Fire protection early warning method and system for converter station

Abstract

The invention discloses a fire protection early warning method and system for a converter station, and belongs to the field of safety protection, and the method comprises the steps: obtaining real-time current data, carrying out the Fourier transform operation of the real-time current data, obtaining a real-time current waveform, and extracting an abnormal current waveform from the real-time current waveform; determining a waveform parameter of the abnormal current waveform, and determining a corresponding probability grade value according to the waveform parameter by using a preset support vector machine; and when the probability grade value is greater than a preset probability grade value, searching a plurality of potential fault points according to the waveform parameters, and triggering corresponding safety early warning processing according to the real-time images of the potential fault points and the temperature time sequence parameters. According to the invention, the equipment in the converter station is monitored at the same time through the current data, the pressure data, the temperature and the real-time image, the fire hazard grade value can be accurately determined, the alarm is triggered, the assessment precision of safety accidents can be improved, different alarms are triggered according to different fire hazard grade values, the timeliness of early warning is improved, and the alarm delay is shortened.

Description

Technical Field

[0001] This invention relates to the technical field of signal alarms, and in particular to a fire protection and early warning method and system for converter stations. Background Technology

[0002] With the continuous expansion of the power grid and the continuous improvement of the voltage level of high-voltage direct current transmission, the number and capacity of power equipment in converter stations are constantly increasing. As the core hub of the power system, the safety of converter stations is directly related to the safe and stable operation of the power grid.

[0003] Because the equipment in a converter station is highly concentrated, operating under high temperature and high pressure, any abnormal signal from any device can easily generate a strong pressure wave due to current coupling, igniting surrounding flammable insulating materials in the high-temperature area and triggering a chain reaction of explosions and fires. Therefore, early fire warnings for converter stations are necessary. One commonly used fire prevention and early warning method involves installing multiple different detection terminals (including temperature sensors and electrical parameter sensors) within the converter station. These terminals monitor real-time voltage, current, and temperature parameters of the equipment, comparing each parameter with its corresponding preset threshold. Based on the comparison results, a fire assessment is made, and an early warning is triggered.

[0004] However, the commonly used methods have the following technical problems: relying on a single parameter to compare with a threshold often triggers an early warning only when the numerical deviation is large, and by then a fire has already occurred. Therefore, the existing early warning methods are delayed and have poor timeliness. Summary of the Invention

[0005] This invention provides a fire protection early warning method and system for converter stations, which can solve the technical problems of existing technologies that use the comparison results of a single parameter and a threshold for early warning, which have delays and poor timeliness.

[0006] A first aspect of this invention provides a fire protection and early warning method for converter stations, the method comprising: Real-time current data is acquired and Fourier transform is performed on the real-time current data to obtain a real-time current waveform. Abnormal current waveforms are extracted from the real-time current waveform. The real-time current data is the real-time data of the equipment in the valve hall of the converter station. The waveform parameters of the abnormal current waveform are determined, and a preset support vector machine is used to determine the corresponding probability level value based on the waveform parameters. The probability level value is the fault level corresponding to the probability of an electric arc caused by equipment overload. When the probability level value is greater than the preset probability level value, several potential fault points are found according to the waveform parameter size, and the pressure wave superposition feature value from the several potential fault points is obtained. The location coordinates of each pressure wave superposition feature value are determined, and the temperature time series parameters and real-time images of each location coordinate are obtained. The fire level value corresponding to each location coordinate is determined based on the real-time images and the temperature time series parameters. When any fire level value is greater than a preset fault level value, the corresponding safety warning processing is triggered based on the fire level value.

[0007] In conjunction with the first aspect, in one implementation, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time-series parameters includes: The preset first image recognition model is invoked to identify the number of objects in the real-time image and the distance between each object and the device, and the object number value and the distance value are obtained respectively. Calculate the quantity weight value based on the number of objects and the interval distance value; Calculate the temperature change value of the temperature time series parameter, and use the temperature change value and the quantity weight value to determine the fire level value.

[0008] In conjunction with the first aspect, in one implementation, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time-series parameters includes: The preset second image recognition model is invoked to identify the objects in each real-time image, thereby obtaining several material information. Based on the information of each material, the corresponding ignition temperature value of each material is found, and the K-means algorithm is called to cluster the multiple ignition temperature values ​​to obtain several ignition categories; The number of ignition temperature values ​​for each ignition category is counted to obtain a temperature quantity value, and the temperature peak value of the temperature time sequence parameter is calculated. The fire level is determined using the temperature quantity and the temperature peak value.

[0009] In conjunction with the first aspect, in one implementation, the waveform parameters include: frequency value and amplitude value; The step of using a preset support vector machine to determine the corresponding probability level value based on the waveform parameters includes: The amplitude change rate is calculated using the amplitude value, and the frequency value, the amplitude value, and the amplitude change rate are converted into a vector to obtain a parameter feature vector. A preset support vector machine is invoked to perform classification processing based on the parameter feature vectors to obtain the arc probability value; Based on the arc probability value, a target interval value is determined from several preset interval values, and the probability level value corresponding to the target interval value is determined.

[0010] In conjunction with the first aspect, in one implementation, the step of finding several potential fault points based on the waveform parameter magnitude and obtaining pressure wave superposition feature values ​​from the several potential fault points includes: By determining the location coordinates of the device corresponding to each abnormal current waveform, several potential fault points are obtained; Based on the pressure sensor at the potential fault point, real-time pressure wave waveforms are obtained, and the values ​​of peak superposition, valley offset, and energy distribution are extracted from each real-time pressure wave waveform. The pressure combination value is obtained by combining the peak superposition degree value, the trough offset value, and the energy distribution value. Several pressure wave superposition feature values ​​are then selected from the multiple pressure combination values. The pressure wave superposition feature values ​​are pressure combination values ​​with values ​​greater than a preset pressure value.

[0011] In conjunction with the first aspect, in one implementation, the step of acquiring real-time current data and performing a Fourier transform operation on the real-time current data to obtain a real-time current waveform, and extracting an abnormal current waveform from the real-time current waveform, includes: A number of real-time current data points are acquired and each of the real-time current data points is preprocessed to obtain a number of processed data points. Each of the processed data points is then converted into a real-time current waveform using a fast Fourier transform. The preprocessing includes noise reduction and time-domain transformation. Calculate the total harmonic distortion rate of each of the real-time current waveforms, and determine the abnormal current waveform from the real-time current waveforms based on the total harmonic distortion rate.

[0012] In conjunction with the first aspect, in one implementation, after the step of acquiring the temperature time-series parameters and real-time images for each of the location coordinates, the method further includes: If the temperature time series parameter meets the threshold or the preset BP model is called to identify that the real-time image contains a heat source, the coordinates of the heat source are determined and an isolation boundary region is constructed with the coordinates of the heat source as the center. Shut down the equipment within the isolation boundary area and activate the emergency alarm process.

[0013] A second aspect of this invention provides a fire early warning system for a converter station, the system comprising: The acquisition module is used to acquire real-time current data and perform Fourier transform on the real-time current data to obtain a real-time current waveform, and extract abnormal current waveforms from the real-time current waveform. The real-time current data is the real-time data of the equipment in the valve hall of the converter station. The determination module is used to determine the waveform parameters of the abnormal current waveform and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters, wherein the probability level value is the fault level corresponding to the probability of an arc caused by equipment overload. The extraction module is used to find several potential fault points based on the waveform parameter magnitude when the probability level value is greater than the preset probability level value, and to obtain the pressure wave superposition feature value from the several potential fault points. The early warning module is used to determine the location coordinates of each pressure wave superposition feature value, and to acquire the temperature time series parameters and real-time image of each location coordinate. Based on the real-time image and the temperature time series parameters, the module determines the fire level value corresponding to each location coordinate. When any fire level value is greater than a preset fault level value, the module triggers the corresponding safety early warning process based on the fire level value.

[0014] Compared with existing technologies, the fire protection and early warning method and system for converter stations provided by this invention have the following advantages: This invention can acquire real-time current data and perform Fourier transform on the real-time current data to obtain real-time current waveforms, extract abnormal current waveforms from the real-time current waveforms; determine the waveform parameters of the abnormal current waveforms, and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters; when the probability level value is greater than the preset probability level value, find several potential fault points according to the waveform parameter magnitude, and obtain the pressure wave superposition feature value from the several potential fault points; determine the location coordinates of each pressure wave superposition feature value, and obtain the temperature time series parameters and real-time image of each location coordinate; determine the fire level value corresponding to each location coordinate based on the real-time image and temperature time series parameters; when any fire level value is greater than the preset fault level value, a safety early warning process is triggered. This invention monitors equipment in a converter station simultaneously using current data, pressure data, temperature, and real-time images. It can accurately determine the fire level and trigger alarms, which not only improves the accuracy of safety accident assessment but also allows different fire levels to trigger different alarms, thus reminding technicians to carry out maintenance and repairs earlier. This improves the timeliness of early warnings, shortens alarm delays, and enables technicians to perform maintenance as soon as possible, further reducing equipment risks and ensuring the safety of technicians. Attached Figure Description

[0015] Figure 1 This is a schematic flowchart of a fire protection and early warning method for a converter station provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a fire early warning system for a converter station according to an embodiment of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] With the rapid development of high-voltage technology and the continuous expansion of the power grid, converter stations are carrying increasingly higher voltage levels and power densities. One common type of converter station is the ultra-high voltage (UHV) station. As the core hub of the power system, the safety of UHV stations is directly related to the safe and stable operation of the power grid.

[0018] Because the equipment in the converter station is highly concentrated, the operating environment is not only high-temperature and high-pressure, but also prone to arcing due to current coupling when any equipment experiences a signal anomaly. During the arc expansion phase, the rapid injection of external energy creates a high-temperature and high-pressure plasma channel. The pressure difference at the gas-liquid interface pushes the liquid medium to form a shock wave, and the superposition of shock waves at different velocities creates a primary pressure wave. The intense pressure wave and arc ignite surrounding flammable insulating materials in the high-temperature area, triggering a chain reaction of explosions and fires. For example, when an equipment experiences an overload fault, the current coupling effect causes the fault signal to propagate nonlinearly between equipment, rapidly evolving a localized overload of a single device into a systemic fault involving multiple critical nodes. This fault propagation process generates a strong pressure wave. When the pressure wave undergoes complex refraction and reflection between various components in the valve hall, a multi-point impact superposition effect is formed, further exacerbating the physical damage to the equipment and the destruction of insulating materials. If the high-temperature area generated by the fault process easily ignites surrounding flammable insulating materials, the fire will spread rapidly.

[0019] Therefore, it is necessary to conduct fire early warning for converter stations as early as possible. One commonly used fire protection early warning method is to set up multiple different detection terminals (including temperature sensors and electrical parameter sensors) in the converter station. By monitoring the real-time voltage, current and temperature parameters of the equipment, the monitored parameters are compared with their corresponding preset thresholds. Based on the comparison results, a fire judgment is made and an early warning is triggered.

[0020] However, the commonly used methods have the following technical problems: relying on a single parameter to compare with a threshold often triggers an early warning only when the numerical deviation is large, and by then a fire has already occurred. Therefore, the existing early warning methods are delayed and have poor timeliness.

[0021] To address the aforementioned issues, the following specific embodiments will provide a detailed description and explanation of a fire protection and early warning method and system for converter stations provided in this application.

[0022] To address the technical issues of delays and poor timeliness in existing technologies that rely on comparing a single parameter with a threshold for early warning, this paper refers to... Figure 1 The diagram shows a flowchart of a fire protection and early warning method for a converter station according to an embodiment of the present invention.

[0023] In one embodiment, the fire protection early warning method for the converter station can be applied to the central control platform of the converter station. Multiple sensing terminals (such as temperature sensors, pressure sensors, cameras, and other various detection instruments) can be installed within the converter station. The central control platform can communicate with each sensing terminal, collecting different data through the terminals. Based on the data collected by the sensing terminals, the central control platform performs fire or fault analysis and immediately triggers an early warning or alarm when a fire or fault risk is identified, thereby improving the timeliness of the early warning and preventing major accidents.

[0024] As an example, the fire protection early warning method for the converter station may include: S11. Acquire real-time current data and perform Fourier transform on the real-time current data to obtain a real-time current waveform. Extract abnormal current waveforms from the real-time current waveforms. The real-time current data is the real-time data of the equipment in the converter station valve hall.

[0025] In one embodiment, one of the reasons why converter stations are prone to fire accidents is the generation of electric arcs due to equipment failure, and the high temperature of the arcs can ignite surrounding objects, causing a fire. Therefore, real-time current data of the equipment in the converter station's valve hall can be acquired. Based on this real-time current data, it can be determined whether there is equipment failure or whether there is faulty equipment. For example, it can be determined whether the real-time current data has abnormal values; if so, it can be determined that there is faulty equipment.

[0026] If there is a faulty device, perform a Fourier transform operation on the real-time current data to convert the real-time current data into a real-time current waveform, and then extract the abnormal current waveform from the real-time current waveform.

[0027] Alternatively, the current waveform of each device can be detected by an inspection instrument to obtain the real-time current waveform directly, and then the abnormal current waveform can be extracted based on the frequency or amplitude of the real-time current waveform.

[0028] To accurately extract abnormal current waveforms and improve the accuracy of subsequent analysis and early warning, as an example, the process of acquiring real-time current data and performing a Fourier transform operation on the real-time current data to obtain a real-time current waveform, and then extracting the abnormal current waveform from the real-time current waveform, may include the following sub-steps: S111. Acquire several real-time current data and preprocess each real-time current data to obtain several processed data. Convert each processed data into a real-time current waveform through fast Fourier transform. The preprocessing includes noise reduction processing and time-domain transformation processing.

[0029] S112. Calculate the total harmonic distortion rate of each of the real-time current waveforms, and determine the abnormal current waveform from the real-time current waveforms based on the total harmonic distortion rate.

[0030] Since there are multiple devices in the valve hall of the converter station, each device can be equipped with a detection instrument to obtain real-time current data of each device, thus obtaining several real-time current data points.

[0031] The acquired real-time current signals in the time domain are preprocessed, including noise reduction and time-domain transformation. Preprocessing removes DC components and noise interference, resulting in a standardized current signal sequence. A Fast Fourier Transform (FFT) is then performed on the standardized current signal sequence to convert each time-domain real-time current signal to a frequency-domain real-time current signal, yielding the real-time current waveform. The amplitude and phase information of the frequency components of the real-time current waveform are then obtained. This amplitude and phase information is used to determine the amplitude proportions of the fundamental and harmonic frequencies, thus obtaining the spectral distribution data.

[0032] The total harmonic distortion (THD) rate is calculated using the fundamental and harmonic components of the spectral distribution data. Specifically, it can be calculated as the ratio (expressed as a percentage) of the root mean square (RMS) value of the harmonic content to the RMS value of its fundamental component. If the THD rate exceeds a preset distortion threshold, an abnormal waveform distortion state is identified. The abnormal current waveform can be identified as the real-time current waveform with a THD rate exceeding the preset distortion threshold.

[0033] For spectral distribution data that indicates abnormal waveform distortion, harmonic components with amplitudes exceeding a preset threshold ratio of the fundamental amplitude are extracted. The frequency and amplitude values ​​corresponding to these harmonic components are recorded. The maximum amplitude value is identified from the recorded data as the abnormal fluctuation amplitude, and the corresponding frequency value is identified as the abnormal fluctuation frequency.

[0034] Specifically, real-time current signal data acquisition is typically achieved through current transformers. These transformers are installed at key locations in various equipment within the valve hall, converting high-voltage, high-current signals into smaller, more measurable current signals. The acquisition system can obtain thousands of sampling points per second, generating continuous real-time current data.

[0035] In one possible implementation, the preprocessing mainly targets interference components in the original signal. DC component removal is achieved by calculating the average value over one period and subtracting it from the original signal. Noise interference elimination employs moving averages or low-pass filtering to retain the main frequency components of the current signal while filtering out high-frequency noise. The preprocessed, standardized current signal sequence exhibits clearer waveform characteristics, laying the foundation for subsequent frequency domain analysis. The fast Fourier transform (FFT) can decompose the time-varying current waveform into sinusoidal components of different frequencies. The fundamental frequency is typically 50Hz, while harmonics are integer multiples of the fundamental frequency, such as the second harmonic at 100Hz, the third harmonic at 150Hz, etc. Each frequency component has corresponding amplitude and phase information; the amplitude reflects the intensity of that frequency component in the original signal, and the phase represents the time offset of that component relative to a reference point.

[0036] It's important to note that total harmonic distortion (THD) is a crucial indicator of current waveform quality. Its calculation involves a comprehensive evaluation of all harmonic components: first, the squares of each harmonic amplitude are calculated, then they are summed and the square root is taken, and finally compared with the fundamental amplitude. When equipment is operating normally, the current waveform is close to a sine wave, resulting in a low THD; however, when an anomaly occurs, the waveform becomes distorted, manifesting as an increase in harmonic content. The preset distortion threshold is typically determined based on the specific requirements of the power system and the characteristics of the equipment.

[0037] Specifically, identifying abnormal fluctuations requires in-depth analysis of spectral data. When the total harmonic distortion rate exceeds a threshold, harmonic components with larger amplitudes are further screened out. These significant harmonic components are often associated with specific causes of anomalies.

[0038] For example, the 5th and 7th harmonics may originate from the nonlinear characteristics of the converter, while the 3rd harmonic may be related to unbalanced loads. By recording the frequency and amplitude of these major harmonics, the characteristics of abnormal fluctuations can be accurately located. The maximum amplitude value directly reflects the severity of the anomaly, while the corresponding frequency value helps to analyze the root cause of the anomaly, providing crucial information for subsequent fault diagnosis and handling.

[0039] S12. Determine the waveform parameters of the abnormal current waveform, and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters, wherein the probability level value is the fault level corresponding to the probability of an electric arc caused by equipment overload.

[0040] After identifying one or more abnormal current waveforms, waveform parameters for each abnormal current waveform can be obtained, including the amplitude of the abnormal current waveform. Next, the amplitude of the abnormal current waveform can be input into a preset support vector machine to assess the probability of arcing. Based on this probability, the level of fire accident severity is determined to ascertain whether further early warning analysis is necessary.

[0041] In an optional embodiment, the waveform parameters include: frequency value and amplitude value; the step of determining the corresponding probability level value using a preset support vector machine based on the waveform parameters may include the following sub-steps: S121. Calculate the amplitude change rate using the amplitude value, and convert the frequency value, the amplitude value, and the amplitude change rate into a vector to obtain the parameter feature vector.

[0042] S122. Call the preset support vector machine to perform classification processing based on the parameter feature vector to obtain the arc probability value.

[0043] S123. Determine a target interval value from several preset interval values ​​based on the arc probability value, and determine the probability level value corresponding to the target interval value.

[0044] After obtaining the waveform parameters of the abnormal current waveform, the waveform parameters can include fluctuation frequency and amplitude data. The amplitude value can be obtained, the difference between the amplitude value at the current moment and the amplitude value at the adjacent moment can be calculated, and then the ratio of the difference to the time interval can be calculated to obtain the amplitude change rate.

[0045] After converting the frequency value, amplitude value, and amplitude change rate into vectors to obtain the parameter feature vector, the support vector machine algorithm can be used to classify the parameters based on the parameter feature vector. The probability value of overload arcing of the equipment can be obtained by calculating the distance between the decision function value and the classification boundary.

[0046] Multiple preset interval values ​​can be set in advance, for example, 0-10 as the first interval value, 10-20 as the second interval value, 20-30 as the third interval value, and so on. Then, each interval value corresponds to a probability level value. The probability level value of the 0-10 interval value is 1, the probability level value of the 10-20 interval value is 2, and finally, the preset interval value corresponding to the arc probability value can be determined.

[0047] Alternatively, the arc probability value can be compared with a preset probability threshold. If the probability value is less than the first threshold, the probability level is determined to be 1. If the probability value is between the first threshold and the second threshold, the probability level is determined to be 2. If the probability value is greater than the second threshold, the probability level is determined to be 3.

[0048] Specifically, acquiring the fluctuation frequency and amplitude data of abnormal current waveforms is fundamental to arc fault diagnosis. This data originates from prior spectral analysis and includes the specific values ​​of each frequency component and its corresponding amplitude information. The amplitude change rate is calculated to reflect the dynamic characteristics of the current fluctuations; it is obtained by dividing the amplitude difference between adjacent sampling times by the time interval.

[0049] For example, if the amplitude is 100A at the current moment and changes to 120A after 0.01 seconds, then the rate of change of amplitude is 2000A / s. This rate of change can reflect the abrupt change characteristics of the electric arc.

[0050] It should be noted that frequency values ​​typically fall within a specific range; for example, arc faults often generate characteristic frequencies between 2kHz and 10kHz. Amplitude values ​​reflect the severity of the anomaly, while the amplitude change rate reflects the speed of fault development. This invention can construct a feature vector from amplitude, frequency, and amplitude change rate together, and then call a preset support vector machine to perform classification processing based on the above three feature vectors to obtain the probability of arcing. By comprehensively considering information from multiple dimensions, the accuracy of subsequent assessments can be further improved.

[0051] In one possible implementation, the support vector machine (SVM) calculates probability values ​​by using the sigmoid function or other probability mapping methods to calculate the distance to the classification boundary, and then converts the distance into a probability value between 0 and 1. When the distance is positive and large, the probability is close to 1, indicating that an electric arc is highly likely to occur; when the distance is negative and the absolute value is large, the probability is close to 0, indicating that the probability of an electric arc is very small.

[0052] Alternatively, probability levels can be categorized using segmented thresholds. The first threshold is typically set around 0.3, and the second threshold around 0.7. This grading method allows equipment maintenance personnel to take appropriate measures based on different risk levels. A low probability level means the equipment is basically normal and only requires routine monitoring; a medium probability level indicates a potential risk and requires increased inspections; a high probability level indicates that an arc fault is about to occur or has already occurred, requiring immediate action.

[0053] S13. When the probability level value is greater than the preset probability level value, find several potential fault points according to the waveform parameter size, and obtain the pressure wave superposition feature value from the several potential fault points.

[0054] When the probability level value is greater than the preset probability level value, it indicates a higher probability of arcing, and also a higher probability of fire and malfunction. To determine the location of potential faults as early as possible, several potential fault points can be identified based on the waveform parameter magnitude. These potential fault points represent the locations of the faulty equipment. Because each device corresponds to a real-time current data point, if that device malfunctions, its real-time current data will become an abnormal current waveform. The probability level value calculated based on the waveform parameters of this abnormal current waveform will be greater than the preset probability level value. Therefore, the location of the device with this abnormal current waveform can be obtained, thus identifying a potential fault point.

[0055] If the probability level value is lower than the preset probability level value, the equipment can be inspected or technicians can be notified to check it, further reducing the risk of an accident. For example, when the probability level value is lower than the preset probability level value, technicians can be notified to conduct a routine location check on the equipment.

[0056] Specifically, identifying potential fault locations can also rely on the correlation between frequency characteristics and the physical characteristics of the equipment. Equipment in different locations has different inherent frequencies; the inherent frequency of a converter valve might be around 3kHz, while the inherent frequency of a busbar might be around 5kHz. When a detected abnormal frequency is close to the inherent frequency of a particular piece of equipment, that equipment becomes the focus of investigation. Simultaneously, the priority and scope of investigation can be determined by probability levels: at high probability levels, focus on inspecting the equipment with the highest frequency matching degree and its adjacent equipment; at medium probability levels, expand the inspection scope to the entire area; at low probability levels, only routine inspections are needed. This hierarchical and zoned fault location method greatly improves the efficiency and accuracy of fault diagnosis.

[0057] In one embodiment, the step of finding several potential fault points based on the waveform parameter magnitude and obtaining pressure wave superposition feature values ​​from the several potential fault points may include the following sub-steps: S131. Determine the location coordinates of the device corresponding to each abnormal current waveform to obtain several potential fault points.

[0058] S132. Based on the pressure sensor of the potential fault point, obtain the real-time pressure wave waveform, and extract the peak superposition value, valley offset value and energy distribution value from each real-time pressure wave waveform.

[0059] S133. Combine the peak superposition degree value, the trough offset value and the energy distribution value to obtain a pressure combination value, and select several pressure wave superposition feature values ​​from multiple pressure combination values. The pressure wave superposition feature value is a pressure combination value with a value greater than a preset pressure value.

[0060] Because of a device failure, the generated data may trigger several device anomalies, and several abnormal current waveforms may be selected. The corresponding location coordinates of several abnormal or faulty devices can be obtained, thereby identifying several potential fault points.

[0061] Meanwhile, because the device with the abnormal current waveform is a faulty device, the data generated by the downstream device based on the erroneous signal may not be incorrect. If there is no error, the waveform corresponding to this device is a normal waveform. Even if the data from the downstream device is correct, the arc may still affect this device, becoming a fault location. To further investigate these devices, several potential fault points can be found by using the magnitude of waveform parameters as a filtering condition. For example, from normal real-time current waveforms, find current waveforms that meet the waveform parameters (their frequency is greater than the frequency value of the waveform parameter, and their amplitude is greater than the amplitude value of the waveform parameter), and use the location of the device with this current waveform as a potential fault point. Through the above search process, several potential fault points can be found. Then, the coordinates of the potential fault points can be connected together to construct a fault propagation path. When connecting, the coordinates of the potential fault points can be connected together according to the direction of signal transmission or the direction of current transmission to obtain the fault propagation path.

[0062] Next, pressure wave superposition feature values ​​corresponding to potential fault points can be extracted from the fault propagation path. The real-time pressure can be determined by the pressure wave superposition feature values ​​to determine whether there is an electric arc affecting the generation of pressure waves at that location, so as to facilitate subsequent accident analysis.

[0063] In one specific operating method, real-time pressure wave data can be acquired from the pressure sensors of devices corresponding to potential fault points along the fault propagation path. The pressure wave data is a continuous sequence of sampled values ​​showing pressure changes over time. Time-domain analysis is performed on the acquired pressure wave data to convert it into a waveform, obtaining the real-time pressure wave waveform. The maximum value points in the real-time pressure wave waveform are identified as peak positions, and the minimum value points are identified as trough positions. For the identified peak and trough positions, the arrival time difference of the same pressure wave peak detected by different sensors is calculated. Based on the time difference, the degree of peak superposition is determined, and the trough time offset is calculated to obtain the trough offset value. The energy distribution value is obtained by summing the squares of the pressure wave amplitudes within a preset time window.

[0064] It should be noted that the pressure wave data is acquired at a frequency of tens of thousands of times per second, capturing pressure changes at the microsecond level. These continuous sampled values ​​reflect the instantaneous fluctuations in air pressure. When an electric arc discharges, a sharp pressure rise occurs, forming a shock wave that propagates outwards. Time-domain analysis identifies extreme points by comparing the magnitudes of adjacent sampled values: when the pressure value at a point is greater than that at the two preceding and following sampled points, that point is a peak; conversely, it is a trough. The calculation of the peak arrival time difference reveals the propagation characteristics of the pressure wave.

[0065] For example, sensor A detects a wave peak at time t1, and sensor B detects a wave peak with the same characteristics at time t2. The time difference t2-t1 reflects the time required for the pressure wave to propagate from near A to near B. When multiple sensors detect a pressure wave simultaneously, wave peaks at certain locations may overlap, forming a higher pressure peak. The degree of overlap is quantified by comparing the ratio of the theoretical single-wave peak value to the actual measured peak value.

[0066] In one embodiment, the energy distribution value can be expressed as follows: ; In the above formula, E d The value represents the energy distribution, M represents the number of sampling points within the preset time window, and P represents the energy distribution value. k Δt represents the pressure wave amplitude at the k-th sampling point, and Δt represents the sampling time interval. This formula calculates the energy distribution characteristic value within a time window by summing the squares of all pressure wave amplitudes within that time window. The combined pressure value is formed by combining the peak overlap value, trough offset value, and energy distribution value.

[0067] As explained above, the potential fault points to be screened could be normal equipment, abnormal equipment, or equipment affected by abnormal equipment. Therefore, different equipment has different pressure combination values. Pressure combination values ​​greater than a preset pressure value can be selected from multiple pressure combination values ​​as pressure wave superposition characteristic values.

[0068] S14. Determine the location coordinates of each pressure wave superposition feature value, and obtain the temperature time series parameters and real-time image of each location coordinate. Determine the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters. When any fire level value is greater than the preset fault level value, trigger the corresponding safety warning processing based on the fire level value.

[0069] In one embodiment, the device corresponding to each superimposed characteristic value of the pressure wave can be identified, and the location coordinates of the device can be obtained. Temperature time-series parameters are then collected using the temperature sensor of the device corresponding to each location coordinate, and real-time images are acquired using the camera of the device corresponding to each location coordinate. Based on the real-time images, it can be determined whether there is a fire source, and based on the temperature time-series parameters, whether a fire has occurred, thereby determining the fire level value corresponding to each location coordinate.

[0070] Because there are several identifiable superimposed characteristic values ​​of the pressure waves, there are also several corresponding calculated fire severity values. When any fire severity value exceeds the preset fault severity value, it indicates a fire risk and can trigger a safety warning.

[0071] Alternatively, the location of the pressure shock source can be determined based on the energy distribution value. This location is the location of the equipment that generates pressure due to a fault or electric arc. The location of the pressure shock source can be determined, and then temperature timing parameters and real-time images can be obtained through the temperature sensor and camera of the equipment at the location of the pressure shock source.

[0072] Specifically, the locations of the three sensors with the highest energy can be identified based on the energy distribution values ​​in the superimposed characteristic values ​​of the pressure wave. Using the arrival time difference of the wave crests detected by these three sensors and the known spatial coordinates of the sensors, the location of the impact source can be determined by calculating the relationship between pressure wave propagation time and distance. The pressure arrival time difference can be expressed as follows: ; In the above formula, Δt ij This represents the time difference t between the arrival of the wave peak detected by sensor i and sensor j. j t represents the arrival time of the wave peak at sensor j. i d represents the peak arrival time of sensor i. sj d represents the distance from the impact source to sensor j. si denoted by , where represents the distance from the impact source to sensor i, and v represents the propagation speed of the pressure wave.

[0073] The location of the pressure impact source can be represented by the following formula: ; In the above formula, x s y s and z s The x represents the spatial coordinates of the impact source. i y i and z i Let v represent the known spatial coordinates of the i-th sensor, v represent the propagation speed of the pressure wave, and t represent the propagation speed of the pressure wave. i d represents the time it takes for the pressure wave to reach the i-th sensor, and d0 represents the reference distance constant.

[0074] Alternatively, the location of the pressure shock source can be determined using the above method, the coordinates of the pressure shock source location can be obtained, the location coordinates can be obtained, and then the temperature time series parameters and real-time images of the coordinates of each pressure shock source location can be obtained. Based on the real-time images and the temperature time series parameters, the fire level value corresponding to each location coordinate can be determined.

[0075] Furthermore, calculating the trough offset is equally important. Normally, the trough of a pressure wave should appear at a fixed time interval after the crest, but due to reflection and interference effects, the trough position will shift. This offset is obtained by calculating the difference between the actual trough time and the theoretical trough time, reflecting the degree of interference encountered by the pressure wave during propagation. The energy distribution value is calculated using the energy integration method. Within a preset time window, a period containing the complete pressure waveform is selected, and the pressure value at each sampling point is squared and summed. The square of the pressure represents the energy density at that moment, and the summation reflects the total energy received at that sensor location. The spatial difference in energy distribution directly indicates the approximate direction of the impact source.

[0076] The determination of the impact location is based on the principle of acoustic localization. After identifying the three sensors with the highest energy, a set of localization equations is constructed using their spatial coordinates and the time difference of arrival of the wave crests. Assuming that the distances from the impact source to the three sensors are d1, d2, and d3, respectively, and the pressure wave propagation speed is v, the distance difference can be calculated using the time difference.

[0077] When the calculated fire severity value is greater than the preset fault severity value, a safety warning can be triggered. Different fire severity values ​​can correspond to different alarm information, and different safety warnings can be executed according to the fire severity value.

[0078] For example, if the fire level is 1, the safety warning process can be a warning broadcast, which can be issued to remind construction workers to pay attention.

[0079] The fire level is 2. The safety warning procedure can be to broadcast an emergency alarm, issue a work stoppage warning, prompt back-end personnel to suspend work, and evacuate technicians from potential fault locations so that maintenance personnel can take emergency measures.

[0080] If the fire level is 3, notify technical personnel to activate the emergency response, cut off the power to the equipment, and arrange evacuation.

[0081] In one embodiment, the location has a large number of objects piled up, which could easily cause a fire if a malfunction or electric arc occurs. To perform fire level analysis based on the number of objects, in one embodiment, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters may include the following sub-steps: S21. Call the preset first image recognition model to identify the number of objects in the real-time image and the distance between each object and the device, and obtain the object number value and the distance value respectively.

[0082] S22. Calculate the quantity weight value based on the number of objects and the interval distance value.

[0083] S23. Calculate the temperature change value of the temperature time series parameter, and use the temperature change value and the quantity weight value to determine the fire level value.

[0084] In one embodiment, a preset first image recognition model (e.g., a pre-trained CNN image recognition model) can be invoked to identify objects in a real-time image and count the number of objects. The identified objects may include the device and objects piled around the device. Then, the distance between the objects and the device can be calculated to obtain the object count and distance value, respectively.

[0085] In practice, a first image recognition model (such as a pre-trained CNN image recognition model) can be preset to identify objects in the real-time image and count the number of objects to obtain the object count value. Next, the image distance between each object and the device is identified, and then the image distance is converted into the actual distance according to the conventional image conversion method to obtain the interval distance value.

[0086] The greater the distance between an object and equipment, the lower the probability of it being directly ignited due to a malfunction; conversely, the closer the distance, the higher the probability of it being directly ignited. A quantity weight value can be calculated based on the number of objects and the distance between them.

[0087] In one embodiment, the quantity weight value can be calculated as follows: Q = (D1 + D2 + ... + Di) / i; Where Di is the reciprocal of the distance between each object and the device, and i is the number of objects.

[0088] Next, the maximum and minimum values ​​can be obtained from the temperature time series parameters. The difference between the maximum and minimum values ​​is calculated to obtain the temperature change value. Finally, the product of the temperature change value and the quantity weight value is calculated to obtain the fire probability value. The fire level value is determined based on the magnitude of the fire probability value.

[0089] Optionally, multiple preset value ranges can be set in advance. For example, 0-10 is the first range, 10-20 is the second, 20-30 is the third, and so on. Each range then corresponds to a fire rating value. For example, the fire rating value for the 0-10 range is 1, and the fire rating value for the 10-20 range is 2. Finally, the fire rating value can be determined based on the magnitude of the fire probability value.

[0090] In one embodiment, the object piled at this location is made of flammable material, and contains a large amount of flammable and explosive material, resulting in a high probability of fire. To perform fire level analysis based on the material properties, in another embodiment, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time-series parameters may include the following sub-steps: S31. Call the preset second image recognition model to perform material recognition on the objects in each of the real-time images to obtain several material information.

[0091] S32. Based on each of the material information, find the corresponding ignition temperature value for each material, and call the K-means algorithm to cluster the multiple ignition temperature values ​​to obtain several ignition categories.

[0092] S33. Count the number of ignition temperature values ​​for each ignition category to obtain the temperature quantity value, and calculate the temperature peak value of the temperature time sequence parameter.

[0093] S34. The fire level value is determined using the temperature quantity value and the temperature peak value.

[0094] The preset second image recognition model (such as a pre-trained CNN image recognition model) is invoked to perform material recognition on the objects in each real-time image, and several material information is obtained.

[0095] It should be noted that the first preset image recognition model is used to identify objects, the number of objects, and the distance between objects, while the second preset image recognition model is used to identify material information. Although both can be CNN model frameworks, they can be trained using different data.

[0096] Next, based on the information of each material, the corresponding ignition temperature value of each material can be found. The ignition temperature value is the combustible temperature value of the material.

[0097] Because there are many objects, there may be multiple ignition temperature values ​​to be searched, and these values ​​may vary widely; some materials may have an ignition temperature of 150 degrees Celsius, while others may have an ignition temperature of 500 degrees Celsius. To improve processing efficiency, the K-means algorithm can be used to cluster the multiple ignition temperature values, resulting in several ignition temperature categories. Each category corresponds to a temperature range. For example, using 100 degrees Celsius as the category range, after clustering, there are three ignition temperature categories: the first category has a temperature range of 0-100 degrees Celsius, the second has a range of 100-200 degrees Celsius, and the third has a range of 200-300 degrees Celsius.

[0098] Next, the number of ignition temperature values ​​within each ignition category can be counted to obtain the temperature quantity value, and the temperature peak value of the temperature time sequence parameter can be calculated.

[0099] Finally, the fire probability value is calculated using the temperature quantity value and the temperature peak value, and then the fire level value is determined based on the fire probability value.

[0100] In this embodiment, the fire probability value can be calculated as follows: K=T*(e A1 +eA2 +…+e Aj ).

[0101] Where K is the fire probability value, T is the peak temperature, e is the calculation constant (specifically an integer greater than 1, optionally a value between 1 and 2), Aj is the temperature quantity value, and j is the number of ignition point categories.

[0102] Alternatively, referring to the above embodiments, multiple preset interval values ​​can be pre-defined. For example, 0-10 is the first interval value, 10-20 is the second interval value, 20-30 is the third interval value, and so on. Each interval value then corresponds to a fire level value. For example, the fire level value for the 0-10 interval is 1, and the fire level value for the 10-20 interval is 2. Finally, the fire level value can be determined based on the magnitude of the fire probability value.

[0103] In one embodiment, a malfunction in a high-voltage electrical device may momentarily trigger an electric arc, which could then cause a fire. To address the fire promptly, as an example, after the step of acquiring the temperature time-series parameters and real-time images for each of the stated location coordinates, the method further includes the following sub-steps: S41. If the temperature time series parameter meets the threshold or the preset BP model is called to identify that the real-time image contains a heat source, the coordinates of the heat source are determined and an isolation boundary region is constructed with the coordinates of the heat source as the center.

[0104] S42. Shut down the equipment within the isolation boundary area and activate the emergency alarm process.

[0105] In one embodiment, when the maximum value of several temperature time-series parameters exceeds a preset high-temperature threshold or when a preset BP model is invoked to identify a heat source in the real-time image, it is determined that a fire has occurred. The coordinates of the heat source can be determined, and the coordinates of the camera can be obtained. Finally, an isolation boundary region is constructed centered on the coordinates of the heat source.

[0106] In one practical operation, the fire intensity level can be assessed simultaneously, and the radius can be determined based on the fire intensity level. For example, a fire intensity level of 1 corresponds to a radius of 50 meters, and a fire intensity level of 2 corresponds to a radius of 100 meters. An isolation boundary zone is constructed with the heat source coordinates as the center and the aforementioned distances as the radius. Finally, the power supply to the equipment within the isolation boundary zone can be cut off to shut down the equipment. Simultaneously, an emergency alarm can be activated to notify back-end personnel and safety vehicles within the isolation boundary zone, and maintenance personnel can be notified to take emergency measures.

[0107] In this embodiment, the present invention provides a fire protection and early warning method for a converter station. Its advantages are as follows: the present invention can acquire real-time current data and perform Fourier transform on the real-time current data to obtain a real-time current waveform; extract abnormal current waveforms from the real-time current waveform; determine the waveform parameters of the abnormal current waveform; and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters; when the probability level value is greater than a preset probability level value, find several potential fault points based on the waveform parameter magnitude, and obtain pressure wave superposition feature values ​​from these potential fault points; determine the location coordinates of each pressure wave superposition feature value, and obtain the temperature time series parameters and real-time image for each location coordinate; determine the fire level value corresponding to each location coordinate based on the real-time image and temperature time series parameters; and trigger a safety early warning process when any fire level value is greater than a preset fault level value. This invention monitors equipment in a converter station simultaneously using current data, pressure data, temperature, and real-time images. It can accurately determine the fire level and trigger alarms, which not only improves the accuracy of safety accident assessment but also allows different fire levels to trigger different alarms, thus reminding technicians to carry out maintenance and repairs earlier. This improves the timeliness of early warnings, shortens alarm delays, and enables technicians to perform maintenance as soon as possible, further reducing equipment risks and ensuring the safety of technicians.

[0108] This invention also provides a fire early warning system for converter stations, see [link to relevant documentation]. Figure 2 The diagram shows a structural schematic of a fire early warning system for a converter station according to an embodiment of the present invention.

[0109] As an example, the fire early warning system of the converter station may include: The acquisition module 201 is used to acquire real-time current data and perform Fourier transform on the real-time current data to obtain a real-time current waveform, and extract abnormal current waveforms from the real-time current waveform, wherein the real-time current data is the real-time data of the equipment in the valve hall of the converter station. The determination module 202 is used to determine the waveform parameters of the abnormal current waveform and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters, wherein the probability level value is the fault level corresponding to the probability of an electric arc caused by equipment overload. Extraction module 203 is used to find several potential fault points according to the waveform parameter size when the probability level value is greater than the preset probability level value, and obtain the pressure wave superposition feature value from the several potential fault points. The early warning module 204 is used to determine the location coordinates of each pressure wave superposition feature value, and to acquire the temperature time series parameters and real-time image of each location coordinate. Based on the real-time image and the temperature time series parameters, the fire level value corresponding to each location coordinate is determined. When any fire level value is greater than a preset fault level value, the corresponding safety early warning processing is triggered based on the fire level value.

[0110] Optionally, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time-series parameters includes: The preset first image recognition model is invoked to identify the number of objects in the real-time image and the distance between each object and the device, and the object number value and the distance value are obtained respectively. Calculate the quantity weight value based on the number of objects and the interval distance value; Calculate the temperature change value of the temperature time series parameter, and use the temperature change value and the quantity weight value to determine the fire level value.

[0111] Optionally, determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time-series parameters includes: The preset second image recognition model is invoked to identify the objects in each real-time image, thereby obtaining several material information. Based on the information of each material, the corresponding ignition temperature value of each material is found, and the K-means algorithm is called to cluster the multiple ignition temperature values ​​to obtain several ignition categories; The number of ignition temperature values ​​for each ignition category is counted to obtain a temperature quantity value, and the temperature peak value of the temperature time sequence parameter is calculated. The fire level is determined using the temperature quantity and the temperature peak value.

[0112] Optionally, the waveform parameters include: frequency value and amplitude value; The step of using a preset support vector machine to determine the corresponding probability level value based on the waveform parameters includes: The amplitude change rate is calculated using the amplitude value, and the frequency value, the amplitude value, and the amplitude change rate are converted into a vector to obtain a parameter feature vector. A preset support vector machine is invoked to perform classification processing based on the parameter feature vectors to obtain the arc probability value; Based on the arc probability value, a target interval value is determined from several preset interval values, and the probability level value corresponding to the target interval value is determined.

[0113] Optionally, the step of finding several potential fault points based on the waveform parameter magnitude and obtaining pressure wave superposition feature values ​​from the several potential fault points includes: By determining the location coordinates of the device corresponding to each abnormal current waveform, several potential fault points are obtained; Based on the pressure sensor at the potential fault point, real-time pressure wave waveforms are obtained, and the values ​​of peak superposition, valley offset, and energy distribution are extracted from each real-time pressure wave waveform. The pressure combination value is obtained by combining the peak superposition degree value, the trough offset value, and the energy distribution value. Several pressure wave superposition feature values ​​are then selected from the multiple pressure combination values. The pressure wave superposition feature values ​​are pressure combination values ​​with values ​​greater than a preset pressure value.

[0114] Optionally, the step of acquiring real-time current data and performing a Fourier transform operation on the real-time current data to obtain a real-time current waveform, and extracting an abnormal current waveform from the real-time current waveform, includes: A number of real-time current data points are acquired and each of the real-time current data points is preprocessed to obtain a number of processed data points. Each of the processed data points is then converted into a real-time current waveform using a fast Fourier transform. The preprocessing includes noise reduction and time-domain transformation. Calculate the total harmonic distortion rate of each of the real-time current waveforms, and determine the abnormal current waveform from the real-time current waveforms based on the total harmonic distortion rate.

[0115] Optionally, the system further includes: A region construction module is used to determine the heat source coordinates and construct an isolation boundary region centered on the heat source coordinates after the step of obtaining the temperature time series parameters and real-time images of each location coordinate. If the temperature time series parameters meet the threshold or the preset BP model is called to identify that the real-time image contains a heat source, the heat source coordinates are determined and an isolation boundary region is constructed with the heat source coordinates as the center. The shutdown module is used to shut down devices within the isolation boundary area and initiate emergency alarm processing.

[0116] Those skilled in the art will understand that, for ease of description and brevity, the specific working process of the device described above can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0117] Furthermore, this application also provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the fire protection and early warning method for converter stations as described in the above embodiments.

[0118] Furthermore, embodiments of this application also provide a computer-readable storage medium storing a computer-executable program, the computer-executable program being used to cause a computer to execute the fire protection and early warning method for converter stations as described in the above embodiments.

[0119] In the description of the embodiments of the present invention, it should be noted that the terms "above," "below," etc., indicating the orientation or positional relationship are based on the orientation or positional relationship shown in the accompanying drawings, and are only for the convenience of describing the embodiments of the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the present invention. When an element such as a layer, region, or substrate is referred to as being "above" or "on top of" another element, it may be directly on the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly on" or "above" another element, there is no intermediate element. It should also be understood that when an element is referred to as being "below" or "under" another element, it may be directly below or under the other element, or there may be an intermediate element. Conversely, when an element is referred to as being "directly below" or "under" another element, there is no intermediate element. Unless otherwise explicitly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to fixed connections, detachable connections, or integral connections; they can refer to mechanical connections or electrical connections; they can refer to direct connections or indirect connections through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0120] Those skilled in the art will understand that embodiments of this application may also include computer program products. Therefore, this application may take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0121] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), devices, and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0122] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0123] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0124] The above description is only a preferred embodiment of the present invention. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the technical principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A fire prevention and early warning method for a converter station, characterized in that, The method includes: Real-time current data is acquired and Fourier transform is performed on the real-time current data to obtain a real-time current waveform. Abnormal current waveforms are extracted from the real-time current waveform. The real-time current data is the real-time data of the equipment in the valve hall of the converter station. The waveform parameters of the abnormal current waveform are determined, and a preset support vector machine is used to determine the corresponding probability level value based on the waveform parameters. The probability level value is the fault level corresponding to the probability of an electric arc caused by equipment overload. When the probability level value is greater than the preset probability level value, several potential fault points are found according to the waveform parameter size, and the pressure wave superposition feature value from the several potential fault points is obtained. The location coordinates of each pressure wave superposition feature value are determined, and the temperature time series parameters and real-time images of each location coordinate are obtained. The fire level value corresponding to each location coordinate is determined based on the real-time images and the temperature time series parameters. When any fire level value is greater than a preset fault level value, the corresponding safety warning processing is triggered based on the fire level value.

2. The fire protection and early warning method for converter stations according to claim 1, characterized in that, The step of determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters includes: The preset first image recognition model is invoked to identify the number of objects in the real-time image and the distance between each object and the device, and the object number value and the distance value are obtained respectively. Calculate the quantity weight value based on the number of objects and the interval distance value; Calculate the temperature change value of the temperature time series parameter, and use the temperature change value and the quantity weight value to determine the fire level value.

3. The fire protection and early warning method for converter stations according to claim 1, characterized in that, The step of determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters includes: The preset second image recognition model is invoked to identify the objects in each real-time image, thereby obtaining several material information. Based on the information of each material, the corresponding ignition temperature value of each material is found, and the K-means algorithm is called to cluster the multiple ignition temperature values ​​to obtain several ignition categories; The number of ignition temperature values ​​for each ignition category is counted to obtain a temperature quantity value, and the temperature peak value of the temperature time sequence parameter is calculated. The fire level is determined using the temperature quantity and the temperature peak value.

4. The fire protection and early warning method for converter stations according to claim 1, characterized in that, The waveform parameters include: frequency value and amplitude value; The step of using a preset support vector machine to determine the corresponding probability level value based on the waveform parameters includes: The amplitude change rate is calculated using the amplitude value, and the frequency value, the amplitude value, and the amplitude change rate are converted into a vector to obtain a parameter feature vector. A preset support vector machine is invoked to perform classification processing based on the parameter feature vectors to obtain the arc probability value; Based on the arc probability value, a target interval value is determined from several preset interval values, and the probability level value corresponding to the target interval value is determined.

5. The fire protection and early warning method for converter stations according to claim 1, characterized in that, The step of finding several potential fault points based on the waveform parameter magnitude and obtaining the pressure wave superposition feature value from the several potential fault points includes: By determining the location coordinates of the device corresponding to each abnormal current waveform, several potential fault points are obtained; Based on the pressure sensor at the potential fault point, real-time pressure wave waveforms are obtained, and the values ​​of peak superposition, valley offset, and energy distribution are extracted from each real-time pressure wave waveform. The pressure combination value is obtained by combining the peak superposition degree value, the trough offset value, and the energy distribution value. Several pressure wave superposition feature values ​​are then selected from the multiple pressure combination values. The pressure wave superposition feature values ​​are pressure combination values ​​with values ​​greater than a preset pressure value.

6. The fire protection and early warning method for converter stations according to claim 1, characterized in that, The process of acquiring real-time current data and performing a Fourier transform operation on the real-time current data to obtain a real-time current waveform, and extracting abnormal current waveforms from the real-time current waveform, includes: A number of real-time current data points are acquired and each of the real-time current data points is preprocessed to obtain a number of processed data points. Each of the processed data points is then converted into a real-time current waveform using a fast Fourier transform. The preprocessing includes noise reduction and time-domain transformation. Calculate the total harmonic distortion rate of each of the real-time current waveforms, and determine the abnormal current waveform from the real-time current waveforms based on the total harmonic distortion rate.

7. The fire protection and early warning method for converter stations according to any one of claims 1-6, characterized in that, After the step of acquiring the temperature time-series parameters and real-time images for each of the location coordinates, the method further includes: If the temperature time series parameter meets the threshold or the preset BP model is called to identify that the real-time image contains a heat source, the coordinates of the heat source are determined and an isolation boundary region is constructed with the coordinates of the heat source as the center. Shut down the equipment within the isolation boundary area and activate the emergency alarm process.

8. A fire early warning system for a converter station, characterized in that, The system includes: The acquisition module is used to acquire real-time current data and perform Fourier transform on the real-time current data to obtain a real-time current waveform, and extract abnormal current waveforms from the real-time current waveform. The real-time current data is the real-time data of the equipment in the valve hall of the converter station. The determination module is used to determine the waveform parameters of the abnormal current waveform and use a preset support vector machine to determine the corresponding probability level value based on the waveform parameters, wherein the probability level value is the fault level corresponding to the probability of an electric arc caused by equipment overload. The extraction module is used to find several potential fault points based on the waveform parameter magnitude when the probability level value is greater than the preset probability level value, and to obtain the pressure wave superposition feature value from the several potential fault points. The early warning module is used to determine the location coordinates of each pressure wave superposition feature value, and to acquire the temperature time series parameters and real-time image of each location coordinate. Based on the real-time image and the temperature time series parameters, the module determines the fire level value corresponding to each location coordinate. When any fire level value is greater than a preset fault level value, the module triggers the corresponding safety early warning process based on the fire level value.

9. The fire early warning system for converter stations according to claim 8, characterized in that, The step of determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters includes: The preset first image recognition model is invoked to identify the number of objects in the real-time image and the distance between each object and the device, and the object number value and the distance value are obtained respectively. Calculate the quantity weight value based on the number of objects and the interval distance value; Calculate the temperature change value of the temperature time series parameter, and use the temperature change value and the quantity weight value to determine the fire level value.

10. The fire early warning system for converter stations according to claim 8, characterized in that, The step of determining the fire level value corresponding to each location coordinate based on the real-time image and the temperature time series parameters includes: The preset second image recognition model is invoked to identify the objects in each real-time image, thereby obtaining several material information. Based on the information of each material, the corresponding ignition temperature value of each material is found, and the K-means algorithm is called to cluster the multiple ignition temperature values ​​to obtain several ignition categories; The number of ignition temperature values ​​for each ignition category is counted to obtain a temperature quantity value, and the temperature peak value of the temperature time sequence parameter is calculated. The fire level is determined using the temperature quantity and the temperature peak value.

Images