Remote control method and system for fire pump of photovoltaic power station

Abstract

The invention provides a remote control method and system for a fire pump of a photovoltaic power station. Temperature information, smoke information and flame information are detected through a sensor array; the temperature information is converted into a temperature sequence, the smoke information is converted into a smoke concentration gradient, and the flame information is converted into arc pulse frequency; inputting the temperature sequence, the smoke concentration gradient and the arc pulse frequency to a time sequence analysis model, and superposing the component attenuation rate to determine a hot spot risk index; when the hot spot risk index is larger than a first threshold value, a first-stage early warning pre-pressurization water pump system is started; when the hot spot risk index is greater than a second threshold value, triggering a secondary response, and linking the direct current circuit breaker to break and start the water pump; when the hot spot risk index is equal to a third threshold value, three-level response is activated, full-field spraying and power grid off-network are executed, through remote sensor detection and fire risk identification, control over a fire pump is remotely achieved, fire extinguishing is conducted on a fire in time and efficiently, and the remote fire extinguishing efficiency is effectively improved.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic power station fire protection, and in particular to a remote control method and system for a photovoltaic power station fire pump. Background Art

[0002] Photovoltaic power station fires are highly prevalent. According to incomplete statistics, 35% of PV power station fires are caused by DC arcs, and 25% by the hot spot effect. Traditional temperature-based smoke detectors have a response lag of >2 minutes. Current firefighting methods have the following drawbacks: general fire pump controls are not adapted to the high-voltage characteristics of photovoltaic DC, and spraying can cause electric shocks; fixed-threshold warning models have a false alarm rate of >40% (due to interference from solar reflection and welding operations); and they lack precise module-level positioning capabilities, making it impossible to disconnect faulty strings.

[0003] Therefore, how to improve the remote control of fire water pumps in photovoltaic power plants has become a technical problem that technical personnel in this field urgently need to solve. Summary of the Invention

[0004] The present invention provides a remote control method and system for a fire water pump of a photovoltaic power station, which are used to solve the defect of low fire fighting efficiency of photovoltaic power stations in the prior art.

[0005] In a first aspect, the present invention provides a method for remotely controlling a fire pump in a photovoltaic power station, comprising: Deploy sensor arrays on photovoltaic module back panels, combiner boxes, and cable trenches, and detect temperature information, smoke information, and flame information through the sensor arrays; Performing protocol conversion through an embedded communication gateway to convert the temperature information into a temperature sequence, the smoke information into a smoke concentration gradient, and the flame information into an arc pulse frequency; Inputting the temperature sequence, the smoke concentration gradient, and the arc pulse frequency into a timing analysis model, and superimposing component attenuation rates to determine a hot spot risk index; When the hot spot risk index is greater than a first threshold, starting a first-level warning pre-pressurization water pump system; When the hot spot risk index is greater than a second threshold, a secondary response is triggered, the DC circuit breaker is disconnected and the water pump is started; When the hot spot risk index is equal to the third threshold, a third-level response is activated, performing full-field spraying and grid disconnection.

[0006] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the method includes performing protocol conversion through an embedded communication gateway, converting the temperature information into a temperature sequence, converting the smoke information into a smoke concentration gradient, and converting the flame information into an arc pulse frequency, including: Perform linear normalization on the temperature information to generate a temperature sequence arranged by timestamp; Discrete sampling and interpolation calculation are performed on smoke information to generate a continuously changing smoke concentration gradient; The light intensity fluctuation characteristics of the flame information are subjected to Fourier transform, and the fundamental frequency component is extracted as the arc pulse frequency.

[0007] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the temperature sequence, the smoke concentration gradient, and the arc pulse frequency are input into a timing analysis model, and the component attenuation rate is superimposed to determine the hot spot risk index, including: extracting multi-scale features of the temperature sequence, the smoke concentration gradient, and the arc pulse frequency; The multi-scale features are identified through a CNN-BiLSTM-attention fusion model, and a hot spot score is output; Identify the multi-scale features by constructing an attenuation model and output a component attenuation rate; The hot spot score and the component attenuation rate are nonlinearly superimposed to obtain a hot spot risk index.

[0008] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the multi-scale features are identified by a CNN-BiLSTM-attention fusion model, and a hot spot score is output, including: Slice the multi-scale features into time windows, extract statistical features, frequency domain features, and time series features, and construct a standardized three-dimensional input tensor; Extracting local spatiotemporal features from the standardized three-dimensional input tensor through a three-layer one-dimensional convolutional network to generate a fused feature map; A two-layer bidirectional LSTM network is used to capture the temporal dependency of the fused feature map. Each layer is followed by layer normalization and residual connection to output sequence features. Applying a dual attention mechanism of time dimension and feature dimension to the sequence features, focusing on key temporal segments and feature dimensions, and generating weighted aggregate features; The weighted aggregate features are generated into a probability distribution through a fully connected classification network to determine the hot spot score.

[0009] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the method of identifying the multi-scale features by establishing an attenuation model and outputting the component attenuation rate includes: Based on the physical mechanisms of light-induced degradation, potential-induced degradation and temperature fatigue, a theoretical degradation calculation model is established to determine the theoretical degradation amount; Input the multi-scale feature vector into the data-driven model, fit the attenuation trend in the complex environment through nonlinear mapping, and output the measured attenuation reflecting the real-time operation abnormality; The theoretical attenuation and the measured attenuation are weighted and fused to determine the component attenuation rate.

[0010] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the hot spot risk index is obtained by nonlinearly superimposing the hot spot score and the component attenuation rate, including: ; in, Characterizes the current abnormal signal strength, The increase in basic risk due to component aging, Characterize the synergistic amplification effect of abnormal signals × component aging; 、 and represents the weight, represents the hot spot score, Represents the component attenuation rate.

[0011] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, the method for starting a first-level warning pre-pressurized water pump system includes: Send pre-pressurization instructions to the water pump inverter, start the water pump auxiliary motor, adjust the opening of the electric control valve, and verify the pipe network pressure sensor reading; The three-dimensional coordinates and risk value of the warning area are displayed on the operation and maintenance terminal.

[0012] According to a method for remotely controlling a fire water pump in a photovoltaic power station provided by the present invention, triggering a secondary response, disconnecting a DC circuit breaker, and starting the water pump comprises: Cut off the faulty string DC circuit breaker; Start the corresponding zone water pump and verify the outlet pressure; Push the escape route heat map to the operation and maintenance terminal.

[0013] According to a method for remotely controlling a fire pump in a photovoltaic power station provided by the present invention, activating a three-level response, executing full-site spraying and disconnecting from the power grid, includes: Start the main spray pump. If the pipe network pressure does not meet the standard after the target spraying time, switch to the backup pump. According to the hot spot positioning system, open the solenoid valve group in the risk area and expand it to fully open the solenoid valves in the entire field to achieve a complete spray coverage. During the no-dead-angle spray coverage process, the DC circuit breaker of the combiner box, the AC circuit breaker of the inverter and the high-voltage side switch of the isolation transformer are disconnected in sequence, and the grounding switch is closed in conjunction to achieve electrical isolation between the photovoltaic system and the power grid.

[0014] In a second aspect, the present invention further provides a photovoltaic power station fire water pump remote control system, comprising: A detection module is used to deploy a sensor array on the photovoltaic module backplane, combiner box and cable trench, and detect temperature information, smoke information and flame information through the sensor array; a conversion module, configured to perform protocol conversion through an embedded communication gateway, convert the temperature information into a temperature sequence, convert the smoke information into a smoke concentration gradient, and convert the flame information into an arc pulse frequency; a determination module, configured to input the temperature sequence, the smoke concentration gradient, and the arc pulse frequency into a timing analysis model, and superimpose component attenuation rates to determine a hot spot risk index; The fire protection module is used to start the first-level warning pre-pressurized water pump system when the hot spot risk index is greater than the first threshold; when the hot spot risk index is greater than the second threshold, trigger the second-level response, disconnect the DC circuit breaker and start the water pump; when the hot spot risk index is equal to the third threshold, activate the third-level response, execute full-site spraying and grid disconnection.

[0015] In a third aspect, the present invention further provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the remote control method for a fire water pump in a photovoltaic power station as described above is implemented.

[0016] In a fourth aspect, the present invention further provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the above-described methods for remotely controlling a fire water pump in a photovoltaic power station.

[0017] In a fifth aspect, the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements any of the above-described methods for remotely controlling a fire water pump in a photovoltaic power station.

[0018] The present invention provides a remote control method and system for fire pumps in a photovoltaic power station. The method and system deploy sensor arrays on the back panel of photovoltaic modules, junction boxes and cable trenches, and detect temperature information, smoke information and flame information through the sensor arrays. Protocol conversion is performed through an embedded communication gateway to convert temperature information into a temperature sequence, smoke information into a smoke concentration gradient, and flame information into an arc pulse frequency. The temperature sequence, smoke concentration gradient and arc pulse frequency are input into a timing analysis model, and the component attenuation rate is superimposed to determine the hot spot risk index. When the hot spot risk index is greater than a first threshold, a first-level warning pre-pressurized water pump system is started. When the hot spot risk index is greater than a second threshold, a second-level response is triggered, and the DC circuit breaker is disconnected and the water pump is started. When the hot spot risk index is equal to a third threshold, a third-level response is activated, and full-field sprinklers and grid disconnection are executed. Through remote sensor detection and fire risk identification, the fire pump is remotely controlled, and the fire is extinguished in a timely and efficient manner, effectively improving the efficiency of remote firefighting. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the technical solutions in the present invention or the prior art, a brief introduction is given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0020] Figure 1 This is a flow chart of a method for remotely controlling a fire pump in a photovoltaic power station provided in this embodiment; Figure 2 This is a schematic diagram of the structure of the remote control device for the fire pump of a photovoltaic power station provided in this embodiment; Figure 3 Schematic diagram of the structure of the electronic device provided in this embodiment. DETAILED DESCRIPTION

[0021] To make the objectives, technical solutions, and advantages of the present invention more clear, the technical solutions of the present invention will be clearly and completely described below in conjunction with the accompanying drawings. Obviously, the embodiments described are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0022] Figure 1 1 is a flow chart of the remote control method for a fire water pump in a photovoltaic power station provided in this embodiment.

[0023] like Figure 1As shown, the photovoltaic power station fire water pump remote control method provided by the embodiment of the present invention mainly includes the following steps: 101. Deploy sensor arrays on the back panels of photovoltaic modules, junction boxes and cable trenches, and use the sensor arrays to detect temperature information, smoke information and flame information.

[0024] Specifically, the first step is sensor deployment. For the photovoltaic module backsheet, the sensor uses a high-temperature-resistant (-40°C to +85°C) and high-precision (±0.5°C) infrared temperature sensor (such as a thermocouple or MEMS sensor) designed to withstand the long-term high-temperature outdoor environment of the photovoltaic backsheet. The sensor probe is tightly attached to the center of the backsheet or to a concentrated heat source (such as the location of the battery string solder joint) using thermally conductive silicone or metal clamps to ensure undiminished temperature signal acquisition. One sensor is deployed for every 20-30 modules, forming a regional monitoring unit.

[0025] Combiner box: Install a digital temperature sensor (such as the DS18B20) inside the combiner box. Secure the probe to the busbar or circuit breaker inside the combiner box (a hotspot). Use a miniature ionization or photoelectric smoke sensor (such as the Honeywell HSSD) and install it on the top of the box (where smoke rises). Maintain a 5-10 cm distance between the sensor and electrical components to prevent electromagnetic interference. Secure the wiring connector with waterproof sealant to meet IP65 protection standards.

[0026] Cable trenches: Deploy a distributed fiber optic temperature measurement system (DTS). The temperature measurement fiber is laid parallel to the cable along the trench wall, with a minimum monitoring spacing of ≤1m, to achieve distributed, real-time temperature scanning along the entire line. Ceiling-mounted smoke detectors (such as linear optical beam smoke detectors) are installed every 5-8m along the trench. The detector beam axis is 1.5-2m above the trench bottom, covering the space above the cable layer. Ultraviolet flame sensors (such as UV-IR dual-spectrum sensors) are installed at ignition-prone areas such as cable joints and branch nodes, providing 360° horizontal coverage.

[0027] 102. Protocol conversion is performed through the embedded communication gateway to convert temperature information into a temperature sequence, smoke information into a smoke concentration gradient, and flame information into an arc pulse frequency.

[0028] Converting temperature information into a temperature sequence primarily involves linear normalization, generating a timestamp-ordered temperature sequence. The target normalization range is specified (typically [0, 1], but adjustable as needed). The extreme values of the temperature data are then calculated. For static scenarios (such as batch processing of historical data), all temperature values are traversed to obtain the global minimum and maximum values. For real-time processing, a sliding window (e.g., the last hour's data) can be used to dynamically calculate the minimum and maximum values within the current window. This allows the extreme values to be updated over time to adapt to real-time data changes.

[0029] For each raw temperature value, a normalized value is calculated using a linear normalization formula. To avoid division by zero errors, the normalized value can be uniformly set to 0.5 or retained as is. All records with normalized temperature values are sorted in ascending timestamp order to ensure that the temperature sequence strictly follows the chronological order of data acquisition. The resulting sequence consists of each element containing an ordered "timestamp-normalized temperature value" pair, forming a one-dimensional ordered array representing the temperature sequence (e.g., a list arranged in acquisition order), providing time-aligned, standardized data for subsequent analysis.

[0030] Converting smoke information into a smoke concentration gradient involves discrete sampling and interpolation of the smoke information to generate a continuously varying smoke concentration gradient. Discrete sampling includes: Determining the sampling interval: Based on actual monitoring needs, discrete sampling of smoke information is performed at fixed time intervals (e.g., every 10 seconds) or spatial intervals (e.g., every 5 meters in a cable trench). For example, a smoke sensor located at the top of a combiner box records smoke concentration values every 10 seconds. Data collection: Smoke concentration data is collected using a smoke sensor (e.g., an ionization or photoelectric sensor) at a set sampling point and time, resulting in a discrete sequence of smoke concentration values, such as [C1, C2, C3...], with each value corresponding to a specific sampling time or location.

[0031] Interpolation: Select an interpolation method. Common interpolation methods include linear, polynomial, and spline. For smoke concentration data, linear interpolation is simple and efficient, suitable for quickly generating continuously changing gradients. Polynomial or spline interpolation can better fit complex trends. Calculate interpolation points. Between discrete sampling points, calculate the smoke concentration value at the intermediate point using the selected interpolation method. For example, using linear interpolation, if the concentrations C1 and C2 at two adjacent sampling points at times t1 and t2 are known, calculate the smoke concentration at any time t between t1 and t2 to fill in the gaps between the discrete data points.

[0032] Generate a smoke concentration gradient. Data integration: Integrate the original discrete sampled data with the interpolated data, arranging them in time or space to form a continuous smoke concentration data sequence. Calculate the gradient: For the integrated continuous data, calculate the ratio of the concentration difference between adjacent data points to the corresponding time or space interval to obtain the rate of change of smoke concentration at different locations or times, i.e., the smoke concentration gradient. This ultimately generates a continuously changing smoke concentration gradient for subsequent analysis and early warning.

[0033] Converting flame information to arc pulse frequency involves performing a Fourier transform on the intensity fluctuation characteristics of the flame information and extracting the fundamental frequency component as the arc pulse frequency. Flame intensity signal acquisition: Using an ultraviolet flame sensor or UV-IR dual-spectrum flame detector, the time-varying flame intensity fluctuation signal is collected at a high sampling rate (e.g., above 10kHz) to ensure complete capture of transient intensity changes and form a discrete time-domain signal sequence. Fourier transform calculation: A discrete Fourier transform (DFT) is performed on the collected time-domain intensity signal. In practical applications, the fast Fourier transform (FFT) algorithm is often used to improve computational efficiency. This converts the time-domain signal into a frequency-domain signal and determines the amplitude distribution corresponding to different frequency components. Fundamental frequency component identification: Within the frequency-domain signal, the frequency component with the largest amplitude is identified. This component is the fundamental frequency component, representing the primary periodic characteristic of the flame intensity fluctuation. Since arc pulses have a stable periodicity, their frequency corresponds to the fundamental frequency component. Frequency extraction and application: The frequency value of the fundamental frequency component is extracted as the arc pulse frequency, which is used to distinguish real flames from interfering light sources (such as sunlight, ordinary lights, etc.). When the detected frequency is within the characteristic frequency range of the arc pulse, a flame alarm signal is triggered.

[0034] 103. Input the temperature sequence, smoke concentration gradient and arc pulse frequency into the timing analysis model, and superimpose the component attenuation rate to determine the hot spot risk index.

[0035] Specifically, the process of determining the hot spot risk index is to extract the multi-scale features of temperature series, smoke concentration gradient and arc pulse frequency; identify the multi-scale features through the CNN-BiLSTM-attention fusion model and output the hot spot score; identify the multi-scale features through the component attenuation model and output the component attenuation rate; nonlinearly superimpose the hot spot score and the component attenuation rate to obtain the hot spot risk index.

[0036] The multi-scale feature extraction for temperature series, smoke concentration gradient, and arc pulse frequency involves: time-domain feature engineering, including statistical calculations such as mean, standard deviation, kurtosis, and skewness; dynamic features such as temperature change rate, smoke concentration gradient, and arc pulse interval; and threshold features such as the duration and amplitude of exceeding the safety threshold. Frequency-domain feature extraction involves wavelet transform decomposition into approximate and detail coefficients (a four-layer decomposition); power spectral density calculation for calculating energy distribution in different frequency bands (0.1-10Hz, 10-100Hz, and 100Hz-1kHz); and arc characteristic frequency analysis for extracting the characteristic energy ratio in the 50kHz-200kHz band. Time-series correlation analysis and cross-correlation analysis are then performed to investigate the time-lagged relationship between temperature and smoke concentration; conditional entropy calculation to quantify the information dependence between the three signal types; and Granger causality testing to identify leading indicators.

[0037] Among them, the CNN-BiLSTM-attention fusion model is used to identify multi-scale features and output hot spot scores, including: slicing multi-scale features according to time windows, extracting statistical features, frequency domain features and time series features, and constructing a standardized three-dimensional input tensor; extracting local spatiotemporal features of the standardized three-dimensional input tensor through a three-layer one-dimensional convolutional network to generate a fusion feature map; using a two-layer bidirectional LSTM network to capture the temporal dependency of the fusion feature map, and each layer is followed by layer normalization and residual connection to output sequence features; applying a dual attention mechanism of time dimension and feature dimension to the sequence features, focusing on key time series segments and feature dimensions, and generating weighted aggregate features; generating a probability distribution of the weighted aggregate features through a fully connected classification network to determine the hot spot score.

[0038] Specifically, the one-dimensional convolutional layer design is as follows: the first convolution layer has 16 kernels of size 3, stride 1, and ReLU activation function; the second convolution layer has 32 kernels of size 5, stride 1, and ReLU activation function; and the third convolution layer has 64 kernels of size 7, stride 1, and ReLU activation function. Each layer is followed by batch normalization and dropout (0.2) to prevent overfitting. Feature maps are then generated: the first layer outputs a feature map of size [number of samples × 28 × 16]; the second layer outputs a feature map of size [number of samples × 24 × 32]; and the third layer outputs a feature map of size [number of samples × 18 × 64]. A feature fusion mechanism is then implemented to concatenate the convolutional feature maps of the three types of time series data. A channel-wise attention mechanism is applied to adaptively adjust the weights of each channel.

[0039] To capture temporal dependencies, a bidirectional long short-term memory network (BiLSTM) with 128 hidden units in the first layer processes temporal information bidirectionally. A second BiLSTM layer with 64 hidden units further extracts high-level temporal features. Each layer is followed by layer normalization and residual connections. The final output is sequence features.

[0040] Focus on key time series segments. Calculate the importance score of each time step using time-dimensional attention, obtain the attention weight, and then perform weighted aggregation. Calculate the importance score of each feature dimension using feature-dimensional attention, obtain the feature weight, and perform weighted aggregation to obtain the final feature. The weighted aggregated feature of the attention weight is fused with the weighted aggregated feature of the feature weight to obtain the final weighted aggregated feature.

[0041] Through the fully connected classification network, the Softmax function is applied to convert the output into a probability distribution and output the hot spot score.

[0042] Among them, multi-scale features are identified by establishing an attenuation model to output the component attenuation rate, including: establishing a theoretical attenuation calculation model based on the physical mechanisms of light-induced attenuation, potential-induced attenuation and temperature fatigue to determine the theoretical attenuation amount; inputting multi-scale feature vectors into the data-driven model, fitting the attenuation trend in a complex environment through nonlinear mapping, and outputting the measured attenuation amount that reflects real-time operation abnormalities; performing weighted fusion of the theoretical attenuation amount and the measured attenuation amount to determine the component attenuation rate.

[0043] Specifically, based on the physical mechanisms of light-induced degradation (LID), potential-induced degradation (PID) and temperature fatigue, the impact of factors such as light, voltage, and temperature on the performance of photovoltaic modules is analyzed, and mathematical expressions under each degradation mechanism are derived. These expressions are integrated to construct a comprehensive theoretical degradation calculation model. By substituting data such as the initial parameters of the modules and environmental conditions, the theoretical attenuation is calculated.

[0044] Multi-scale data is collected during the operation of photovoltaic modules, such as microscopic material parameters, macroscopic environmental data (light intensity, temperature, humidity), and module electrical parameters (current, voltage). This data is then used to construct a multi-scale feature vector. This feature vector is then fed into a data-driven model (such as a neural network or support vector machine). Leveraging the model's nonlinear mapping capabilities, the model fits module degradation trends in complex environments and outputs measured attenuation values that reflect any real-time module operational anomalies.

[0045] Appropriate weights are assigned to the theoretical and measured attenuation values based on factors such as their reliability and stability. The theoretical and measured attenuation values are then combined through weighted summation to produce a module attenuation rate that accurately reflects the degree of attenuation of PV modules and is used to evaluate module performance and lifespan.

[0046] The hot spot risk index is obtained by nonlinearly superimposing the hot spot score and the component attenuation rate, as shown in formula (1): (1); in, Characterizes the current abnormal signal strength, The increase in basic risk due to component aging, Characterize the synergistic amplification effect of abnormal signals × component aging; 、 and represents the weight, represents the hot spot score, Represents the component attenuation rate.

[0047] 104. When the hot spot risk index is greater than the first threshold, the first-level warning pre-pressurization water pump system is activated.

[0048] When the hot spot risk index is greater than the first threshold of 6, a pre-pressurization command is sent to the water pump inverter to start the water pump auxiliary motor, adjust the opening of the electric control valve, and verify the pipe network pressure sensor reading; the three-dimensional coordinates and risk value of the warning area are displayed on the operation and maintenance terminal.

[0049] 105. When the hot spot risk index is greater than the second threshold, a secondary response is triggered, the DC circuit breaker is disconnected and the water pump is started.

[0050] When the hot spot risk index is greater than the second threshold of 8, the faulty string DC circuit breaker is disconnected; the corresponding zone water pump is started and the outlet pressure is verified; and the escape path thermal map is pushed to the operation and maintenance terminal.

[0051] 106. When the hot spot risk index is equal to the third threshold, the third level response is activated, and full-field spraying and grid disconnection are executed.

[0052] When the hot spot risk index is equal to the third threshold value of 10, the main spray pump is started. If the pipeline pressure does not meet the standard after the target spraying time, it is switched to the backup pump. The solenoid valve group in the risk area is opened according to the hot spot positioning system and expanded to fully open the solenoid valves in the entire field to form a no-dead-angle spray coverage. During the no-dead-angle spray coverage process, the DC circuit breaker of the junction box, the AC circuit breaker of the inverter and the high-voltage side switch of the isolation transformer are disconnected in sequence, and the grounding switch is closed in conjunction to achieve electrical isolation between the photovoltaic system and the power grid.

[0053] The technical solution of the present invention has the following advantages: By deploying sensor arrays on the backplanes of photovoltaic modules, junction boxes and cable trenches, three key fire-fighting parameters, temperature, smoke and flame, are collected in real time, covering the core parts of the photovoltaic system that are prone to heat and fire, thereby achieving early perception of fire hazards.

[0054] The embedded communication gateway is used to standardize the raw monitoring data (such as temperature series, smoke concentration gradient, and arc pulse frequency) to improve data availability and analysis accuracy, providing a reliable basis for subsequent risk assessment.

[0055] Multi-scale features are extracted through the CNN-BiLSTM-attention fusion model. Combined with the component attenuation rate (weighted fusion of theoretical attenuation and measured attenuation), a nonlinear superposition hot spot risk index is constructed to comprehensively reflect the collaborative risks of equipment aging and real-time anomalies, avoiding misjudgment of a single indicator.

[0056] Improve efficiency through a tiered response mechanism: Level 1 Warning: The pre-pressurized water pump system is prepared in advance to shorten emergency response time. Risk areas are located using three-dimensional coordinates, facilitating rapid investigation by operations and maintenance personnel. Level 2 Response: The DC circuit breaker disconnects the faulty string, isolating the fire source. Zoned water pumps are activated for precise firefighting, and escape routes are provided to ensure personnel safety. Level 3 Response: Full-site sprinkler control is synchronized with grid disconnection. Switching between primary and backup pumps and full-coverage spraying with solenoid valves ensures comprehensive firefighting. Electrical isolation prevents secondary hazards (such as electric shock and arc reignition).

[0057] Overall, through a closed-loop system of multi-dimensional monitoring, intelligent assessment, and hierarchical linkage, early warning and precise handling of fire risks in photovoltaic power stations are achieved, significantly improving the intelligence level and reliability of the fire protection system. At the same time, it takes into account equipment operation and maintenance and personnel safety, and has strong engineering application value and industry demonstration significance.

[0058] Based on the same general inventive concept, the present invention also protects a photovoltaic power station fire water pump remote control system. The photovoltaic power station fire water pump remote control system described below and the photovoltaic power station fire water pump remote control method described above can refer to each other.

[0059] Figure 2 This is a schematic diagram of the structure of the photovoltaic power station fire water pump remote control system provided in this embodiment.

[0060] like Figure 2 As shown, this embodiment provides a photovoltaic power station fire water pump remote control system, including: Detection module 201 is used to deploy sensor arrays on the photovoltaic module backplane, combiner box and cable trench, and detect temperature information, smoke information and flame information through the sensor array; The conversion module 202 is used to perform protocol conversion through the embedded communication gateway to convert the temperature information into a temperature sequence, the smoke information into a smoke concentration gradient, and the flame information into an arc pulse frequency; Determination module 203, for inputting temperature sequence, smoke concentration gradient and arc pulse frequency into a time series analysis model, and superimposing component attenuation rate to determine hot spot risk index; Firefighting module 204 is used to start the first-level warning pre-pressurized water pump system when the hot spot risk index is greater than the first threshold; when the hot spot risk index is greater than the second threshold, trigger the second-level response, disconnect the DC circuit breaker and start the water pump; when the hot spot risk index is equal to the third threshold, activate the third-level response, execute full-site spraying and grid disconnection.

[0061] Figure 3 Schematic diagram of the structure of the electronic device provided in this embodiment.

[0062] like Figure 3 As shown, the electronic device may include: a processor (processor) 310, a communication interface (Communications Interface) 320, a memory (memory) 330 and a communication bus 340, wherein the processor 310, the communication interface 320, and the memory 330 communicate with each other through the communication bus 340. The processor 310 can call the logic instructions in the memory 330 to execute the remote control method of the fire water pump of the photovoltaic power station, which includes: deploying a sensor array on the back panel of the photovoltaic module, the junction box and the cable trench, and detecting temperature information, smoke information and flame information through the sensor array; performing protocol conversion through the embedded communication gateway, converting the temperature information into a temperature sequence, converting the smoke information into a smoke concentration gradient, and converting the flame information into an arc pulse frequency; inputting the temperature sequence, the smoke concentration gradient and the arc pulse frequency into the timing analysis model, and superimposing the component attenuation rate to determine the hot spot risk index; when the hot spot risk index is greater than the first threshold, starting the first-level warning pre-pressurization water pump system; when the hot spot risk index is greater than the second threshold, triggering the second-level response, linking the DC circuit breaker to disconnect and start the water pump; when the hot spot risk index is equal to the third threshold, activating the third-level response, executing full-field spraying and grid disconnection.

[0063] Furthermore, the logic instructions in the aforementioned memory 330 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the portion that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product, stored in a storage medium, includes several instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, a mobile hard drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0064] On the other hand, the present invention also provides a computer program product, which includes a computer program, which can be stored on a non-transitory computer-readable storage medium. When the computer program is executed by a processor, the computer can execute the photovoltaic power station fire water pump remote control method provided by the above methods, the method including: deploying a sensor array on the photovoltaic module backplane, junction box and cable trench, and detecting temperature information, smoke information and flame information through the sensor array; performing protocol conversion through an embedded communication gateway to convert the temperature information into a temperature sequence, the smoke information into a smoke concentration gradient, and the flame information into an arc pulse frequency; inputting the temperature sequence, the smoke concentration gradient and the arc pulse frequency into a timing analysis model, and superimposing the component attenuation rate to determine the hot spot risk index; when the hot spot risk index is greater than a first threshold, starting a first-level warning pre-pressurization water pump system; when the hot spot risk index is greater than a second threshold, triggering a second-level response, linking the DC circuit breaker to disconnect and start the water pump; when the hot spot risk index is equal to a third threshold, activating a third-level response, executing full-field spraying and grid disconnection.

[0065] On the other hand, the present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, is implemented to execute the photovoltaic power station fire water pump remote control method provided by the above-mentioned methods, the method comprising: deploying a sensor array on the photovoltaic module backplane, junction box and cable trench, and detecting temperature information, smoke information and flame information through the sensor array; performing protocol conversion through an embedded communication gateway, converting the temperature information into a temperature sequence, converting the smoke information into a smoke concentration gradient, and converting the flame information into an arc pulse frequency; inputting the temperature sequence, the smoke concentration gradient and the arc pulse frequency into a timing analysis model, and superimposing the component attenuation rate to determine the hot spot risk index; when the hot spot risk index is greater than a first threshold, starting a first-level warning pre-pressurization water pump system; when the hot spot risk index is greater than a second threshold, triggering a second-level response, linking the DC circuit breaker to disconnect and start the water pump; when the hot spot risk index is equal to a third threshold, activating a third-level response, executing full-field spraying and grid disconnection.

[0066] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0067] Through the above description of the embodiments, those skilled in the art will clearly understand that each embodiment can be implemented using software plus a necessary general-purpose hardware platform, or of course, hardware. Based on this understanding, the essence of the above technical solution, or the portion that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, or an optical disk, and includes a number of instructions for causing a computer device (such as a personal computer, server, or network device) to execute the methods described in each embodiment or certain portions of the embodiments.

[0068] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A remote control method for a fire pump in a photovoltaic power station, characterized in that: include: Deploy sensor arrays on photovoltaic module back panels, combiner boxes, and cable trenches, and detect temperature information, smoke information, and flame information through the sensor arrays; Performing protocol conversion through an embedded communication gateway to convert the temperature information into a temperature sequence, the smoke information into a smoke concentration gradient, and the flame information into an arc pulse frequency; Inputting the temperature sequence, the smoke concentration gradient, and the arc pulse frequency into a timing analysis model, and superimposing component attenuation rates to determine a hot spot risk index; When the hot spot risk index is greater than a first threshold, starting a first-level warning pre-pressurization water pump system; When the hot spot risk index is greater than a second threshold, a secondary response is triggered, the DC circuit breaker is disconnected and the water pump is started; When the hot spot risk index is equal to the third threshold, a third-level response is activated, performing full-field spraying and grid disconnection.

2. The photovoltaic power station fire water pump remote control method according to claim 1 is characterized in that: The method of performing protocol conversion through an embedded communication gateway to convert the temperature information into a temperature sequence, convert the smoke information into a smoke concentration gradient, and convert the flame information into an arc pulse frequency includes: Perform linear normalization on the temperature information to generate a temperature sequence arranged by timestamp; Discrete sampling and interpolation calculation are performed on smoke information to generate a continuously changing smoke concentration gradient; The light intensity fluctuation characteristics of the flame information are subjected to Fourier transform, and the fundamental frequency component is extracted as the arc pulse frequency.

3. The photovoltaic power station fire water pump remote control method according to claim 1, characterized in that: The step of inputting the temperature sequence, the smoke concentration gradient, and the arc pulse frequency into a timing analysis model and superimposing the component attenuation rate to determine the hot spot risk index includes: extracting multi-scale features of the temperature sequence, the smoke concentration gradient, and the arc pulse frequency; The multi-scale features are identified through a CNN-BiLSTM-attention fusion model, and a hot spot score is output; Identify the multi-scale features by constructing an attenuation model and output a component attenuation rate; The hot spot score and the component attenuation rate are nonlinearly superimposed to obtain a hot spot risk index.

4. The photovoltaic power station fire water pump remote control method according to claim 3 is characterized in that: The multi-scale features are identified by the CNN-BiLSTM-attention fusion model, and the hot spot score is output, including: Slice the multi-scale features into time windows, extract statistical features, frequency domain features, and time series features, and construct a standardized three-dimensional input tensor; Extracting local spatiotemporal features from the standardized three-dimensional input tensor through a three-layer one-dimensional convolutional network to generate a fused feature map; A two-layer bidirectional LSTM network is used to capture the temporal dependency of the fused feature map. Each layer is followed by layer normalization and residual connection to output sequence features. Applying a dual attention mechanism of time dimension and feature dimension to the sequence features, focusing on key temporal segments and feature dimensions, and generating weighted aggregate features; The weighted aggregate features are generated into a probability distribution through a fully connected classification network to determine the hot spot score.

5. The photovoltaic power station fire water pump remote control method according to claim 3, characterized in that: The identifying the multi-scale features by constructing an attenuation model and outputting a component attenuation rate includes: Based on the physical mechanisms of light-induced degradation, potential-induced degradation and temperature fatigue, a theoretical degradation calculation model is established to determine the theoretical degradation amount; Input the multi-scale feature vector into the data-driven model, fit the attenuation trend in the complex environment through nonlinear mapping, and output the measured attenuation reflecting the real-time operation abnormality; The theoretical attenuation and the measured attenuation are weighted and fused to determine the component attenuation rate.

6. The photovoltaic power station fire water pump remote control method according to claim 3, characterized in that: The hot spot risk index is obtained by nonlinearly superimposing the hot spot score and the component attenuation rate, including: ； in, Characterizes the current abnormal signal strength, The increase in basic risk due to component aging, Characterize the synergistic amplification effect of abnormal signals × component aging; 、 and represents the weight, represents the hot spot score, Represents the component attenuation rate.

7. The photovoltaic power station fire water pump remote control method according to any one of claims 1 to 6, characterized in that: The starting of the first-level early warning pre-pressurization water pump system includes: Send pre-pressurization instructions to the water pump inverter, start the water pump auxiliary motor, adjust the opening of the electric control valve, and verify the pipe network pressure sensor reading; The three-dimensional coordinates and risk value of the warning area are displayed on the operation and maintenance terminal.

8. The photovoltaic power station fire water pump remote control method according to any one of claims 1 to 6, characterized in that: The triggering of the secondary response, which triggers the DC circuit breaker to disconnect and start the water pump, includes: Cut off the faulty string DC circuit breaker; Start the corresponding zone water pump and verify the outlet pressure; Push the escape route heat map to the operation and maintenance terminal.

9. The photovoltaic power station fire water pump remote control method according to any one of claims 1 to 6, characterized in that: The activation of the three-level response, execution of full-site spraying and grid disconnection, includes: Start the main spray pump. If the pipe network pressure does not meet the standard after the target spraying time, switch to the backup pump. According to the hot spot positioning system, open the solenoid valve group in the risk area and expand it to fully open the solenoid valves in the entire field to achieve a complete spray coverage. During the no-dead-angle spray coverage process, the DC circuit breaker of the combiner box, the AC circuit breaker of the inverter and the high-voltage side switch of the isolation transformer are disconnected in sequence, and the grounding switch is closed in conjunction to achieve electrical isolation between the photovoltaic system and the power grid.

10. A remote control system for a photovoltaic power station fire pump, characterized in that: include: A detection module is used to deploy a sensor array on the photovoltaic module backplane, combiner box and cable trench, and detect temperature information, smoke information and flame information through the sensor array; a conversion module, configured to perform protocol conversion through an embedded communication gateway, convert the temperature information into a temperature sequence, convert the smoke information into a smoke concentration gradient, and convert the flame information into an arc pulse frequency; a determination module, configured to input the temperature sequence, the smoke concentration gradient, and the arc pulse frequency into a timing analysis model, and superimpose component attenuation rates to determine a hot spot risk index; The fire protection module is used to start the first-level warning pre-pressurized water pump system when the hot spot risk index is greater than the first threshold; when the hot spot risk index is greater than the second threshold, trigger the second-level response, disconnect the DC circuit breaker and start the water pump; when the hot spot risk index is equal to the third threshold, activate the third-level response, execute full-site spraying and grid disconnection.

Images