Power distribution network short-circuit parameter real-time monitoring method, system and equipment based on edge calculation and medium

Abstract

The invention discloses a power distribution network short-circuit parameter real-time monitoring method, system and equipment based on edge calculation and a medium, and belongs to the technical field of power system monitoring and protection, and the method comprises the steps: collecting data information in real time at each monitoring node of a power distribution network, generating a state vector reflecting a node operation state, and inputting the state vector to an edge calculation unit, processing through an internal first calculation model to obtain a power grid fault evaluation result of a corresponding node, and compressing and uploading the power grid fault evaluation result to a cloud; based on the uploaded data and the system operation state, a system performance evaluation result is generated in real time, and an adaptive optimization algorithm is used for dynamically adjusting internal model parameters of the edge calculation unit. According to the method, the composite algorithm model fusing feature extraction and parameter calculation is deployed at the edge node, the problems that a traditional centralized processing mode is large in data transmission delay and central node calculation pressure is concentrated are solved, and the technical effect of reducing bandwidth occupation of a backbone communication network is achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power system monitoring and protection, in particular to a power distribution network short-circuit parameter real-time monitoring method, system, device and medium based on edge computing. BACKGROUND

[0002] Real-time monitoring of power distribution network short-circuit parameters is crucial to ensure the safe and stable operation of the power grid. Traditional monitoring methods generally rely on centralized computing architecture, that is, all measurement data is remotely transmitted to the main station for unified processing. This method has inherent defects: first, the remote transmission of massive data causes significant communication bandwidth pressure and significant data transmission delay, making it difficult to meet the millisecond-level real-time requirements of fault monitoring; second, the computing pressure is highly concentrated on the central node, which easily forms a performance bottleneck and has poor system scalability.

[0003] To overcome the above problems, some existing distributed solutions attempt to decentralize some computing tasks. However, these solutions still have insufficient computing efficiency and accuracy when dealing with high-frequency sampling data and complex network topologies. More importantly, with the large-scale integration of distributed energy sources (such as photovoltaic and wind power), the short-circuit current level, direction, and characteristics of the power distribution network exhibit significant dynamic changes. Traditional methods and existing distributed solutions lack effective adaptive mechanisms and cannot perform online calibration and optimization of model parameters, resulting in decreased monitoring accuracy and insufficient adaptability in the new power distribution network environment. SUMMARY

[0004] In view of the above problems, the present application is proposed.

[0005] Therefore, the technical problem solved by the present application is how to overcome the defects of poor real-time performance and heavy communication burden caused by centralized processing in existing power distribution network short-circuit parameter monitoring methods, while addressing the insufficient adaptive capacity and decreased computing accuracy of existing distributed methods when facing high-frequency data, complex topologies, and distributed energy source integration.

[0006] To solve the above technical problems, the present application provides the following technical solutions: a power distribution network short-circuit parameter real-time monitoring method based on edge computing, comprising, At each monitoring node of the power distribution network, real-time data information is collected and a state vector reflecting the node's operating state is generated; The state vector is input into an edge computing unit, which is processed by an internal first computing model to obtain a power grid fault assessment result for the corresponding node; The power grid fault assessment result is compressed and the compressed data is uploaded to the cloud; Based on the uploaded data and system operating state, a system performance assessment result is generated in real time, Based on the system performance evaluation result, the internal model parameters of the edge computing unit are dynamically adjusted using an adaptive optimization algorithm.

[0007] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, the state vector reflecting the running state of the node is generated by real-time collection of data information at each monitoring node of the power distribution network. The electrical quantity data is measured in real time by the sensing device deployed at the node. Based on the electrical quantity data, the state vector capable of representing the comprehensive running state of the node is calculated and combined. The data collection frequency of the state vector is dynamically configured according to the system monitoring requirement.

[0008] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, the state vector is input to the edge computing unit and processed by the internal first calculation model. The state vector is distributedly calculated by the internal first calculation model in the edge computing unit. The fault analysis result of the system fault characteristic is synchronously output by the first calculation model. The first calculation model comprises a composite algorithm structure for data feature extraction and parameter calculation.

[0009] In the traditional method, a large amount of raw data needs to be transmitted back to the center for processing, and the link is long and the delay is large.

[0010] In the present application, the feature extraction and parameter calculation tasks originally performed sequentially in the cloud are converted into parallel processing at the data source by deploying the first calculation model on the edge side. The unnecessary data transmission time is eliminated, and the response time is fundamentally compressed. The depth of fault information processing is ensured, and the output result contains qualitative judgment and quantitative analysis at the same time.

[0011] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, the fault analysis result is compressed and the compressed data is uploaded to the cloud. The fault analysis result is calculated and processed by using a data compression algorithm. In the compression process, the reconstruction error of the data is constrained within a preset error threshold by an error control mechanism. The compressed data is uploaded to the cloud for global state analysis and optimization decision of the power distribution network.

[0012] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, wherein the system performance evaluation result is generated in real time based on the uploaded data and the system running state, obtain the time delay error, communication overhead and reliability index data of the system; Based on the obtained system data, the comprehensive performance index is calculated according to the predefined weight coefficient; The comprehensive performance index is output as a quantitative basis for reflecting the overall running state of the system.

[0013] The running condition of the power grid is dynamically changing, and any preset and fixed model parameter cannot always remain optimal.

[0014] The application continuously monitors the performance indicators such as calculation delay and accuracy deviation of itself, and automatically starts the optimization algorithm to fine-tune the key parameters such as weight and correction coefficient in the model once the performance degradation is found by evaluation, so as to ensure that it always matches the actual running characteristics of the current power grid.

[0015] In addition, the current optimization mechanism can dynamically allocate computing power according to the task priority by adjusting the scheduling strategy of the computing resource, and can ensure the absolute priority execution of the core function when the system load is too high or abnormal.

[0016] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, wherein the system performance evaluation result is generated in real time based on the uploaded data and the system running state, According to the system performance evaluation result, the gradient of the loss function weight is calculated; Based on the gradient and combined with the adaptive learning rate and momentum factor, the update amount of the model parameter is calculated; The internal model weight parameters of the edge computing unit are iteratively updated by using the update amount.

[0017] As a preferred scheme of the power distribution network short-circuit parameter real-time monitoring method based on edge computing, wherein the running of the first calculation model includes adaptive scheduling of computing resources; Monitor the computing tasks concurrently executed in the first calculation model, and obtain the computing demand of each task; Based on the computing demand of each task and the total available computing resources of the edge computing unit, the computing resources are allocated in proportion; The resource allocation proportion is dynamically adjusted according to the task priority.

[0018] The application provides a power distribution network short-circuit parameter real-time monitoring system based on edge computing.

[0019] To solve the above technical problems, the application provides the following technical scheme: a power distribution network short-circuit parameter real-time monitoring system based on edge computing, comprising a data collection module, a power grid fault evaluation module, a compression module, a performance evaluation module and an updating module; The data collection module is used to collect data information in real time at each monitoring node of the power distribution network and generate a state vector reflecting the running state of the node. The power grid fault evaluation module is used to input the state vector into an edge computing unit, process it through an internal first calculation model and obtain the power grid fault evaluation result of the corresponding node. The compression module is used to compress the power grid fault evaluation result and upload the compressed data to the cloud. The performance evaluation module is used to generate a system performance evaluation result in real time based on the uploaded data and the system running state. The updating module is used to dynamically adjust the internal model parameters of the edge computing unit based on the system performance evaluation result using an adaptive optimization algorithm.

[0020] The application provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and wherein the processor implements the steps of the power distribution network short-circuit parameter real-time monitoring method based on edge computing when executing the computer program.

[0021] The application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program implements the steps of the power distribution network short-circuit parameter real-time monitoring method based on edge computing when executed by a processor.

[0022] The application has the following beneficial effects: the application deploys a composite algorithm model integrating feature extraction and parameter calculation at an edge node, solves the problems of large data transmission delay and concentrated calculation pressure of the central node in the traditional centralized processing mode, and reduces the bandwidth occupation of the backbone communication network. BRIEF DESCRIPTION OF DRAWINGS

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the application, the following will briefly introduce the drawings needed in the embodiment description. Obviously, the drawings in the following description are only some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0024] Figure 1A general flowchart of a power distribution network short-circuit parameter real-time monitoring method based on edge computing is provided for an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to make the above-mentioned purposes, features and advantages of the present application more obvious and easy to understand, the specific embodiments of the present application will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative labor should fall within the scope of protection of the present application.

[0026] Embodiment 1, refer to Figure 1 For an embodiment of the present application, the embodiment provides a power distribution network short-circuit parameter real-time monitoring method based on edge computing, comprising: The existing power distribution network short-circuit parameter monitoring technology mainly relies on a centralized computing architecture, requiring all monitoring nodes to upload the collected raw electrical quantity data to the central master station for unified processing. This architecture has inherent defects: the remote transmission of massive data causes huge communication bandwidth pressure and significant data transmission delay, making it difficult to meet the millisecond-level real-time requirements of short-circuit fault monitoring; the computing pressure is highly concentrated on the central node, which easily forms a performance bottleneck, resulting in poor system scalability.

[0027] Although some distributed solutions have been proposed to attempt to decentralize some computing tasks, these solutions still have deficiencies when dealing with high-frequency sampling data and complex network topologies. The computing model is often single-functioned, only capable of performing limited calculations or judgments, and unable to simultaneously complete deep extraction of fault features and accurate calculation of short-circuit parameters, resulting in limited comprehensiveness and accuracy of fault diagnosis. More importantly, with the large-scale access of distributed energy, the short-circuit current level, direction and characteristics of the power distribution network present significant dynamic changes, while the model parameters of existing methods are mostly fixed presets, lacking effective online calibration and adaptive mechanisms, making it difficult to ensure monitoring accuracy in the new power distribution network environment.

[0028] In addition, existing solutions generally lack intelligent management mechanisms for edge-side computing resources. The allocation of computing resources is usually static and cannot be dynamically adjusted according to task priority and system state, making it difficult to ensure that critical tasks can obtain sufficient computing resources in the event of a sudden fault, posing a risk to system reliability.

[0029] Therefore, the present application proposes a power distribution network short-circuit parameter real-time monitoring method based on edge computing.

[0030] S1 In each monitoring node of the power distribution network, real-time data information is collected, and a state vector reflecting the running state of the node is generated; S2 inputs the state vector to the edge computing unit, processes through an internal first calculation model, and obtains a power grid fault evaluation result of the corresponding node; S3 performs compression processing on the power grid fault evaluation result, and uploads the compressed data to the cloud; S4 generates a system performance evaluation result in real time based on the uploaded data and the system running state, S5 dynamically adjusts the internal model parameters of the edge computing unit based on the system performance evaluation result using an adaptive optimization algorithm.

[0031] Embodiment 2 is an embodiment of the present application, which provides a power distribution network short-circuit parameter real-time monitoring method based on edge computing based on the above embodiment, comprising:

[0032] Step S1, real-time data information is collected at each monitoring node of the power distribution network, and a state vector reflecting the node running state is generated, including the following steps: S11, real-time measurement of electrical quantity data is performed through the sensing device deployed at the node.

[0033] Through the high-precision sensor deployed at the node, the original electrical quantity data such as voltage and current are synchronously and real-time measured.

[0034] S12, calculation is performed based on the electrical quantity data and combined into a state vector capable of representing the comprehensive running state of the node.

[0035] S13, the data collection frequency of the state vector is dynamically configured according to system monitoring requirements.

[0036] The components of the state vector are real-time collected by high-precision sensors, and the sampling frequency can be dynamically adjusted according to monitoring requirements, and a typical value is 100 times per second.

[0037] In the present application, S12 calculates and combines the state vector capable of representing the comprehensive running state of the node based on the electrical quantity data is based on the real-time collected voltage and current instantaneous values, and the voltage effective value and current effective value are obtained through effective value calculation; Based on the voltage effective value, the current effective value and the voltage-current phase difference, the active power and the reactive power are obtained through the power calculation formula; The voltage-current phase difference is calculated by analyzing the zero-crossing point of the voltage and current waveforms or using a phase detection method; The system frequency is calculated by analyzing the periodic change of the voltage waveform; The calculated node voltage effective value, current effective value, short-circuit impedance complex value, active power, reactive power, voltage-current phase angle difference and system frequency components are combined in a predetermined order to form a state vector, and the expression is: wherein, is the node voltage effective value, is the current effective value, is the short-circuit impedance complex value, is the active power, is the reactive power, is the voltage-current phase angle difference, is the system frequency.

[0038] The state vector reflects the electrical characteristics and operating state of the edge node.

[0039] In an optional embodiment, the S12 calculates and combines the electrical quantity data into a state vector capable of representing the comprehensive operating state of the node is In another optional embodiment, the S12 calculates and combines the electrical quantity data into a state vector capable of representing the comprehensive operating state of the node is to perform fast Fourier transform on the collected voltage and current instantaneous values by a digital signal processor, directly and simultaneously extract the fundamental voltage effective value, the fundamental current effective value, the voltage-current fundamental phase difference and each harmonic component; calculate the active power, the reactive power and the system frequency based on the extracted fundamental parameters; and combine the fundamental voltage effective value, the fundamental current effective value, the fundamental active power, the fundamental reactive power, the voltage-current fundamental phase difference, the system frequency and the harmonic distortion rate of each order to form a state vector.

[0040] The application realizes synchronous and accurate collection of electrical quantity data, comprehensive and comprehensive representation of the operating state of the node by integrating multiple key electrical parameters into a unified state vector, and adaptive balance between monitoring accuracy and system resource consumption by dynamically adjusting the sampling frequency.

[0041] Step S2, input the state vector into the edge computing unit, process through the internal first calculation model to obtain the power grid fault assessment result of the corresponding node, including the following steps: S21, the edge computing unit adopts the internal first calculation model to perform distributed calculation on the state vector; A plurality of edge computing units are deployed in the distribution network, and each unit is responsible for monitoring one or more nodes. Each edge unit independently runs the first calculation model to perform parallel processing on the locally collected state vector without waiting for data from other nodes or centralized scheduling.

[0042] When a certain edge unit receives the local state vector, it immediately starts the first calculation model for processing. The model internally includes two core components, a feature extraction unit and a parameter calculation unit.

[0043] The feature extraction unit is responsible for identifying and extracting key fault feature information from the state vector, including amplitude change characteristics, phase relationship, frequency offset, and other multi-dimensional information of electrical quantities.

[0044] The parameter calculation unit calculates the key parameters of the short-circuit impedance of the node based on the extracted features, combined with the preset calculation rules and dynamic correction mechanism.

[0045] The calculation process of each edge unit is independent and synchronized, forming a truly distributed computing architecture. Each unit adjusts the weight parameters and calculation accuracy within the model according to the characteristics of the local state vector to adapt to different operating conditions. After calculation, each edge unit outputs the local fault analysis result, including the short-circuit parameter estimate and the fault feature vector.

[0046] In order to realize the cooperation between multiple edge units, the system adopts a distributed fusion mechanism.

[0047] When global fault location or regional analysis is required, adjacent edge units exchange their calculation result summary information, rather than raw data.

[0048] The fusion process is achieved through a weighted average method, and the output results of each edge unit are given different fusion weights according to their operating reliability and measurement accuracy. The system evaluates the data quality and calculation reliability of each edge unit in real time, and normalizes the evaluation results as fusion weights. Units with high reliability and measurement accuracy have greater weights, and their calculation results dominate in the fusion process.

[0049] S22, synchronously outputting the fault analysis result of the system fault characteristics through the first calculation model.

[0050] Based on the state vector input, the equivalent short-circuit impedance complex value is calculated in real time through the short-circuit parameter calculation module in the first calculation model.

[0051] Through the fault feature extraction module in the first calculation model, a multi-dimensional fault feature vector is extracted from the state vector.

[0052] The equivalent short-circuit impedance complex value and the multi-dimensional fault feature vector are jointly output as the fault analysis result.

[0053] The calculation formula of the equivalent short-circuit impedance complex value is: wherein, is the equivalent short-circuit impedance complex value, is the effective value of the fault point voltage, is the total fault current flowing through the fault point, is the initial operating current before the fault occurs, is the reactance angle, reflecting the reactance characteristic of short-circuit impedance, is the comprehensive correction coefficient, used to compensate the influence of distributed power supply, load fluctuation and other factors.

[0054] By introducing initial current compensation and correction coefficient, the calculation accuracy is significantly improved. The correction coefficient is updated in real time using an adaptive algorithm, ensuring the accuracy of the calculation results.

[0055] To realize fast identification and accurate positioning of short-circuit faults, the present application designs a multi-dimensional fault feature extraction method. By analyzing the electrical quantity change characteristics during the fault process, a fault feature vector is constructed: where, is the fault current per unit value, is the fault current instantaneous value, is the system rated current; is the fault voltage per unit value, is the fault voltage instantaneous value, is the system rated voltage; is the phase angle change before and after the fault, reflecting the transient characteristics of the system; is the frequency change rate, used to represent the dynamic response characteristics of the system. The current feature vector comprehensively reflects the key features of the fault process, providing a basis for fault diagnosis. The feature extraction uses a sliding window technique, and the window length can be dynamically adjusted according to the fault type.

[0056] S23, the first calculation model contains a composite algorithm structure for data feature extraction and parameter calculation.

[0057] In the first calculation model in the present embodiment S2, an improved embedded processor is used in the edge computing unit, which has data acquisition, processing and communication functions. Its basic processing model can be represented as: where, is the processing output result of the edge unit, including the short-circuit parameter estimation value and the state evaluation result, is the weight matrix of the processing model, reflecting the importance of different input quantities, is the input state vector, is the model bias term, used to correct system errors, is the nonlinear activation function.

[0058] An improved ReLU function is used to improve the model expression ability, is the comprehensive error in the measurement and processing process, including sensor error, quantization error and model error.

[0059] The first computing model of an optional embodiment S2 is a recurrent neural network architecture based on a long short-term memory network (LSTM), which models the time sequence characteristics of the state vector through its gating mechanism; the state vector is input into the network in a time sequence, and the dynamic change law before and after the fault occurs is automatically learned; the estimation value of the short-circuit impedance and the classification probability of the fault type are synchronously output by the fully connected layer in the network.

[0060] In another optional embodiment S2, the first computing model is a multi-task learning framework, which constructs a composite structure of a shared bottom feature extraction network and multiple special task output branches; the input state vector is processed by the shared network for general feature extraction, and then by the short-circuit parameter calculation branch and the fault feature recognition branch; an attention mechanism is established between the two branches, so that the fault recognition result can feedback and correct the short-circuit parameter calculation process, improving the overall accuracy of the model.

[0061] In the present embodiment S21, distributed computing is to complete most of the computing tasks at the edge nodes. Local feature extraction is the key link, and the mathematical expression is: In the formula, is the extracted feature vector, reflecting the local electrical characteristics; is the feature weight coefficient, the initial value of αj is determined by offline training, the value range is [0, 1], and the normalization constraint Σαj=1 is satisfied, and the recursive least squares method (RLS) is used in the online optimization process; is the original sampling data, including voltage, current and other basic measurement quantities; is the feature dimension, is a non-linear correction factor, used to enhance the expression ability of the model to non-linear characteristics; is a non-linear activation function, using an improved Sigmoid function to ensure the stability of feature extraction.

[0062] The feature extraction method fully considers the correlation between electrical quantities and can effectively extract short-circuit feature information. The feature extraction result is uploaded to the upper node after data compression, reducing the communication burden. The distributed processing architecture not only improves the computing efficiency, but also enhances the reliability and scalability of the system. By reasonably configuring the computing resources, the system can flexibly cope with different scales of monitoring demand.

[0063] In an optional embodiment S21, the distributed computing is to use a federated learning architecture, each edge computing unit trains the first computing model based on local data, and only uploads the model parameter update to the cloud; the cloud aggregates the model updates from multiple edge nodes, generates an improved global model, and then distributes it to each node.

[0064] Through the way of cyclic iteration, the synergistic evolution of all edge node computing models is realized under the premise of ensuring data privacy.

[0065] In another optional embodiment S21, the distributed computing is in a hierarchical collaborative computing mode, the monitoring area is divided into multiple computing domains, a master edge node and multiple slave nodes are arranged in each domain, the slave nodes send the feature vectors to the master node for data fusion and comprehensive analysis in the domain after completing local feature extraction, and the master node uploads the refined results to the cloud after completing fault diagnosis in the domain.

[0066] The distributed computing architecture is adopted, the reasonable distribution of computing load is realized, the core computing task is transferred from the center node to the edge node, the cloud computing pressure is effectively reduced, the system response time is controlled within milliseconds, and the monitoring real-time performance is improved.

[0067] Through the design of the composite algorithm structure, parallel processing of feature extraction and parameter calculation is realized, the model can output the short-circuit impedance estimation value and the multi-dimensional fault feature vector synchronously, and the comprehensiveness and accuracy of fault diagnosis are improved. Through the cooperation of the improved short-circuit impedance calculation model and the adaptive optimization mechanism, the system can automatically adapt to the change of power grid operation state caused by distributed energy access and the like, continuously maintain high-precision monitoring capability, and significantly enhance the environmental adaptability of the system.

[0068] Step S3, the fault evaluation result of the power grid is compressed, and the compressed data is uploaded to the cloud, comprising the following steps: S31, the fault analysis result is calculated and processed by using a data compression algorithm.

[0069] The adaptive quantization function dynamically adjusts the quantization level according to the data characteristics and the system communication state, and differentiates the key data such as short-circuit impedance value and fault feature vector; then the compression transformation matrix optimized by principal component analysis is used to reduce the dimension of the data, and the most representative fault feature information is retained.

[0070] Meanwhile, the system supports online switching of lossless and lossy compression modes, adopts the lossy mode to ensure real-time transmission of key data when the communication bandwidth is tight, and adopts the lossless mode to ensure data integrity when the bandwidth is sufficient.

[0071] S32, in the compression process, the reconstruction error of the data is constrained within the preset error threshold by an error control mechanism.

[0072] Error monitoring and control mechanisms are provided at each stage of the compression process: a maximum quantization error is set at the quantization stage, information loss is controlled by retaining the contribution rate of principal components at the dimension reduction stage, and key fault feature data is protected through an error compensation term.

[0073] The system monitors the reconstruction error in real time, and when the error approaches the preset threshold of 2%, the compression parameters are automatically adjusted to prioritize the accuracy of key parameters such as short-circuit impedance and fault current, ensuring that the compressed data still meets the accuracy requirements for subsequent analysis.

[0074] S33, upload the compressed data to the cloud for global state analysis and optimization decision-making of the power distribution network.

[0075] After uploading the compressed data, the cloud performs data decompression and quality verification, and then performs fusion analysis based on the compressed data uploaded by multiple nodes, achieving global fault diagnosis and fault section positioning of the power distribution network.

[0076] At the same time, a complete fault case library is established to support historical data tracing and fault mode analysis. More importantly, based on the comparative analysis of long-term accumulated compressed data and corresponding decompressed data, the compression algorithm parameters and error control strategies at the edge are continuously optimized, forming a complete closed loop from data compression, transmission to decompression application.

[0077] In the compression process in the S3 embodiment of the present application, considering the limitation of communication bandwidth in the edge computing environment, a lightweight data compression algorithm is developed to significantly reduce the transmission burden while ensuring data quality.

[0078] The mathematical expression of the compression algorithm is: Q(X) = round(X / Δ)Δ In the formula, is the compressed data, is the original data, is an adaptive quantization function that dynamically adjusts the quantization level according to data characteristics, round is a rounding function, Δ is a quantization step that is adaptively adjusted according to the dynamic range of data, and X max , X min are the maximum and minimum measurement values in the current data block, respectively; b is the number of quantization bits, which is dynamically selected between 6-12 bits according to the communication bandwidth state. is a compression transformation matrix, which is obtained by principal component analysis optimization to realize data dimension reduction, To compress the error compensation term, for controlling the reconstruction error. The compression algorithm has the characteristics of low computational complexity and high compression ratio, and the compression ratio can reach 10:1, while ensuring that the reconstruction error is within 2%. The algorithm supports both lossless and lossy compression modes, which can be flexibly switched according to application requirements. Through data compression, the communication burden is significantly reduced, and the system real-time performance is improved.

[0079] In an optional embodiment, the compression mode of S3 is to adopt a hierarchical compression strategy based on feature importance, and different compression precisions are adopted for different feature components in the fault analysis result; a lossless compression mode is adopted for key parameters such as short-circuit impedance, and a lossy compression mode is adopted for auxiliary monitoring data; through a dynamic bit allocation algorithm, the overall data amount is effectively reduced while the accuracy of key data is ensured.

[0080] In another optional embodiment, the compression mode of S3 is to adopt a method combining time series prediction and residual coding, and the time sequence characteristics of fault data are utilized to improve the compression efficiency; a short-term prediction model is established to generate a predicted value, and only the residual between the actual value and the predicted value is encoded and transmitted; when the system is in a steady state, the compression ratio is increased, and when a transient process is detected, the sampling rate is automatically increased and a more conservative compression strategy is adopted to ensure data integrity.

[0081] The present application efficiently processes the fault analysis result through a lightweight data compression algorithm, and compresses the monitoring data originally requiring a large amount of bandwidth to one-tenth of the original size, directly solving the core problem of excessive communication burden of the traditional monitoring method.

[0082] In the compression process, an error compensation mechanism is introduced to strictly control the reconstruction error within 2%, ensuring that the compressed data still maintains sufficient accuracy when restored in the cloud.

[0083] By uploading the compressed data to the cloud for global analysis, the powerful computing capability of the cloud is fully utilized to realize accurate perception and optimization decision of the global state of the power distribution network, and the dependence of the system on the communication network is reduced by reducing the amount of transmitted data, thereby enhancing the adaptability of the system in a poor communication environment.

[0084] Step S4, based on the uploaded data and the system running state, real-time generation of system performance evaluation results includes the following steps: S41, obtain the time delay error, communication overhead and reliability index data of the system.

[0085] By monitoring the processing delay of the edge computing unit, the data transmission amount of the communication module and the system running state, three key performance indicators are collected in real time: time delay error E delay (including calculation delay and communication delay), communication overhead C commReflecting data transmission volume) and system reliability index R s .

[0086] System reliability is an important indicator to measure the performance of the monitoring system. The present application establishes a multi-level reliability evaluation model: Among them, is the overall reliability of the system, is the reliability of the i-th edge computing unit, is the failure rate of the i-th edge computing unit, is the running time, is the total number of edge computing units in the system.

[0087] The model considers various failure factors such as hardware failure, software error and communication interruption, and through real-time monitoring of reliability indicators, potential risks are discovered in time. The system adopts redundancy design and fault switching mechanism to ensure that the basic functions can still be maintained in the case of single point failure. The reliability evaluation results are used to guide the development of system maintenance and upgrade strategies.

[0088] S42, based on the obtained system data, according to the pre-defined weight coefficient, the comprehensive performance index is calculated.

[0089] After obtaining the system operation data, the comprehensive performance evaluation model provided by the inventor is used for calculation. The mathematical expression of the model is: Among them, is the comprehensive performance index, is the delay error, including calculation delay and communication delay; is the communication overhead, reflecting the data transmission volume; is the system reliability index; , , is the weight coefficient, the initial value is set: ω1=0.5 (delay error weight), ω2=0.3 (communication overhead weight), ω3=0.2 (reliability weight) satisfies the constraint condition: ω1+ω2+ω3=1, and ω i ∈[0,1], through adaptive algorithm dynamic adjustment, meet the needs of different application scenarios.

[0090] For real-time fault protection scenarios with high requirements: ω1=0.6, ω2=0.25, ω3=0.15.

[0091] For remote power distribution networks with limited communication: ω1=0.3, ω2=0.5, ω3=0.2.

[0092] For the power supply scene of important loads with high reliability requirements: ω1=0.3, ω2=0.2, ω3=0.5.

[0093] The evaluation model can reflect the system running state in real time, and provide basis for optimization decision.

[0094] S43, output the comprehensive performance index as a quantitative basis reflecting the overall running state of the system.

[0095] The calculated comprehensive performance index is a quantitative representation of the system running state, and reflects the overall performance level of the monitoring system in real time. After the index is output, the decision basis for subsequent adaptive optimization is provided, so that the system can dynamically adjust the running parameters according to the current performance status, and continuously maintain the best working state.

[0096] Step S5, dynamically adjusting the internal model parameters of the edge computing unit based on the system performance evaluation result includes the following steps: S51, calculating the gradient of the loss function weight according to the system performance evaluation result.

[0097] By analyzing the difference between the system performance evaluation result and the expected performance target, a loss function is constructed and the gradient of the function with respect to the model weight is calculated.

[0098] The parameter optimization algorithm based on the improved gradient descent method. This algorithm can dynamically adjust the model parameters according to the running state, and the optimization expression is: In the formula, is the weight parameter at the next moment, is the weight parameter at the current moment, is the adaptive learning rate, which is dynamically adjusted according to the change of the loss function, and the initial value η0=0.01, is the momentum factor, which is used to accelerate convergence and avoid local optimum, and λ is set to a fixed value of 0.9, which has the physical meaning of retaining 90% of the historical gradient information, accelerating convergence and suppressing oscillation, is the gradient of the loss function with respect to the current weight, which reflects the direction and amplitude of parameter update, is the parameter change amount.

[0099] The adaptive adjustment strategy adopts the AdaGrad method: Where g(τ) is the gradient value of the weight parameter at the τth moment, and ε is a numerical stability term, which is taken as 1×10 -8 , the actual value range of η(t) is [0.0001, 0.1].

[0100] When the gradient is large (system performance deviation is large), η is automatically reduced to prevent oscillation; when the gradient is small (close to the optimal point), η remains a small value to achieve fine adjustment.

[0101] Loss function The calculation accuracy, response time, resource consumption, and robustness loss indicators are comprehensively considered.

[0102] Loss function The specific calculation method is: The specific calculation formula of each component is: Calculation accuracy loss : Where, is the short-circuit impedance predicted by the model, is the actual measured value, and N is the sample number.

[0103] Response time loss : Where, is the actual calculation time, is the time threshold (typical value 50ms).

[0104] Resource consumption loss : Where, is the actual CPU usage, is the total CPU capacity, is the actual memory usage, is the total memory capacity, β1=0.6, β2=0.4, indicating the relative importance of CPU and memory.

[0105] Robustness loss : Where, M is the number of test conditions, is the error under the jth disturbance, is the error threshold (set to 2%).

[0106] The weight coefficients of each loss component: α1=0.4 (accuracy weight), α2=0.3 (time weight), α3=0.2 (resource weight), α4=0.1 (robustness weight), satisfying Σα i =1.

[0107] S52, calculate the update amount of the model parameter based on the gradient and combined with the adaptive learning rate and the momentum factor.

[0108] S52, update the internal model weight parameter of the edge computing unit iteratively using the update amount.

[0109] Apply the calculated parameter update amount to the internal model of the edge computing unit to complete the iterative update of the weight parameter from to .

[0110] Ensure that the model can continuously evolve according to the actual running state of the system, so that the short-circuit parameter calculation model and the feature extraction model always maintain the best performance.

[0111] Embodiment 3 is an embodiment of the present application, which provides an edge computing-based real-time monitoring system for short-circuit parameters of a power distribution network, comprising a data collection module, a power grid fault assessment module, a compression module, a performance assessment module, and an update module The data collection module is at each monitoring node of the power distribution network, which collects data information in real time and generates a state vector reflecting the running state of the node; The power grid fault assessment module inputs the state vector into the edge computing unit, processes it through the internal first calculation model, and obtains the power grid fault assessment result of the corresponding node; The compression module compresses the power grid fault assessment result and uploads the compressed data to the cloud; The performance assessment module generates a system performance assessment result in real time based on the uploaded data and the system running state, The update module dynamically adjusts the internal model parameters of the edge computing unit based on the system performance assessment result using an adaptive optimization algorithm.

[0112] The embodiment also provides an electronic device suitable for the edge computing-based real-time monitoring method for short-circuit parameters of a power distribution network, comprising a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute the computer executable instructions to realize the edge computing-based real-time monitoring method for short-circuit parameters of a power distribution network proposed in the above embodiment.

[0113] The embodiment also provides a storage medium having a computer program stored thereon, which is executed by a processor to realize the edge computing-based real-time monitoring method for short-circuit parameters of a power distribution network proposed in the above embodiment.

[0114] The storage medium proposed in the embodiment and the edge computing-based real-time monitoring method for short-circuit parameters of a power distribution network proposed in the above embodiment belong to the same inventive concept, and the technical details not described in detail in the embodiment can be referred to the above embodiment, and the embodiment has the same beneficial effects as the above embodiment.

[0115] From the above description of the embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software and necessary universal hardware, and of course can also be implemented by hardware, but in many cases the former is a better implementation. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a floppy disk, a read-only memory (ROM), a random access memory (RAM), a FLASH memory, a hard disk, or an optical disc, and includes a number of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of various embodiments of the present application.

[0116] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present application but not limit the present application, and although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or replaced equivalently without departing from the spirit and scope of the present application, and all of them should be covered in the scope of the claims of the present application.

Claims

1. A method for real-time monitoring of short-circuit parameters in distribution networks based on edge computing, characterized in that: include, At each monitoring node of the power distribution network, data information is collected in real time, and a state vector reflecting the operating status of the node is generated. The state vector is input to the edge computing unit and processed by the internal first computing model to obtain the power grid fault assessment result of the corresponding node. The power grid fault assessment results are compressed, and the compressed data is uploaded to the cloud. Based on the uploaded data and system operating status, system performance evaluation results are generated in real time. Based on the system performance evaluation results, the internal model parameters of the edge computing unit are dynamically adjusted using an adaptive optimization algorithm.

2. The real-time monitoring method for short-circuit parameters in distribution networks based on edge computing as described in claim 1, characterized in that: The process of collecting data information in real time at each monitoring node of the distribution network and generating a state vector reflecting the operating status of the node includes: Electrical quantity data is measured in real time through sensing devices deployed at nodes; Based on electrical quantity data, a state vector that can characterize the overall operating status of the node is calculated and combined. The data acquisition frequency of the state vector is dynamically configured according to the system monitoring requirements.

3. The real-time monitoring method for short-circuit parameters of distribution networks based on edge computing as described in claim 2, characterized in that: The step of inputting the state vector into the edge computing unit and processing it through the internal first computing model includes: The edge computing unit uses an internal first computing model to perform distributed computation on the state vector; The fault analysis results of the system fault characteristics are output synchronously through the first calculation model; The first computational model contains a composite algorithm structure for data feature extraction and parameter calculation.

4. The real-time monitoring method for short-circuit parameters of distribution networks based on edge computing as described in claim 3, characterized in that: The process of compressing the power grid fault assessment results and uploading the compressed data to the cloud includes: Data compression algorithms are used to process the fault analysis results. During the compression process, the reconstruction error of the data is constrained within a preset error threshold through an error control mechanism; The compressed data is uploaded to the cloud for the power distribution network to perform global status analysis and optimization decisions.

5. The method for real-time monitoring of distribution network short-circuit parameters based on edge computing as described in claim 4, characterized in that: The real-time generation of system performance evaluation results based on uploaded data and system operating status includes: Acquire data on system latency error, communication overhead, and reliability metrics; Based on the acquired system data, the comprehensive performance index is calculated according to the predefined weighting coefficients. The comprehensive performance indicators are output as a quantitative basis to reflect the overall operating status of the system.

6. The method for real-time monitoring of distribution network short-circuit parameters based on edge computing as described in claim 5, characterized in that: The method of dynamically adjusting the internal model parameters of the edge computing unit using an adaptive optimization algorithm based on system performance evaluation results includes... Calculate the gradient of the loss function weights based on the system performance evaluation results; The update amount of model parameters is calculated based on gradient and combined with adaptive learning rate and momentum factor; The internal model weight parameters of the edge computing unit are iteratively updated using the update amount.

7. The method for real-time monitoring of distribution network short-circuit parameters based on edge computing as described in claim 6, characterized in that: The operation of the first computing model includes adaptive scheduling of computing resources; Monitor concurrent computing tasks within the first computing model and obtain the computing requirements of each task; Based on the computational requirements of each task and the total available computational resources of the edge computing units, computational resources are allocated proportionally. The resource allocation ratio is dynamically adjusted based on task priority.

8. A real-time monitoring system for distribution network short-circuit parameters based on edge computing, employing the real-time monitoring method for distribution network short-circuit parameters based on edge computing as described in any one of claims 1 to 7, characterized in that, include: The system includes a data collection module, a power grid fault assessment module, a compression module, a performance assessment module, and an update module. The data collection module collects data information in real time at each monitoring node of the power distribution network and generates a state vector reflecting the operating status of the node. The power grid fault assessment module inputs the state vector to the edge computing unit, processes it through the internal first calculation model, and obtains the power grid fault assessment result for the corresponding node. The compression module compresses the power grid fault assessment results and uploads the compressed data to the cloud. The performance evaluation module generates system performance evaluation results in real time based on uploaded data and system operating status. The update module dynamically adjusts the internal model parameters of the edge computing unit based on the system performance evaluation results using an adaptive optimization algorithm.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the real-time monitoring method for short-circuit parameters of distribution networks based on edge computing as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the real-time monitoring method for short-circuit parameters of distribution networks based on edge computing as described in any one of claims 1 to 7.

Images