Fire-fighting electrical early warning method and system based on wireless networking

Abstract

The invention discloses a fire-fighting electrical early warning method and system based on wireless networking, particularly relates to the technical field of electrical fire monitoring, and is used for solving the problem of multi-node data space-time misalignment caused by transmission delay and data packet disorder in an existing wireless early warning system. Electrical parameter data with a local timestamp is acquired through a wireless sensor node, and after a transient characteristic waveform is extracted by a cloud platform, time synchronization verification is carried out by adopting a time service reference node and a dynamic fault-tolerant window mechanism; respectively generating a certainty index and a path conformity confidence index for the data which do not pass the verification through a recurrence plot analysis and transfer entropy algorithm, and finally carrying out fault analysis based on dual-index weighted correction; while the convenience of wireless networking is maintained, the accuracy and reliability of fault early warning are effectively improved through a multi-level data verification and correction mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electrical fire monitoring, and in particular to a fire-fighting electrical early warning method and system based on wireless networking. BACKGROUND

[0002] In the field of fire safety, electrical fire monitoring is an important preventive measure. In the prior art, electrical fire early warning systems using wireless networking have been applied. Such systems are usually composed of wireless sensors distributed in the power distribution system, wireless transmission networks, and cloud early warning platforms. The wireless sensors are responsible for collecting electrical parameters in the line and uploading data to the cloud platform through the wireless communication network; the cloud platform processes the received data using data analysis algorithms to monitor and warn of abnormal conditions in the electrical line. This wireless networking-based approach avoids the complexity of wiring in traditional wired solutions and is particularly suitable for old renovation and distributed building scenarios.

[0003] However, the existing fire-fighting electrical early warning method based on wireless networking still has significant defects. The inherent transmission mechanism of the wireless communication network introduces uncertain delays and packet reordering, resulting in a deviation between the timestamp sequence of data from sensor nodes at different spatial locations after transmission to the cloud and the actual physical fault waveform sequence. Precise early warning of electrical faults, especially the location of the fault source and the determination of the nature of the fault, highly depends on the collaborative analysis of transient characteristic waveforms collected by multiple nodes with strict spatio-temporal correlation. In the prior art, the cloud early warning model when processing these spatio-temporal data that are inaccurate due to transmission asynchrony, the physical basis of its analysis criteria is destroyed, leading to misjudgment or omission of the early warning conclusion, seriously weakening the reliability of the system. SUMMARY

[0004] The present application provides a fire-fighting electrical early warning method and system based on wireless networking to address the technical problems in the prior art.

[0005] The technical solution of the present application to solve the above technical problems is as follows:

[0006] The fire-fighting electrical early warning method based on wireless networking comprises:

[0007] S1, collecting electrical parameter data of the electrical line by multiple wireless sensor nodes and attaching local timestamps;

[0008] S2, the cloud early warning platform receives the electrical parameter data to extract transient characteristic waveform data of multiple nodes within the same fault period;

[0009] S3, selecting one wireless sensor node upstream on the selected topology as a time reference node, calculating a dynamic fault tolerance window for each node downstream based on the time stamp of the time reference node and the theoretical propagation delay, and determining whether it falls within the fault tolerance time window;

[0010] S4, constructing a recurrence plot of the transient characteristic waveform data of the nodes that do not fall within the fault tolerance time window and the upstream and downstream nodes, generating a certainty degree index through recursive quantization analysis;

[0011] S5, based on the time stamp sequence of the transient characteristic waveform data of the nodes that do not fall within the fault tolerance time window and the upstream and downstream nodes, using the transfer entropy algorithm to calculate the information flow direction and strength between nodes, and through comparison of the actual information flow direction and the theoretical causal flow direction based on the power grid topology, generating a path consistency confidence index;

[0012] S6, according to the certainty degree index and the path consistency confidence index, weighting and correcting the transient characteristic waveform data of the nodes that do not fall within the fault tolerance time window, and then performing electrical fault early warning analysis.

[0013] Further, the electrical parameter data of the electrical line is collected by the plurality of wireless sensor nodes and is additionally attached with local time stamps, including:

[0014] The wireless sensor nodes collect current and voltage data in the electrical line as electrical parameter data at a set sampling frequency;

[0015] After each data collection, each wireless sensor node obtains the current time from the built-in clock chip and attaches the corresponding local time stamp to the collected electrical parameter data.

[0016] Further, the cloud early warning platform receives the electrical parameter data to extract the transient characteristic waveform data of multiple nodes within the same fault period, including:

[0017] The cloud early warning platform receives the electrical parameter data uploaded by each wireless sensor node with local time stamps;

[0018] Detecting the fault start time based on the change of the current effective value in the electrical parameter data and determining the fault period;

[0019] The current waveform data of each node within the fault period is extracted from the electrical parameter data with local time stamps as the transient characteristic waveform data.

[0020] Further, one wireless sensor node upstream on the selected topology is selected as a time reference node, and the dynamic fault tolerance window for each node downstream is calculated based on the time stamp of the time reference node and the theoretical propagation delay, and it is determined whether it falls within the fault tolerance time window, including:

[0021] determining a wireless sensor node located at the most upstream of the electrical line as a time reference node according to the power grid topology;

[0022] calculating a dynamic fault-tolerant window of each downstream node based on the time stamp of the time reference node, adding a theoretical propagation delay of the fault waveform from the time reference node to each downstream node, and superimposing a preset transmission delay tolerance value;

[0023] judging whether the time stamp of the transient characteristic waveform data of each downstream node falls within the dynamic fault-tolerant window of the corresponding node.

[0024] Further, when there is a node that does not fall into the fault-tolerant time window, a recurrence plot of the transient characteristic waveform data of the node that does not fall into the fault-tolerant time window and the upstream and downstream nodes is constructed, and a determinism index is generated through recursive quantization analysis, including:

[0025] obtaining the transient characteristic waveform data of the node that does not fall into the fault-tolerant time window and the directly connected upstream and downstream nodes in the power grid topology;

[0026] reconstructing the transient characteristic waveform data of each node in phase space, and constructing a recurrence plot by calculating the Euclidean distance between the waveform data points of each node;

[0027] Based on the characteristics of the diagonal line structure in the recurrence plot, the proportion of the deterministic line in the recurrence plot is calculated as the determinism index.

[0028] Further, reconstructing the transient characteristic waveform data of each node in phase space by calculating the Euclidean distance between the waveform data points of each node includes:

[0029] calculating the delay time and embedding dimension for the transient characteristic waveform data of each node, respectively;

[0030] reconstructing the transient characteristic waveform data into state vectors in a high-dimensional phase space according to the delay time and embedding dimension;

[0031] calculating the Euclidean distance between the state vectors and generating a binary recurrence plot matrix according to a preset recurrence threshold.

[0032] Further, based on the transient characteristic waveform data time stamp sequence of the node that does not fall into the fault-tolerant time window and the upstream and downstream nodes, the transfer entropy algorithm is used to calculate the information flow direction and intensity between nodes, and by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology, a path consistency confidence index is generated, including:

[0033] extracting the transient characteristic waveform data time stamp sequence of the node that does not fall into the fault-tolerant time window and the directly connected upstream and downstream nodes in the power grid topology;

[0034] For each two adjacent nodes, the transfer entropy algorithm is used to calculate the information transmission amount from one node to another node;

[0035] The actual information flow direction calculated is matched with the theoretical causal flow direction determined according to the power grid topology connection relationship;

[0036] Based on the matching result, a path consistency confidence index indicating the consistency degree of the actual information flow direction and the theoretical causal flow direction is generated.

[0037] Further, the information flow direction and intensity between nodes are calculated using the transfer entropy algorithm, including:

[0038] For the timestamp sequence of two adjacent nodes, the source node and the historical length parameter of the historical node are set;

[0039] The joint probability distribution and the conditional probability distribution are calculated based on the frequency of the timestamp sequence value;

[0040] The transfer entropy value from the source node to the target node is calculated according to the probability distribution result to represent the information flow direction and intensity;

[0041] The actual information flow direction calculated is matched with the theoretical causal flow direction based on the power grid topology, including:

[0042] The theoretical causal flow direction relationship between nodes is determined according to the power grid topology structure;

[0043] The actual information flow direction calculated by the transfer entropy algorithm is compared with the theoretical causal flow direction one by one;

[0044] The proportion of the number of nodes with consistent flow direction to the total number of node pairs is counted to generate the path consistency confidence index.

[0045] Further, according to the determinacy index and the path consistency confidence index, the transient characteristic waveform data of the nodes not falling into the fault tolerance time window is weighted and corrected for electrical fault warning analysis, including:

[0046] A data credibility evaluation matrix is constructed based on the determinacy index and the path consistency confidence index;

[0047] According to the output result of the data credibility evaluation matrix, the processing mode of retaining, correcting or rejecting the transient characteristic waveform data of the nodes not falling into the fault tolerance time window is selected;

[0048] The processed transient characteristic waveform data is integrated with the transient characteristic waveform data of the nodes normally falling into the fault tolerance time window;

[0049] Based on the integrated multi-node transient characteristic waveform data, electrical fault type identification and fault location analysis are performed.

[0050] In another aspect, the application provides a wireless networking-based fire-fighting electrical early warning system, comprising:

[0051] A data acquisition module for acquiring electrical parameter data of an electrical line through a plurality of wireless sensor nodes and attaching local time stamps;

[0052] A waveform extraction module for receiving electrical parameter data by a cloud early warning platform to extract transient characteristic waveform data of a plurality of nodes within the same fault period;

[0053] A falling-in judgment module for selecting one wireless sensor node upstream on the topology as a timing reference node, calculating the dynamic fault tolerance window of each node downstream based on the time stamp of the timing reference node and the theoretical propagation delay, and judging whether it falls within the fault tolerance time window;

[0054] A recursive analysis module for constructing a recursive graph of transient characteristic waveform data of the node not falling within the fault tolerance time window and upstream and downstream nodes when there is a node not falling within the fault tolerance time window, and generating a determinacy index through recursive quantization analysis;

[0055] A confidence analysis module for calculating the information flow direction and strength between nodes based on the time stamp sequence of the transient characteristic waveform data of the node not falling within the fault tolerance time window and upstream and downstream nodes, and generating a path consistency confidence index by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology;

[0056] A fault early warning module for performing electrical fault early warning analysis on the transient characteristic waveform data of the node not falling within the fault tolerance time window after weighted correction according to the determinacy index and the path consistency confidence index.

[0057] The application has the following beneficial effects:

[0058] 1. The innovative multi-node time synchronization verification and data reliability evaluation mechanism significantly improves the accuracy and reliability of fault early warning, introduces the concept of timing reference node and dynamic fault tolerance window, effectively solves the inherent time asynchrony problem in wireless transmission, ensures the consistency of multi-node transient characteristic waveform data in time and space dimensions, and quantitatively evaluates the data that does not pass the time synchronization verification from two dimensions of waveform determinacy and information flow direction by using recursive graph analysis and transfer entropy algorithm, respectively, to generate comprehensive data quality index, which provides a scientific basis for subsequent data correction and integration. This multi-level and multi-dimensional data processing method fundamentally ensures that the data for fault analysis has high physical authenticity and time and space accuracy.

[0059] 2. By organically combining time fault tolerance judgment, waveform feature analysis and information flow verification, the system can adaptively identify and correct data deviation caused by wireless transmission characteristics, thereby maintaining the convenience of wireless networking while achieving monitoring accuracy close to wired systems. This systematic solution is particularly suitable for distributed electrical fire monitoring scenarios and can maintain stable early warning performance in complex electromagnetic environments and network conditions, providing more reliable technical support for fire safety. BRIEF DESCRIPTION OF DRAWINGS

[0060] Figure 1 Flowchart of the fire electrical early warning method based on wireless networking of the present application;

[0061] Figure 2 Structure diagram of the fire electrical early warning system based on wireless networking of the present application. DETAILED DESCRIPTION

[0062] The technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a 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 skilled in the art without creative labor fall within the scope of protection of the present application.

[0063] Embodiment 1: Figure 1 The fire electrical early warning method based on wireless networking of the present application is given, including:

[0064] S1, collecting electrical parameter data of an electrical line by multiple wireless sensor nodes and adding local time stamps;

[0065] S2, receiving the electrical parameter data by a cloud early warning platform to extract transient feature waveform data of multiple nodes in the same fault period;

[0066] S3, selecting one wireless sensor node upstream of the topology as a time reference node, calculating dynamic fault tolerance windows of downstream nodes based on the time stamp of the time reference node and the theoretical propagation delay, and judging whether the nodes fall within the fault tolerance time windows;

[0067] S4, when there are nodes that do not fall within the fault tolerance time windows, constructing a recurrence plot of the transient feature waveform data of the nodes that do not fall within the fault tolerance time windows and upstream and downstream nodes, and generating a determinacy index through recurrence quantification analysis;

[0068] S5, based on the time stamp sequence of the transient feature waveform data of the nodes that do not fall within the fault tolerance time windows and upstream and downstream nodes, calculating the information flow direction and strength between the nodes by using a transfer entropy algorithm, and generating a path consistency confidence index by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology;

[0069] S6、According to the certainty degree index and the path consistency confidence index, the transient characteristic waveform data of the node not falling into the fault-tolerant time window is weighted and corrected, and then the electrical fault early warning analysis is performed.

[0070] S1, collect electrical parameter data of electrical lines by multiple wireless sensor nodes and add local time stamps, including:

[0071] In the implementation process of the fire-fighting electrical early warning method based on wireless networking, data acquisition and time stamp addition are the basic steps of the whole method, and the specific implementation is as follows. A plurality of wireless sensor nodes are distributedly installed on the electrical lines to be monitored. These wireless sensor nodes are internally provided with high-precision current sensing elements and voltage sensing elements, which can measure the electrical parameters in the line in real time. The wireless sensor nodes work according to a pre-set sampling frequency, and the value of the sampling frequency ranges between 4000 times / s and 10000 times / s. The specific value is determined according to the rated frequency of the monitored electrical line and the highest frequency component of the transient fault characteristics to be captured. For example, for a line with a power frequency of 50Hz, the sampling frequency can be set to 6400 times / s to ensure that the tens of harmonic components and fault transient processes can be reliably captured. The sampling process is controlled by the microcontroller unit inside the wireless sensor node. The current and voltage sensing elements convert the analog signals into digital signals, which are read by the microcontroller unit and temporarily stored in the internal memory of the node to form electrical parameter data.

[0072] After each set of electrical parameter data acquisition is completed, the wireless sensor node immediately obtains the accurate time information of the current time from the high-precision clock chip built-in. The clock chip performs initial time synchronization by receiving the broadcast message of the wireless network time synchronization protocol after the node is powered on, and relies on its own crystal oscillator to keep time counting during subsequent operation. The microcontroller unit binds the read time information with the just collected electrical parameter data, and adds a corresponding local time stamp to this batch of data. The data format of the time stamp is the number of milliseconds calculated from the coordinated universal time epoch. After completing the time stamp addition, the microcontroller unit packs the electrical parameter data with the local time stamp and uploads it to the cloud early warning platform through wireless communication. Each wireless sensor node independently performs the above data acquisition and time stamp addition process without real-time coordination with other nodes, thereby avoiding additional communication overhead. This implementation ensures that the electrical parameter data obtained from the spatially distributed monitoring points are all accurately time-stamped, laying a solid foundation for subsequent analysis of the spatio-temporal correlation between multiple node data. The whole data acquisition and time stamp addition process is completely controlled by the embedded program, ensuring the reliability of the operation and the accuracy of the time sequence.

[0073] S2, the cloud early warning platform receives the electrical parameter data to extract the transient characteristic waveform data of multiple nodes in the same fault period, and the specific implementation includes:

[0074] After the cloud early warning platform receives the electrical parameter data uploaded by each wireless sensor node with a local timestamp, it starts to perform data analysis and fault feature extraction process. The platform first checks and decodes the received data packet, extracts the electrical parameter data and the corresponding local timestamp, and stores these data in the time sequence database in chronological order. The electrical parameter data with local timestamp includes current instantaneous value and voltage instantaneous value. These data are recorded continuously at a fixed sampling frequency, and each data point is labeled with a time mark accurate to milliseconds. The platform maintains a sliding time window to continuously monitor the latest received data. The window length is usually set to several seconds to tens of seconds, depending on the characteristics of the electrical system and the type of possible faults. For example, for power distribution systems, the time window length can be set to 5 seconds to ensure that the transient process of most short circuit faults can be captured completely.

[0075] When detecting the fault starting time based on the current effective value change in the electrical parameter data and determining the fault period, the platform calculates the current effective value sequence using a sliding window. When calculating, the current instantaneous value data in the time window of one power frequency cycle length before the current time is taken, and the current effective value estimate value at the current time is obtained through the operation process of squaring, averaging, and square root. Specifically, for a 50Hz system, take the sampling points in a 20ms time window. Assuming the sampling frequency is 6400Hz, the window contains 128 sampling points. Square the sampling values first, then calculate the arithmetic mean of the 128 square values, and finally square the mean value to get the current effective value. The fault detection threshold is obtained based on the historical current effective value data statistics of the line under normal operation, and is usually set to 1.2 to 1.5 times the historical average value. For example, for a line with a rated current of 100A, the fault detection threshold can be set to 120A. When the current effective value of multiple consecutive sampling points exceeds the threshold, the platform determines the time mark corresponding to the first out-of-limit point as the fault starting time. The determination of the fault period considers the fault feature duration, and is usually set to a time interval of 100ms to 500ms from the fault starting time, which is enough to cover the transient process of most faults.

[0076] When the current waveform data of each node in the fault period is intercepted from the electrical parameter data with local timestamps as transient feature waveform data, the platform first retrieves the electrical parameter data with local timestamps of all nodes in the time period from the time sequence database according to the time range of the fault period. Since there may be slight deviations in the local timestamps of each node, the platform re-samples the data of each node to a unified time axis by using a time alignment algorithm. The specific process of time alignment includes: taking the fault start time as the reference point, interpolating the electrical parameter data with local timestamps of each node to make them have the same time interval and the same time reference point. The accuracy requirement of time alignment is usually controlled within 1 ms, which is sufficient to ensure the accuracy of subsequent waveform analysis. The intercepted transient feature waveform data contains the current instantaneous value sequence of each node in the fault period, which retains the original time relationship and waveform characteristics. The platform stores these transient feature waveform data together with the corresponding node identification information, and each node's data contains complete time sequence information with uniform time stamp interval, ensuring that subsequent analysis can accurately reflect the electrical characteristic changes of each monitoring point when the fault occurs. The entire data processing process is executed in a pipeline manner to ensure processing efficiency and data consistency, providing a reliable data foundation for subsequent fault analysis.

[0077] During data processing, the platform also sets an abnormal data processing mechanism. When it is found that the electrical parameter data with local timestamps of a certain node is missing or abnormal, the platform will interpolate and compensate according to the data of adjacent time points, or use the historical data of the same period of the node for supplementation. For data with too large timestamp deviation, the platform will mark it as suspicious data and perform special processing in subsequent analysis. These measures ensure that even in the case of poor data quality of some nodes, usable transient feature waveform data can still be obtained, improving the robustness and reliability of the system. All parameter settings and algorithm choices in the processing process are based on professional knowledge in the field of electrical engineering, ensuring the implementability and effectiveness of the technical scheme.

[0078] S3, selects one wireless sensor node upstream of the selected topology as a time reference node, calculates the dynamic fault tolerance window of each node downstream based on the timestamp of the time reference node and the theoretical propagation delay, and judges whether it falls within the fault tolerance time window, including:

[0079] In the implementation process of the fire electrical early warning method based on wireless networking, time synchronization verification and fault tolerance analysis as a key step, the specific implementation is as follows. The cloud early warning platform first determines the wireless sensor node located at the upstream of the electrical line as the time reference node according to the preset electrical network topology information. The electrical network topology information is stored in the platform database in the form of a node connection relationship table, which records the installation position, connection order and electrical distance of all wireless sensor nodes. When determining the time reference node, the platform starts from the power line end and selects the first installed wireless sensor node as the time reference node along the power transmission direction of the electrical line. This selection is based on the characteristics of the fault waveform propagation of the power system, that is, the fault transient signal always propagates from the fault point to both sides, and the upstream node can detect the fault signal earliest, so the timestamp thereof is the most reliable reference.

[0080] With the timestamp of the time reference node as the reference, the platform calculates the dynamic fault tolerance window of each downstream node. The calculation process first needs to determine the theoretical propagation delay of the fault waveform from the time reference node to each downstream node. The calculation of the theoretical propagation delay is based on the wave propagation theory of the power line, considering factors such as line length, wave speed and propagation path. The wave speed depends on the line type and structural parameters, for example, for overhead lines, the wave speed is about 98% of the speed of light, that is, about 294 m / μs; for cable lines, the wave speed is about 50% to 60% of the speed of light, that is, about 150 m / μs to 180 m / μs. The platform automatically selects the corresponding wave speed value according to the line type, and calculates the theoretical propagation delay in combination with the electrical distance between nodes. The electrical distance is calculated by the line parameters and topology structure, including line length, branching situation, etc. For example, for an overhead line with a distance of 10 km between two nodes, the theoretical propagation delay is about 34 μs. On the basis of the theoretical propagation delay, the platform also needs to superimpose a preset transmission delay tolerance value, which considers factors such as clock synchronization error, sampling time deviation, measurement error, etc. The transmission delay tolerance value is usually set to 10% to 20% of the theoretical propagation delay, but not less than one sampling interval. For example, for a system with a sampling frequency of 6400 Hz, the sampling interval is 156 μs, and the transmission delay tolerance value is set to at least 156 μs. The calculation formula of the dynamic fault tolerance window is: the start time of the fault tolerance window is equal to the timestamp of the time reference node plus the theoretical propagation delay minus the transmission delay tolerance value, and the end time of the fault tolerance window is equal to the timestamp of the time reference node plus the theoretical propagation delay plus the transmission delay tolerance value.

[0081] When judging whether the time stamp of the transient characteristic waveform data of each downstream node falls within the dynamic fault tolerance window range of the corresponding node, the platform compares the time stamp of the transient characteristic waveform data of each node with the dynamic fault tolerance window time range calculated for it. The time stamp comparison adopts millisecond-level precision to ensure the accuracy of the judgment. For a certain node, if the time stamp of its transient characteristic waveform data falls between the start time and the end time of the dynamic fault tolerance window, it is considered that the time synchronization of the node is normal and the data is reliable; if the time stamp is earlier than the start time of the fault tolerance window or later than the end time of the fault tolerance window, it is considered that the node has time synchronization deviation, and the data reliability needs to be further verified. The platform performs this judgment process for each downstream node and records the judgment results. During the judgment process, the platform also considers special case processing, such as when the dynamic fault tolerance window time range of a certain node crosses the sampling interval boundary, the use of loop judgment logic ensures that no judgment is missed. All judgment results are stored in the analysis result database together with the identification information, time stamp information and fault tolerance window information of the node, providing a basis for subsequent data quality evaluation and fault analysis. The entire time synchronization verification process is completely automated and does not require human intervention, ensuring the consistency of analysis efficiency.

[0082] In the implementation process, the platform also establishes a perfect abnormality handling mechanism. When it is found that the time stamp of the transient characteristic waveform data of a certain node deviates significantly from the dynamic fault tolerance window range, the platform will automatically trigger the resynchronization process. The resynchronization process includes recalibrating the local clock of the node, reacquiring the time stamp information, and re-executing the time synchronization verification. If the time stamp still deviates from the fault tolerance window range after continuous multiple resynchronizations, the platform will mark the node as an abnormal node and start the backup time synchronization scheme. The backup scheme can use the time stamp of the adjacent node as a reference, or use historical contemporaneous data as a compensation benchmark. These abnormality handling mechanisms ensure that even in the case of time synchronization failure of some nodes, the system can still maintain normal operation, improving the reliability and robustness of the entire early warning system. All abnormality handling processes will generate detailed log records, including the abnormality occurrence time, abnormality type, handling measures and handling results, etc. These logs are used for subsequent system maintenance and performance optimization.

[0083] S4, constructing a recursive graph of the transient characteristic waveform data of the node not falling into the fault tolerance time window and the upstream and downstream nodes when there is a node not falling into the fault tolerance time window, generating a deterministic degree index through recursive quantization analysis, specific implementation including:

[0084] When there are nodes not falling into the fault-tolerant time window, the cloud early warning platform initiates the recursive graph construction and deterministic analysis process. The platform first obtains the node identifiers of the nodes not falling into the fault-tolerant time window from the stored fault analysis results, and then obtains the transient characteristic waveform data of these nodes and their directly connected upstream and downstream nodes according to the power grid topology connection relationship. The power grid topology connection relationship is stored in the form of an adjacency list, and each node records the information of its directly adjacent upstream and downstream nodes. When obtaining the transient characteristic waveform data, the platform retrieves the complete waveform data sequence of these nodes within the fault period from the time series database, ensuring that the data of each node contains enough sampling points for subsequent analysis. During the data acquisition process, the platform checks the integrity and quality of the data, and for data segments with missing or abnormal data, linear interpolation or repair methods based on historical data are used for preprocessing to ensure the accuracy of subsequent analysis.

[0085] When reconstructing the phase space of each node's transient characteristic waveform data, the platform first independently calculates two key parameters, delay time and embedding dimension, for each node's data. The delay time is calculated using the autocorrelation function method, which selects the delay corresponding to the first drop of the autocorrelation coefficient to 1-1 / e of the initial value as the optimal delay time by calculating the autocorrelation coefficient between the time series and its delayed version. The embedding dimension is determined using the false nearest neighbor method, which selects the dimension corresponding to the ratio of false nearest neighbors falling below a certain threshold, such as 5%, as the optimal embedding dimension by gradually increasing the embedding dimension. According to the calculated delay time and embedding dimension, the platform reconstructs the one-dimensional transient characteristic waveform data of each node into a sequence of state vectors in high-dimensional phase space. The reconstruction process is achieved through the time delay embedding method, where each state vector is composed of multiple points in the original time series separated by the delay time, and the dimension of the state vector is equal to the embedding dimension.

[0086] When calculating the Euclidean distance between each state vector and generating a binary recursive graph matrix based on a pre-set recursive threshold, the platform first calculates the Euclidean distance between all pairs of state vectors in the reconstructed phase space. The Euclidean distance calculation uses the standard formula, which is the square root of the sum of the squares of the differences between the corresponding dimensions of the two vectors. Then the platform sets the recursive threshold based on the statistical characteristics of all distance values, which is usually a certain proportion of the median or mean of all distance values, such as 10% to 20% of the standard deviation of the distance values. Based on this recursive threshold, the platform converts the distance matrix into a binary recursive graph matrix: when the distance between two state vectors is less than or equal to the recursive threshold, the corresponding position in the matrix is marked as 1, indicating a recursive point; otherwise, it is marked as 0. The rows and columns of the recursive graph matrix correspond to time points, and the 1 value points in the matrix indicate that the system states at these two times in the phase space are similar.

[0087] Based on the characteristics of the diagonal line structure in the recurrence plot, the platform first identifies the diagonal line structure in the recurrence plot. The diagonal line structure is composed of consecutive recurrence points, indicating that the system state evolves along a similar trajectory in the phase space. The platform scans the recurrence plot matrix and counts all diagonal line segments with a length greater than the minimum length threshold of 2 points. The total number of recurrence points contained in these diagonal line segments is calculated. The determinism index is defined as the ratio of the number of recurrence points in the diagonal line structure to the total number of recurrence points in the recurrence plot. This ratio reflects the degree of determinism of the system dynamic behavior. The platform independently calculates the determinism index for each node and stores the calculation results together with the node identification information, providing a basis for subsequent data quality evaluation and weighted processing. The entire recurrence plot construction and determinism analysis process is executed in batch mode, processing all node data uniformly to ensure analysis efficiency and consistency of analysis results.

[0088] In the implementation process, the platform also establishes a perfect parameter optimization mechanism. The calculation of delay time and embedding dimension uses an iterative optimization algorithm, which selects the optimal parameter combination through multiple calculations. The setting of the recurrence threshold uses an adaptive method, which dynamically adjusts the threshold size according to the noise level and signal characteristics of the waveform data. For special cases, such as waveform data containing large noise or abnormal interference, the platform will start data filtering processing, using a digital filter to preprocess the original waveform data, and then reconstruct the phase space after eliminating the noise effect. All these processing processes are equipped with detailed parameter records and process logs to ensure the traceability and repeatability of the analysis process. The platform also sets up an exception handling mechanism, which automatically starts the review process when the data analysis of a certain node produces abnormal results, recalculates the relevant parameters to ensure the reliability of the analysis results. These measures together ensure the accuracy and stability of the recurrence plot analysis and the calculation of the determinism index, providing a high-quality data basis for subsequent fault warning analysis.

[0089] S5, based on the transient characteristic waveform data timestamp sequence of the node and the upstream and downstream nodes that do not fall into the fault-tolerant time window, using the transfer entropy algorithm to calculate the information flow direction and strength between nodes, and by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology, generating a path consistency confidence index, the specific implementation includes:

[0090] Based on the transient characteristic waveform data timestamp sequence of the node not falling into the fault-tolerant time window and the upstream and downstream nodes, the cloud early warning platform performs information flow analysis and path verification process. The platform first extracts the transient characteristic waveform data timestamp sequence of the node not falling into the fault-tolerant time window and the upstream and downstream nodes directly connected in the power grid topology. In the extraction process, the platform determines the node pairs to be analyzed according to the connection relationship of the power grid topology, each node pair containing two directly connected nodes. The timestamp sequence is obtained from the time series database, containing the time mark information of the waveform data of each node in the fault period. The platform pre-processes the timestamp sequence, including time alignment and resampling, to ensure that the sequences of different nodes have the same time reference and sampling interval. In the pre-processing process, the platform uses linear interpolation method to align the timestamp sequence of each node to the unified time grid, and the interval of the time grid is determined according to the original sampling frequency, for example, for a 6400Hz sampling system, the time grid interval is 156μs. The pre-processed timestamp sequence is organized into a time series dataset, each node's sequence contains the same number of data points, and the time points correspond one by one, providing accurate data basis for subsequent information flow analysis.

[0091] For each pair of adjacent nodes, the transfer entropy algorithm is used to calculate the information transmission from one node to another. The transfer entropy algorithm is based on information theory and is used to quantify the directional information transfer between two time series. In the implementation process, first, set the source node and the history length parameter of the historical node for each node pair, the history length parameter determines how many past time information to consider, usually determined by trial and error method or based on data characteristics, for example, set the history length parameter k=2, l=1 to represent considering the information of the last 2 time steps of the source node and the last 1 time step of the target node. Calculate the joint probability distribution and conditional probability distribution based on the frequency of timestamp sequence values, estimate the joint probability and conditional probability by counting the frequency of each value combination. Specifically, the platform discretizes the timestamp sequence values into several intervals, counts the number of each interval combination, and then calculates the corresponding probability estimate. Calculate the transfer entropy value from the source node to the target node according to the probability distribution result, the size of the transfer entropy value represents the strength of the information flow, positive value represents information flow from the source node to the target node, negative value represents reverse flow. In the calculation process, the base of the logarithm is 2, and the result is in bits. The larger the transfer entropy value, the greater the information transmission from one node to another.

[0092] According to the topology connection relationship of the power grid, the theoretical causal flow direction is determined, and the actual information flow direction calculated is matched with the theoretical causal flow direction. The theoretical causal flow direction is determined according to the topology structure of the power grid, and based on the physical characteristics of the power system, the theoretical causal flow direction always propagates from the power supply side to the load side. According to the position relationship of the nodes in the power grid, the platform pre-defines the theoretical flow direction relationship of each node pair, for example, for the nodes connected downstream of the power supply, the theoretical flow direction is from the power supply side node to the load side node. When matching the mode, the actual information flow direction calculated by the transfer entropy algorithm is compared with the theoretical causal flow direction one by one, and for each node pair, it is checked whether the actual calculated information flow direction is consistent with the theoretical flow direction. The statistical significance of the transfer entropy value is considered in the comparison process, and only when the transfer entropy value exceeds the significance threshold, it is considered that the effective information flow direction is detected. The significance threshold is determined by the surrogate data method, and a plurality of groups of surrogate data with disturbed time sequence are generated, the transfer entropy distribution of these data is calculated, and the upper percentile of the distribution is taken as the threshold, for example, the 95% percentile. Through this method, the information flow direction misjudgment caused by random factors can be excluded.

[0093] Based on the matching result, a path consistency confidence index indicating the consistency degree of the actual information flow direction and the theoretical causal flow direction is generated. The platform counts the number of flow direction consistent node pairs in all node pairs, calculates the proportion of the number of flow direction consistent node pairs in the total number of node pairs, and obtains a preliminary consistency index. In order to further improve the reliability of the index, the platform also considers the intensity factor of the information flow, and gives higher weight to the node pairs with larger transfer entropy values. The calculation formula of the path consistency confidence index is the ratio of the number of consistent node pairs to the total number of node pairs, wherein the weight is determined according to the size of the transfer entropy value, for example, the normalized value of the transfer entropy value is used as the weight. The specific calculation process includes: first, calculating the maximum and minimum values of the transfer entropy values of all node pairs, then normalizing the transfer entropy value of each node pair to 0 to 1, and the normalized value is used as the weight of the node pair. For the node pairs with consistent flow direction, the weight is counted in the numerator, and the sum of the weights of all node pairs is used as the denominator, and finally the weighted path consistency confidence index is obtained. The path consistency confidence index obtained finally is a value between 0 and 1, and the closer the value is to 1, the higher the consistency between the actual information flow direction and the theoretical causal flow direction. The platform stores the index together with the node identification information and the analysis process parameters, which provides an important basis for subsequent data quality evaluation and fault analysis.

[0094] The entire analysis process is performed in parallel computing mode, and different computing resources are allocated to different node pairs to improve analysis efficiency. The platform also establishes a perfect quality control mechanism to monitor and handle abnormal situations during the analysis process, ensuring the accuracy and reliability of the analysis results. During the calculation process, the platform records the calculation parameters and intermediate results of each node pair, including the historical length parameter selection process, probability distribution estimation results, transfer entropy calculation values, and significance test results. These detailed information not only serves for subsequent result verification, but also provides data support for system performance optimization. For abnormal situations that occur during the calculation process, such as inaccurate probability estimation or divergent transfer entropy calculation, the platform will automatically adjust the parameters and recalculate, or start the backup analysis method. Through this comprehensive and detailed implementation, the reliability and accuracy of the information flow analysis and path verification process are ensured, providing an important data quality evaluation basis for electrical fault early warning.

[0095] S6、According to the determinacy index and the path consistency confidence index, the transient characteristic waveform data of the nodes not falling into the fault-tolerant time window is weighted and corrected for electrical fault early warning analysis, and the specific implementation includes:

[0096] According to the determinacy index and the path consistency confidence index, the cloud early warning platform performs data quality evaluation and correction processing. The platform first constructs a data reliability evaluation matrix based on the determinacy index and the path consistency confidence index. This matrix is a two-dimensional decision model, with the horizontal axis representing the determinacy index and the vertical axis representing the path consistency confidence index. The matrix divides the value range of these two indexes into multiple intervals, for example, the determinacy index is divided into high determinacy interval (0.8-1.0), medium determinacy interval (0.5-0.8), and low determinacy interval (0-0.5), and the path consistency confidence index is also divided into high confidence interval (0.8-1.0), medium confidence interval (0.5-0.8), and low confidence interval (0-0.5). Each interval combination corresponds to a specific data processing strategy, including complete retention, weighted correction, or complete rejection. The specific assignment of the matrix is determined by combining expert experience and historical data analysis, for example, when both indexes are in the high interval, the retention strategy is adopted, when one index is in the high interval and the other is in the low interval, the correction strategy is adopted, and when both indexes are in the low interval, the rejection strategy is adopted. The construction process of the matrix considers the needs of different application scenarios, and adjusts the interval boundaries and strategy mapping relationship to adapt to different monitoring requirements.

[0097] The output results of the data credibility evaluation matrix are used to select the processing method of the transient characteristic waveform data of the nodes that do not fall within the fault-tolerant time window, including reservation, correction or rejection. For the node data selected for reservation, the platform directly uses the original transient characteristic waveform data without any modification, but records the basis and parameters of the reservation decision. For the node data selected for correction, the platform calculates the correction weight according to the values of the two indicators, and the calculation formula of the correction weight is the geometric mean of the determinacy degree indicator and the path consistency confidence indicator. This calculation method can reflect the influence of the two indicators at the same time. In the correction process, the platform fuses the original waveform data with the waveform data of adjacent nodes by weighting, and the weighting coefficients are determined according to the correction weight. For example, when the correction weight is 0.8, the weight of the original data is 0.8 and the weight of the reference data is 0.2. The selection of reference data is based on the topological relationship of the power grid, and the data of the adjacent nodes with the shortest electrical distance are preferred. For the node data selected for rejection, the platform marks it as unusable data and excludes the influence of these data in subsequent analysis, while starting the data compensation mechanism to fill in the historical same period data or adjacent node data. All processing decisions and parameters are recorded in the data processing log to ensure the traceability of the processing process.

[0098] The processed transient characteristic waveform data is integrated with the transient characteristic waveform data of the nodes that normally fall within the fault-tolerant time window. In the integration process, the platform first establishes a unified time reference to align the waveform data of all nodes to the same time axis. The time alignment is realized by using the cubic spline interpolation method, which ensures that the data of different nodes have the same time resolution and time reference point, and the time alignment accuracy is controlled within 1 millisecond. For the reserved and corrected data, the platform directly integrates them into the integrated data set; for the rejected data, the platform uses the data of adjacent nodes or historical same period data for compensation, and the weight of the compensation data is determined according to the data quality indicator. The data integration adopts the time series merging algorithm to maintain the time synchronization and waveform integrity of each node data, and records the source and processing history of each data point. The integrated data set contains the transient characteristic waveform data of all available nodes, which has consistent time reference and data processing standard, providing a high-quality data basis for subsequent fault analysis. During the integration process, data consistency check is also performed to ensure that the data of different nodes are coordinated in terms of time relationship and waveform characteristics.

[0099] Based on the integrated multi-node transient characteristic waveform data, electrical fault type recognition and fault location analysis are performed. The fault type recognition adopts a multi-feature fusion analysis method to extract multiple characteristic parameters from the integrated waveform data, including waveform amplitude, phase, harmonic content, waveform distortion rate, etc. These characteristic parameters are input into a fault classification model based on machine learning algorithm, and the fault type is recognized by pattern matching algorithm, such as short circuit fault, ground fault, broken line fault, etc. The fault location adopts the principle of traveling wave distance measurement, and the fault point position is calculated by using the propagation time difference of fault transient waveform between nodes. In specific implementation, the platform selects the node that detects the fault feature earliest as the reference point, calculates the time delay of fault waveform propagation to other nodes, and calculates the fault distance combined with the wave velocity parameter, which is determined according to the line type, such as 294 meters per microsecond for overhead line and 150 meters per microsecond for cable line. The whole analysis process adopts multi-algorithm parallel execution mode, and different analysis strategies are adopted for different types of faults to ensure the accuracy and reliability of the analysis results. The analysis results include fault type, fault location, fault severity, etc. These information is finally used to generate warning report and guide fault handling decision. The platform also establishes an analysis result verification mechanism, which continuously optimizes the analysis algorithm and parameter setting by comparing the analysis results of multiple algorithms with the actual fault records.

[0100] In addition, the wireless sensor nodes can transmit the collected electrical parameter data with local time stamp to the cloud warning platform through the new generation of mobile communication network. Specifically, each wireless sensor node is built-in with a 5G communication module, accesses the 5G base station access network, transmits data to the Internet through the mobile core network, and finally reaches the cloud warning platform. The mobile telecommunication service provider specially configures a slice network for the warning data to provide high priority and low latency transmission channel, ensuring the real-time of critical alarm information. At the same time, the fiber broadband network provides a reliable transmission guarantee for the parallel processing and storage of massive waveform data of multiple nodes as the high-speed interconnection channel between cloud data centers. In areas where mobile communication network signal coverage is insufficient, the system automatically switches to the backup fiber network transmission channel to ensure the continuity of data transmission through wired mode. This hybrid networking method that combines new generation mobile communication and fiber broadband operation service not only takes advantage of the flexibility and convenience of wireless networking, but also provides reliable backup transmission path through wired network, significantly improving the communication reliability and service quality of the whole warning system. In addition, the system also integrates mobile voice service, which automatically notifies relevant personnel through voice call when a major electrical fault is detected, providing multiple alarm safeguards. Mobile data communication service supports operation and maintenance personnel to view system status and warning information in real time through mobile terminal, realizing remote monitoring and maintenance.

[0101] Example 2: Figure 2The application provides a structure diagram of a wireless network-based fire-fighting electrical early warning system.

[0102] A data acquisition module is configured to acquire electrical parameter data of the electrical line through the plurality of wireless sensor nodes and attach local time stamps;

[0103] A waveform extraction module is configured to receive the electrical parameter data by the cloud early warning platform to extract transient characteristic waveform data of the plurality of nodes in the same fault period.

[0104] A falling judgment module is configured to select one wireless sensor node upstream of the topology as a timing reference node, calculate dynamic fault tolerance windows of each node downstream based on the time stamp of the timing reference node and the theoretical propagation delay, and judge whether the nodes fall into the fault tolerance time windows.

[0105] A recursive analysis module is configured to construct a recursive graph of the transient characteristic waveform data of the nodes not falling into the fault tolerance time windows and the upstream and downstream nodes when there are nodes not falling into the fault tolerance time windows, and generate a determinacy degree index through recursive quantization analysis.

[0106] A confidence analysis module is configured to calculate the information flow direction and intensity between the nodes based on the time stamp sequence of the transient characteristic waveform data of the nodes not falling into the fault tolerance time windows and the upstream and downstream nodes by using a transfer entropy algorithm, and generate a path consistency confidence index by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology.

[0107] A fault early warning module is configured to perform electrical fault early warning analysis on the transient characteristic waveform data of the nodes not falling into the fault tolerance time windows after weighting correction according to the determinacy degree index and the path consistency confidence index.

[0108] In the embodiments, all the calculations are dimensionless numerical calculations, and the preset parameters and threshold values in the calculations are set by a person skilled in the art according to actual conditions.

[0109] It should be noted that the application can be deployed on the device itself to realize embedded application, or run on a PC terminal or other terminal with a user interface, so as to meet various hardware environments and use requirements.

[0110] The above-described embodiments can be implemented in whole or in part by software, hardware, firmware, or any combination thereof. When implemented by software, the above-described embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the processes or functions according to the embodiments of the present application are wholly or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable devices. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, for example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center in a wireless or wired manner. The wired transmission manner includes optical fiber, twisted pair, coaxial cable, etc. The wireless transmission includes infrared rays, microwaves, etc. The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server, data center, etc. containing one or more available medium collections. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state disk.

[0111] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device, and module can refer to the corresponding processes in the foregoing method embodiments, which will not be described here.

[0112] In several embodiments provided in the present application, it should be understood that the disclosed system, device, and method can be implemented in other ways. For example, the above-described device embodiments are only schematic, for example, the division of the modules is only a logical function division, and actual implementation can have another division manner, for example, a plurality of modules or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the shown or discussed ones can be indirect coupling or communication connection through some interfaces, devices, or modules, and can be electrical, mechanical, or other forms.

[0113] The modules illustrated as separate components can or can not be physically separate, and the components illustrated as modules can or can not be physical modules, which can be located in one place or distributed on a plurality of network modules. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0114] In addition, each functional module in the various embodiments of the present application can be integrated in one processing module, or each module can exist physically independently, or two or more modules can be integrated in one module.

[0115] If the functions are implemented in the form of software functional modules and sold or used as independent products, they can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions to make a computer device (which can be a personal computer, a server or a network device, etc.) execute all or part of the steps of the various embodiments of the method of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various media that can store program codes.

[0116] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0117] Finally: the above is only the preferred embodiment of the present application, and is not used to limit the present application, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A fire electrical early warning method based on wireless networking, characterized in that, include: S1. Collect electrical parameter data of electrical circuits through multiple wireless sensor nodes and attach local timestamps; S2. The cloud-based early warning platform receives electrical parameter data to extract transient characteristic waveform data of multiple nodes within the same fault period. S3. Select a wireless sensor node upstream of the topology as the timing reference node, calculate the dynamic fault tolerance window of each downstream node based on the timestamp of the timing reference node and the theoretical propagation delay, and determine whether it falls within the fault tolerance time window. S4. When there are nodes that do not fall within the fault tolerance time window, construct a recursive graph of the transient characteristic waveform data of the nodes that do not fall within the fault tolerance time window and their upstream and downstream nodes, and generate a deterministic index through recursive quantitative analysis. S5. Based on the transient characteristic waveform data timestamp sequence of nodes that do not fall into the fault tolerance time window and upstream and downstream nodes, the transfer entropy algorithm is used to calculate the information flow direction and intensity between nodes, and by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology, a path consistency confidence index is generated. S6. Based on the determinism index and the path consistency confidence index, the transient characteristic waveform data of nodes that do not fall within the fault tolerance time window are weighted and corrected before electrical fault early warning analysis is performed.

2. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, Electrical parameter data of electrical circuits are collected through multiple wireless sensor nodes and appended with local timestamps, including: Wireless sensor nodes collect current and voltage data from electrical circuits at a set sampling frequency as electrical parameter data. After each data acquisition, each wireless sensor node obtains the current time from the built-in clock chip and adds a corresponding local timestamp to the collected electrical parameter data.

3. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, The cloud-based early warning platform receives electrical parameter data to extract transient characteristic waveform data of multiple nodes within the same fault period, including: The cloud-based early warning platform receives electrical parameter data with local timestamps uploaded by each wireless sensor node; Detect the fault initiation time and determine the fault period based on changes in the effective value of current in electrical parameter data; The current waveform data of each node during the fault period is extracted from the electrical parameter data with local timestamps as transient characteristic waveform data.

4. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, A wireless sensor node upstream of the topology is selected as the timing reference node. Based on the timestamp of the timing reference node and the theoretical propagation delay, the dynamic fault tolerance window of each downstream node is calculated, and it is determined whether it falls within the fault tolerance time window, including: Based on the power grid topology, the wireless sensor node located at the upstream end of the electrical line is selected as the timing reference node; Based on the timestamp of the timing reference node, the theoretical propagation delay of the fault waveform from the timing reference node to each downstream node is added, and a preset transmission delay tolerance value is superimposed to calculate the dynamic fault tolerance window of each downstream node. Determine whether the timestamps of the transient characteristic waveform data of each downstream node fall within the dynamic fault tolerance window range of the corresponding node.

5. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, When nodes that do not fall within the fault tolerance time window exist, a recursive graph of the transient characteristic waveform data of the nodes that do not fall within the fault tolerance time window and their upstream and downstream nodes is constructed. Deterministic indices are generated through recursive quantization analysis, including: Acquire transient characteristic waveform data of nodes that do not fall within the fault-tolerant time window and their corresponding upstream and downstream nodes directly connected in the power grid topology; Phase space reconstruction is performed on the transient characteristic waveform data of each node, and a recursive graph is constructed by calculating the Euclidean distance between the waveform data points of each node; Based on the characteristics of the diagonal structure in the recursive graph, the proportion of deterministic lines in the recursive graph is calculated as a deterministic index.

6. The fire electrical early warning method based on wireless networking according to claim 5, characterized in that, Phase space reconstruction is performed on the transient characteristic waveform data of each node, and a recursive graph is constructed by calculating the Euclidean distance between waveform data points of each node, including: Calculate the delay time and embedding dimension for the transient characteristic waveform data of each node; Based on the delay time and embedding dimension, the transient feature waveform data is reconstructed into a state vector in a high-dimensional phase space; Calculate the Euclidean distance between each state vector and generate a binary recursive graph matrix based on a preset recursion threshold.

7. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, Based on the timestamp sequence of transient characteristic waveform data of nodes that do not fall within the fault-tolerant time window and their upstream and downstream nodes, the transfer entropy algorithm is used to calculate the information flow direction and intensity between nodes. By comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology, a path consistency confidence index is generated, including: Extract the transient characteristic waveform data timestamp sequence of nodes that do not fall within the fault-tolerant time window and the corresponding upstream and downstream nodes directly connected to the nodes in the power grid topology; For each pair of adjacent nodes' timestamp sequences, the transfer entropy algorithm is used to calculate the amount of information transmitted from one node to another. The theoretical causal flow direction is determined based on the power grid topology connection relationship, and the calculated actual information flow direction is matched with the theoretical causal flow direction. Based on the matching results, a path consistency confidence index is generated, which represents the degree of consistency between the actual information flow and the theoretical causal flow.

8. The fire electrical early warning method based on wireless networking according to claim 7, characterized in that, The transfer entropy algorithm is used to calculate the direction and intensity of information flow between nodes, including: For the timestamp sequences of two adjacent nodes, set the historical length parameters for the source node and the historical node; Calculate the joint probability distribution and conditional probability distribution based on the occurrence frequency of timestamp sequence values; The transfer entropy value from the source node to the target node is calculated based on the probability distribution results to characterize the direction and intensity of information flow; Pattern matching between the calculated actual information flow direction and the theoretical causal flow direction based on the power grid topology includes: Determine the theoretical causal flow relationships between nodes based on the power grid topology; The actual information flow direction calculated by the transfer entropy algorithm is compared one by one with the theoretical causal flow direction; The proportion of node pairs with consistent flow direction to the total number of node pairs is used to generate a path consistency confidence index.

9. The fire electrical early warning method based on wireless networking according to claim 1, characterized in that, Based on the determinism index and the path consistency confidence index, the transient characteristic waveform data of nodes that do not fall within the fault tolerance time window are weighted and corrected before electrical fault early warning analysis is performed, including: A data credibility assessment matrix is ​​constructed based on the determinism index and the path consistency confidence index; Based on the output of the data credibility assessment matrix, select the processing method of retaining, correcting or eliminating the transient characteristic waveform data of nodes that do not fall within the fault tolerance time window; The processed transient characteristic waveform data is integrated with the transient characteristic waveform data of nodes that normally fall within the fault-tolerant time window; Electrical fault type identification and fault location analysis are performed based on the integrated multi-node transient characteristic waveform data.

10. A fire electrical early warning system based on wireless networking, used to implement the fire electrical early warning method based on wireless networking as described in any one of claims 1-9, characterized in that, include: The data acquisition module is used to collect electrical parameter data of electrical circuits through multiple wireless sensor nodes and attach local timestamps; The waveform extraction module is used by the cloud-based early warning platform to receive electrical parameter data and extract transient characteristic waveform data of multiple nodes within the same fault period. The fall-in judgment module is used to select a wireless sensor node upstream of the topology as the timing reference node, calculate the dynamic fault tolerance window of each downstream node based on the timestamp of the timing reference node and the theoretical propagation delay, and determine whether it falls within the fault tolerance time window. The recursive analysis module is used to construct a recursive graph of transient characteristic waveform data of nodes that do not fall within the fault tolerance time window and their upstream and downstream nodes when there are nodes that do not fall within the fault tolerance time window. It generates a deterministic index through recursive quantitative analysis. The confidence analysis module is used to calculate the information flow direction and intensity between nodes based on the transient characteristic waveform data timestamp sequence of nodes that do not fall within the fault tolerance time window and upstream and downstream nodes. It also generates a path consistency confidence index by comparing the actual information flow direction with the theoretical causal flow direction based on the power grid topology. The fault early warning module is used to perform electrical fault early warning analysis by weighting and correcting the transient characteristic waveform data of nodes that do not fall within the fault tolerance time window based on the determinism index and the path consistency confidence index.

Images