, to obtain a first set of keywords; a first set of key sentences is obtained by performing two-stage extraction on a sentence in which the first set of keywords is located in the first monitoring state semantic information; the first set of keywords and the first set of key sentences are added into a vector initially empty, to obtain a first monitoring state semantic memory vector.

[0084] Further, the health state identification module 30 is configured to perform the following operation steps:

[0085] A set of health decay coefficients of the set of historical health state identification results sequences is calculated respectively; each health decay coefficient in the set of health decay coefficients is divided by a sum of the set of health decay coefficients, to obtain a second set of resource allocation coefficients.

[0086] Further, the intelligent online monitoring module 40 is configured to perform the following operation steps:

[0087] The first resource allocation coefficient set and the second resource allocation coefficient set are mapped and weighted to determine a target resource allocation coefficient set; and the wireless sensing module is optimized in resource allocation based on the target resource allocation coefficient set to obtain an updated wireless sensing module.

[0088] The foregoing detailed description of the power distribution network intelligent online monitoring method of wireless sensing has enabled those skilled in the art to clearly understand the power distribution network intelligent online monitoring system of wireless sensing in the embodiments. Since the system corresponds to the disclosed method, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0089] The foregoing description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Accordingly, the present application will not be limited to the embodiments shown herein but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for intelligent online monitoring of a power distribution network with wireless sensing, characterized in that, The method comprises: deploying a wireless sensing module on a target power distribution network, wherein the wireless sensing module supports a low-power wide-area network protocol; extracting a set of monitoring data sequences of the wireless sensing module in a preset monitoring window, identifying the health state of the set of monitoring data sequences by using a pre-constructed health state evaluator, and determining a first set of resource allocation coefficients based on the identification result, including: respectively converting the set of monitoring data sequences into a set of monitoring state semantic information sequences; randomly extracting a first monitoring state semantic information sequence from the set of monitoring state semantic information sequences; identifying the health state of the first monitoring state semantic information in the first monitoring state semantic information sequence by using the health state evaluator to obtain a first health state coefficient, including: performing double-stage analysis on the first monitoring state semantic information in the first monitoring state semantic information sequence by using the health state evaluator to obtain a first monitoring state semantic memory vector, including: calling a first-order keyword matching branch of the health state evaluator to extract keywords from the first monitoring state semantic information in the first monitoring state semantic information sequence to obtain a first keyword set; performing second-order extraction on the sentence in which the first keyword set is located in the first monitoring state semantic information to obtain a first key sentence set; adding the first keyword set and the first key sentence set into an initially empty vector to obtain a first monitoring state semantic memory vector; again performing double-stage analysis on the second monitoring state semantic information in the first monitoring state semantic information sequence by using the health state evaluator in combination with the first monitoring state semantic memory vector to obtain a second monitoring state semantic memory vector; in this way, a first target monitoring state semantic memory vector is obtained; identifying the health state of the first target monitoring state semantic memory vector by calling a health state identification network layer of the health state evaluator to determine a first health state coefficient; identifying the health state of the set of monitoring state semantic information sequences by using the health state evaluator to obtain a set of health state coefficient sequences; calculating the mean value of each health state coefficient sequence in the set of health state coefficient sequences to obtain a set of health state coefficient mean values, taking the set of health state coefficient mean values as the identification result, and determining a first set of resource allocation coefficients based on the identification result; calling a set of historical health state identification result sequences of the wireless sensing module in a historical monitoring window, identifying the health decay of the set of historical health state identification result sequences, and determining a second set of resource allocation coefficients; performing resource allocation optimization on the wireless sensing module based on the first set of resource allocation coefficients and the second set of resource allocation coefficients to obtain an updated wireless sensing module, and using the updated wireless sensing module to intelligently monitor the target power distribution network online.

2. The method as claimed in claim 1, wherein, deploying a wireless sensing module on a target power distribution network, wherein the wireless sensing module supports a low-power wide-area network protocol, including: Obtaining a key node set of a target power distribution network, respectively arranging wireless sensors for the key nodes to obtain a wireless sensor set; Respectively sampling the key node set by using the wireless sensor set to obtain a trial sampling data set; Performing data quality analysis on the trial sampling data set, identifying wireless sensors whose data quality analysis results do not meet the requirements, and obtaining an abnormal wireless sensor set; Respectively associating the abnormal wireless sensor set with a sampling enhancer, and collecting the abnormal wireless sensor set and a normal wireless sensor set in the wireless sensor set to obtain a wireless sensing module.

3. The method of claim 2, wherein the wireless sensing of the power distribution grid is performed by a plurality of sensors, and the plurality of sensors are configured to transmit the sensed data to the at least one processor via a wireless communication network. Respectively associating the abnormal wireless sensor set with a sampling enhancer, including: Obtaining a sampling enhancer, wherein the sampling enhancer includes a depth separable convolution branch and a low-resolution self-attention branch; The depth separable convolution branch includes a channel-wise convolution layer and a point-wise convolution layer, which is used to capture local details; The low-resolution self-attention branch performs self-attention calculation by down-sampling the input data to a low-resolution space, up-samples the calculation result, and performs residual connection on the up-sampled result and the input data, and outputs sampling enhancement data for global analysis.

4. The method of claim 1, wherein the wireless sensing of the power distribution grid is performed by a plurality of sensors. Extracting a monitoring data sequence set of the wireless sensing module in a preset monitoring window, including: Obtaining a preset monitoring index set, wherein the preset monitoring index set at least includes voltage amplitude, voltage fluctuation frequency, current amplitude, overload current, active power, reactive power and load data; Monitoring the target power distribution network according to the preset monitoring index set by the wireless sensing module in the preset monitoring window to obtain the monitoring data sequence set.

5. The method as claimed in claim 1, wherein, Retrieving a historical health state recognition result sequence set of the wireless sensing module in a historical monitoring window, performing health attenuation recognition on the historical health state recognition result sequence set, and determining a second resource allocation coefficient set, including: Respectively calculating a health attenuation coefficient set of the historical health state recognition result sequence set; Dividing each health attenuation coefficient in the health attenuation coefficient set by the sum of the health attenuation coefficient set to obtain a second resource allocation coefficient set.

6. The method of wireless sensed power distribution network intelligent online monitoring of claim 1, wherein, Including: Mapping and weighting the first resource allocation coefficient set and the second resource allocation coefficient set to determine a target resource allocation coefficient set; Based on the target resource allocation coefficient set, the resource allocation optimization of the wireless sensing module is performed to obtain an updated wireless sensing module.

Images