Fire linkage control method and system for photovoltaic power station, medium and product

Abstract

The invention belongs to the technical field of fire fighting, and discloses a fire linkage control method and system for a photovoltaic power station, a medium and a product, and the method comprises the steps: converging multi-source heterogeneous sensing data collected by a fixed monitoring equipment group and a mobile inspection equipment group which are disposed in a target photovoltaic power station, and obtaining a power station monitoring data stream; performing feature extraction on the power station monitoring data flow by using a pre-constructed fire identification model, and comparing the feature with a preset fire feature to obtain fire information; based on the fire information, combining a pre-created digital twinborn model of the target photovoltaic power station to generate structured fire situation information; according to the structured fire situation information, a preset emergency response plan is matched and activated, and a linkage control instruction is generated in combination with real-time environment parameters or predicted environment parameters of the target photovoltaic power station; the linkage control instruction is sent to one or more equipment execution units; the current situation that alarm and fire fighting are disjointed is changed, and the emergency efficiency is improved.

Description

Technical Field

[0001] This invention belongs to the field of fire protection technology, and specifically relates to a fire linkage control method, system, medium and product for photovoltaic power plants. Background Technology

[0002] As a crucial component of the modern energy system, photovoltaic power plants play a key role in promoting clean energy development and reducing environmental pollution. Their safe and stable operation not only affects the economic benefits of the power plant itself but also directly impacts the sustainable and healthy development of the new energy industry. However, photovoltaic power plants typically occupy vast areas, have complex equipment and wiring, and are exposed to the outdoor environment for extended periods, making them vulnerable to numerous safety hazards. Among these, the risk of persistent fires caused by equipment aging, severe weather conditions, and external factors is particularly prominent. Once a fire occurs, it can not only cause enormous asset losses but also trigger a series of chain reactions, posing a serious threat to the surrounding environment and the lives of people. Therefore, in the field of photovoltaic power plants, achieving efficient and intelligent fire prevention and control has become the core of ensuring the healthy development of the new energy industry.

[0003] Currently, the main fire prevention and control practice in existing photovoltaic power plants adopts a fixed-point monitoring-based model. This model involves installing traditional fire detectors such as heat and smoke detectors in key areas of the photovoltaic power plant, such as inverter rooms and combiner boxes. By monitoring changes in physical quantities in the environment, such as temperature increases and smoke generation, it determines whether a fire has occurred and issues a fire alarm. However, in practical applications, the fixed-point monitoring-based model fails to effectively integrate fire alarm information with firefighting information, creating information silos. This results in a lack of proactive fire-fighting linkage control mechanisms, severely impacting the speed and efficiency of fire emergency response, significantly increasing the difficulty of fire prevention and control in photovoltaic power plants, and raising their safety risks. Summary of the Invention

[0004] In view of the technical problems existing in the prior art, the present invention provides a fire linkage control method, system, medium and product for photovoltaic power plants, so as to solve the technical problem that the prevention and control mode based on fixed point monitoring in the existing photovoltaic power plant fire prevention and control practice lacks an active fire linkage control mechanism, which seriously affects the speed and efficiency of fire emergency response and greatly increases the difficulty of fire prevention and control in photovoltaic power plants.

[0005] To achieve the above objectives, the technical solution adopted by the present invention is as follows: This invention provides a fire linkage control method for photovoltaic power plants, comprising: The multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station are aggregated to obtain the power station monitoring data stream; Using a pre-built fire identification model, features are extracted from the power plant monitoring data stream and compared with preset fire features to obtain fire information; based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power plant, structured fire situation information is generated. Based on the structured fire situation information, the preset emergency response plan is matched and activated, and combined with the real-time or predicted environmental parameters of the target photovoltaic power station, a linkage control command is generated. Send linkage control commands to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

[0006] Furthermore, structured fire situation information includes fire location information, fire severity level, and fire development trend information; The process of generating structured fire situation information based on fire information and a pre-created digital twin model of the target photovoltaic power station is as follows: Based on fire information, the fire point is located in the pre-created digital twin model of the target photovoltaic power station to obtain the fire point location information; The fire severity level is determined based on the proportion of fire information in the power plant monitoring data stream; Based on the fire location information and the geographic information system built into the pre-created digital twin model of the target photovoltaic power station, the spatial topology of the fire point is obtained; based on the spatial topology of the fire point and the wind direction and speed data of the fire point, the fire development trend information is obtained. Output fire location information, fire severity level, and fire development trend information to obtain structured fire situation information.

[0007] Furthermore, the pre-created digital twin model of the target photovoltaic power station includes a mirror model and a projection model; Among them, the mirror model is a digital twin model that keeps data synchronized with the target photovoltaic power station, and the derivation model is a derivation model of the mirror model after accelerated time stream processing; Structured fire situation information includes first structured fire situation information and second structured fire situation information; Specifically, fire information is used as a perturbation variable, and a mirror model is used to generate the first structured fire situation information; fire information is used as a perturbation variable, and a deductive model is used to generate the second structured fire situation information.

[0008] Furthermore, based on the structured fire situation information, the process of matching and activating the preset emergency response plan, and generating linkage control commands in conjunction with the real-time or preset environmental parameters of the target photovoltaic power station, is as follows: Based on the first structured fire situation information, the preset emergency response plan is matched and activated, and the first linkage control command is generated in combination with the real-time environmental parameters of the target photovoltaic power station. Based on the second structured fire situation information, the preset emergency response plan is matched and activated, and combined with the predicted environmental parameters of the target photovoltaic power station, a second linkage control command is generated.

[0009] Furthermore, the process of sending linkage control commands to one or more device execution units is as follows: Send the first linkage control command to one or more device execution units; During the execution of the first linkage control command by the equipment execution unit, the multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station are re-aggregated to obtain the power station monitoring update data stream; Based on the power plant monitoring update data stream, the updated first structured fire situation information is obtained; The updated first structured fire situation information is compared with the second structured fire situation information to obtain the state deviation value. When the state deviation value is less than the preset state deviation threshold, a second linkage control command is sent to one or more device execution units; When the state deviation value is not less than the preset state deviation threshold, the preset emergency response plan is matched and activated based on the updated first structured fire situation information, and the third linkage control command is generated in combination with the real-time environmental parameters of the target photovoltaic power station. Send a third linkage control command to one or more device execution units to switch the device execution units from executing the first linkage control command to executing the third linkage control command.

[0010] Furthermore, before obtaining the power plant monitoring data stream, the process of aggregating multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups deployed within the target photovoltaic power plant also includes: Scanning monitoring is performed using a fixed monitoring equipment group deployed within the target photovoltaic power station to obtain preliminary scanning monitoring data; Based on the pre-scan detection data, determine whether there are abnormal temperature areas within the target photovoltaic power station, and determine the coordinate information of the abnormal temperature areas; If there is an abnormal temperature area within the target photovoltaic power station, the mobile inspection equipment within the target photovoltaic power station will be dispatched to the abnormal temperature area based on the coordinate information of the abnormal temperature area to begin multi-source heterogeneous sensing data collection.

[0011] The present invention also provides a fire linkage control system for photovoltaic power plants, comprising: The data stream aggregation module is used to aggregate multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups deployed in the target photovoltaic power station to obtain the power station monitoring data stream; The fire situation identification module is used to extract features from the power plant monitoring data stream using a pre-built fire identification model and compare the features with preset fire features to obtain fire information; based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power plant, structured fire situation information is generated. The control command generation module is used to match and activate the preset emergency response plan based on the structured fire situation information, and generate linkage control commands in combination with the real-time or predicted environmental parameters of the target photovoltaic power station. The instruction sending and execution module is used to send linkage control instructions to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

[0012] The present invention also provides an electronic device, comprising: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the fire linkage control method for photovoltaic power plants.

[0013] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the fire linkage control method for photovoltaic power plants.

[0014] The present invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the fire linkage control method for photovoltaic power plants.

[0015] Compared with the prior art, the beneficial effects of the present invention are as follows: The fire linkage control method for photovoltaic power plants provided by this invention changes the current situation of disconnect between alarm and fire fighting, and improves emergency response efficiency. Specifically, by aggregating multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups, it breaks the information silo limitation of the prevention and control mode based on fixed-point monitoring. Based on the pre-constructed fire identification model and the pre-constructed digital twin model of the target photovoltaic power plant, it realizes the accurate extraction of fire information and the structured presentation of the fire situation. This enables the activation of emergency response plans based on structured fire situation information, and generates linkage control commands by combining the real-time or predicted environmental parameters of the target photovoltaic power plant. This drives the active fire-fighting drone and the fixed monitoring equipment group to coordinate and execute fire prevention and control operations, effectively constructing an active fire-fighting linkage control mechanism of monitoring-identification-situation analysis-linkage control. This significantly improves the speed and efficiency of fire emergency response, greatly reduces the difficulty and safety risks of fire prevention and control in photovoltaic power plants, ensures the safe and stable operation of photovoltaic power plants, and provides key technical support for the safe and stable operation of the new energy industry.

[0016] Furthermore, by accurately locating fire points in a pre-constructed digital twin model of the target photovoltaic power station, the fire situation is classified according to its location and the proportion of fire information in the power station's monitoring data stream. Combining the spatial topology of the fire points and wind direction and speed data obtained from the built-in geographic information system in the pre-created digital twin model, the fire development trend is determined. This process integrates and generates structured fire situation information containing fire point location, fire level, and fire development trend information. This not only achieves comprehensive and detailed perception and description of the fire situation at the target photovoltaic power station, but also provides reliable decision-making basis for the precise matching and activation of subsequent emergency response plans and the generation of linkage control commands. This effectively improves the pertinence and effectiveness of fire linkage control, further reduces the difficulty and safety risks of fire prevention and control in photovoltaic power stations, and ensures the safe and stable operation of the photovoltaic power station.

[0017] Furthermore, by introducing a mirror model that keeps data synchronized with the power plant and a deductive model that has undergone accelerated time-stream processing, using fire information as perturbation variables, a first structured fire situation information reflecting the current fire situation and a second structured fire situation information predicting the fire development trend are generated. This achieves accurate mapping of the current fire situation of the target photovoltaic power plant and deduction of its future development trend. It provides dual-dimensional and high-precision decision support for the matching and activation of subsequent emergency response plans and the differentiated generation of linkage control commands. It effectively constructs a three-dimensional fire analysis system that combines real-time monitoring and trend prediction, significantly improving the scientific nature and foresight of fire linkage control, and further reducing the difficulty and safety risks of fire prevention and control in photovoltaic power plants.

[0018] Furthermore, by generating a first linkage control command based on the first structured fire situation information reflecting the current fire situation and combined with real-time environmental parameters, and simultaneously generating a second linkage control command based on the second structured fire situation information predicting the fire development trend and combined with predicted environmental parameters, a two-layer linkage control logic of real-time response + forward-looking prediction is formed. This not only achieves accurate matching between emergency response plans and fire situation and environmental parameters, but also ensures the timeliness and foresight of linkage control commands, effectively improving the pertinence and effectiveness of fire linkage control, effectively optimizing the overall efficiency of photovoltaic power station fire emergency response, and significantly reducing fire prevention and control risks.

[0019] Furthermore, during the execution of the first linkage control command by the equipment execution unit, multi-source heterogeneous sensing data is re-aggregated to obtain the power plant monitoring update data stream, thereby obtaining the updated first structured fire situation information, which is then compared with the second structured fire situation information to obtain the state deviation value. Based on the comparison result of the state deviation value and the preset threshold, the second linkage control command is sent or the third linkage control command is generated and sent, realizing dynamic closed-loop control of the fire linkage control process. This effectively avoids the problem of control command failure caused by dynamic changes in the fire situation or prediction deviations, improves the adaptability and execution accuracy of the linkage control command, further optimizes the operational efficiency of the real-time response + forward prediction dual-layer linkage control mechanism, and significantly enhances the timeliness and reliability of photovoltaic power plant fire emergency response.

[0020] Furthermore, by using fixed monitoring equipment to conduct preliminary scanning and monitoring before aggregating multi-source heterogeneous sensing data, and then identifying and determining abnormal temperature areas and their coordinates, mobile inspection equipment is then dispatched to these abnormal areas to collect data. This avoids the information limitations of fixed monitoring and the blindness of mobile inspection, achieving preliminary collaboration and precise linkage between fixed monitoring and mobile inspection. This not only significantly improves the targeting, effectiveness, and efficiency of multi-source heterogeneous sensing data collection, but also enables early identification and locking of fire hazards in photovoltaic power plants. This lays a high-quality data foundation for the accurate judgment of subsequent fire identification models, the scientific generation of structured fire situation information, and the efficient issuance of linkage control commands.

[0021] The fire linkage control system, electronic equipment, computer-readable storage medium, and computer program products for photovoltaic power plants provided by this invention possess all the advantages of the aforementioned fire linkage control methods for photovoltaic power plants. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 A flowchart of a fire linkage control method for photovoltaic power plants provided in Example 1; Figure 2 This is a structural block diagram of the hardware system used to implement the fire linkage control method for photovoltaic power plants in Example 1; Figure 3 This is a schematic diagram of the use of the fixed monitoring equipment group in Example 1; Figure 4 This is a schematic diagram illustrating the use of the mobile monitoring equipment group in Example 1; Figure 5 This is a schematic diagram illustrating the use of the active firefighting drone in Example 1; Figure 6 This is a structural block diagram of a fire linkage control system for photovoltaic power plants provided in Example 2; Figure 7 This is a structural block diagram of the electronic device provided in Example 3. Detailed Implementation

[0024] To make the technical problems, technical solutions, and beneficial effects solved by this application clearer, the technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0025] As attached Figure 1 As shown, this invention provides a fire linkage control method for photovoltaic power plants, comprising the following steps: Step 1: Aggregate the multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station to obtain the power station monitoring data stream.

[0026] Step 2: Using a pre-built fire identification model, extract features from the power plant monitoring data stream and compare them with preset fire features to obtain fire information; based on the fire information, combine it with a pre-created digital twin model of the target photovoltaic power plant to generate structured fire situation information.

[0027] Step 3: Based on the structured fire situation information, match and activate the preset emergency response plan, and generate linkage control commands by combining the real-time or predicted environmental parameters of the target photovoltaic power station.

[0028] Step 4: Send linkage control commands to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

[0029] In the above implementation, by aggregating multi-source heterogeneous sensing data from fixed monitoring equipment groups and mobile inspection equipment groups, comprehensive integration and interoperability of power plant monitoring data are achieved. Based on a pre-constructed fire identification model, accurate extraction and judgment of fire information are completed, and structured fire situation information is generated by combining the digital twin model of the photovoltaic power plant, providing accurate decision-making basis for subsequent emergency response. By matching and activating preset emergency response plans, and generating linkage control commands based on real-time or predicted environmental parameters of the power plant, the linkage control commands are sent to the equipment execution units of the active fire-fighting drones and fixed monitoring equipment groups, constructing a full-link active fire-fighting linkage control mechanism from fire perception and situation analysis to command execution. This effectively realizes the fusion of fire alarm information and fire-fighting information, significantly improves the speed and efficiency of emergency response to photovoltaic power plant fires, significantly reduces the difficulty of fire prevention and control and the safety risks of the power plant, ensures the safe and stable operation of the photovoltaic power plant, and thus promotes the sustainable and healthy development of the new energy industry.

[0030] The following specific embodiments further explain the fire linkage control method for photovoltaic power plants provided by the present invention: Example 1 As attached Figure 1 As shown in the figure, this embodiment 1 provides a fire linkage control method for photovoltaic power plants, including the following steps: S101. Aggregate the diverse and heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station to obtain the power station monitoring data stream.

[0031] Please see the appendix Figure 2 , attached Figure 2 The hardware system used to implement the fire linkage control method for photovoltaic power plants described in Embodiment 1 serves as the execution subject of the method in this application. In the subsequent description of this application, unless a specific execution subject is specifically indicated, the steps are assumed to be executed by this hardware system, and will not be described again here.

[0032] The fixed monitoring equipment group includes fixed monitoring equipment permanently or semi-permanently installed at predetermined key locations in the target photovoltaic power station; among which, the predetermined key locations in the target photovoltaic power station include combiner boxes, inverters, or step-up substations, the fixed monitoring equipment includes infrared thermal imaging cameras, visible light high-definition cameras, smoke detectors, and temperature detectors; as shown in the attached document. Figure 3 As shown, attached Figure 3 The diagram below shows the usage of the fixed monitoring equipment group in Embodiment 1.

[0033] The mobile monitoring equipment group includes mobile monitoring devices capable of moving within the target photovoltaic power plant on demand or along a preset path; wherein the mobile monitoring devices employ inspection drones or ground inspection drones equipped with preset mission payloads, such as high-definition zoom cameras and thermal imagers; as shown in the attached document. Figure 4 As shown, attached Figure 4 The diagram below shows the usage of the mobile monitoring equipment group in Embodiment 1.

[0034] It should be noted that the data collected by the fixed monitoring equipment group and the mobile inspection equipment group can be of the same or different types, the difference being that the data sources or the angles from which they are collected are different.

[0035] In some implementations, communication connections are established with fixed or mobile monitoring equipment groups within the target photovoltaic power plant via wired or wireless networks (such as 5G networks or Wi-Fi) to aggregate multi-source heterogeneous sensing data. For example, real-time reception of wide-area monitoring images and temperature or smoke readings from key nodes from fixed cameras, along with close-range reconnaissance images and location information transmitted by drones during inspection missions, is achieved. By preprocessing the real-time received data, such as decoding, format conversion, and time / space synchronization, it is ultimately integrated into a unified real-time data stream containing multi-dimensional information, serving as the power plant monitoring data stream and providing comprehensive and seamless on-site sensing data for subsequent intelligent identification and analysis.

[0036] In other implementations, fixed monitoring equipment sets have unique advantages in data acquisition, as they can cover a large area and continuously collect data; while mobile inspection equipment sets also have their own characteristics, with a relatively smaller data collection range, but the collected data is often more detailed and clear. The data collected by the two can complement each other, play a synergistic role, and jointly provide more comprehensive and effective data support for subsequent work.

[0037] S102. Using a pre-built fire identification model, feature extraction is performed on the power station monitoring data stream, and the features are compared with preset fire features to obtain fire information. Based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power station, structured fire situation information is generated.

[0038] The pre-built fire identification model uses deep learning algorithms such as CNN and YOLO trained on a large amount of data. The pre-built fire identification model can automatically identify fire-related features from the input power plant monitoring data stream, such as images, videos or sensor data.

[0039] A pre-created digital twin model of a target photovoltaic power station refers to a three-dimensional virtual model that is precisely mapped to a physical photovoltaic power station. Its creation process includes: based on the pre-acquired operation and environmental data of the photovoltaic power station, using digital twin technology to construct a digital twin model of the target photovoltaic power station as the pre-created digital twin model of the target photovoltaic power station. It not only includes the static geographical information and equipment layout of the target photovoltaic power station, but also can synchronize dynamic data such as operating status.

[0040] Structured fire situation information includes fire location information, fire severity level, and fire development trend information. In essence, it integrates scattered fire alarm information into a standardized intelligence format that contains multiple key fields (such as fire location, fire severity level, and fire development trend) and can be directly interpreted by machines and used for decision-making.

[0041] In this embodiment 1, the process of generating structured fire situation information based on fire information and a pre-created digital twin model of the target photovoltaic power station includes: locating the fire point in the pre-created digital twin model of the target photovoltaic power station based on the fire information to obtain the fire point location information; obtaining the fire level based on the proportion of fire information in the power station monitoring data stream; obtaining the spatial topology of the fire point based on the fire point location information and the built-in geographic information system in the pre-created digital twin model of the target photovoltaic power station; obtaining the fire point development trend information based on the spatial topology of the fire point and the wind direction and speed data of the fire point; and outputting the fire point location information, fire level, and fire point development trend information to obtain structured fire situation information.

[0042] In some implementations, when the power plant monitoring data stream in S101 is continuously input, the pre-built fire identification model analyzes the power plant monitoring data stream in real time to extract suspected fire features; the extracted suspected fire features are compared with preset fire features; if the matching degree exceeds a preset threshold, it is confirmed as a fire and fire information is obtained; when it is confirmed as a fire, the fire information is input into the pre-created digital twin model of the target photovoltaic power plant, and the three-dimensional coordinates of the fire point are parsed as the fire location information using the precise spatial information built into the digital twin model, and the fire level and fire point development trend information are generated, and finally the fire point location information, fire level and fire point development trend information are output to obtain structured fire situation information.

[0043] In some specific embodiments, S102 specifically includes: SA1 decodes and aligns the power plant monitoring data stream with time and space stamps to separate image / video data and sensor time-series data.

[0044] Decoding refers to the process of converting compressed media data (such as H.264 format video) into raw pixel data (such as RGB matrix) that can be directly processed by a computer; spatiotemporal stamp alignment refers to calibrating and synchronizing time stamps from different devices and different types of data packets based on a unified, high-precision master clock to ensure that all data are comparable in the time dimension; image / video data refers to visual information collected by devices such as cameras and infrared thermal imagers in frames; sensor time-series data refers to a sequence of physical quantity measurements that change continuously over time, collected by devices such as temperature, smoke, and gas sensors.

[0045] After aggregating the raw data streams from multiple cameras and sensors within the target photovoltaic power plant, step SA1 first functions as a data preprocessing center. It identifies the encoding format of each data block, for example, using decoding libraries like FFmpeg to restore video data streams into frame-by-frame images, and parsing sensor data to extract specific values ​​according to their protocols. Crucially, it aligns the timestamps by reading the timestamps inherent in each data packet and correcting all data timestamps to the same timeline based on the pre-calibrated offset between each device's clock and the system's master clock. For example, the 100th frame of a drone video and the 50th temperature reading from a fixed sensor, even with network transmission delays, can be accurately identified as events occurring at the same time after alignment. Finally, after processing by step SA1, the originally chaotic single data stream is clearly separated into two independent but time-synchronized data channels: one for visual information and the other for physical quantity sensing information.

[0046] SA2. Input image / video data into the convolutional neural network module in the pre-built fire recognition model to extract visual feature vectors of flames and smoke.

[0047] Specifically, frame by frame of decoded and spatiotemporally aligned image / video data is fed into a Convolutional Neural Network (CNN) module. This CNN module contains multiple convolutional layers, pooling layers, and fully connected layers. During processing, shallow convolutional layers first identify some basic features, such as edges and color blocks (specifically, the bright yellow of flames or the grayish-white of smoke). As the number of network layers increases, deeper layers combine these basic features to identify more complex patterns, such as "the outline of rising smoke" or "the irregularly jumping shape of flames." Finally, after the image information undergoes nonlinear transformation by the entire CNN module, it is compressed into a fixed-length numerical array, i.e., a visual feature vector.

[0048] SA3. Input the sensor time-series data into the Long Short-Term Memory (LSTM) network module in the pre-built fire identification model to extract the time-series feature vectors of temperature and gas concentration. The LSTM module is the part of the model that is specifically responsible for processing sequence data. The time-series feature vector is a set of high-dimensional values ​​output by the LSTM module, which quantifies the dynamic change trend, fluctuation pattern and anomaly degree of sensor data over a period of time, rather than just the instantaneous value at the current moment.

[0049] Specifically, a continuous, time-synchronized series of temperature and gas concentration readings, such as the sampled values ​​per second over the past 60 seconds, is input as a sequence into the LSTM module. Unlike CNNs, which process spatial information, the core of the LSTM module lies in its internal memory units and input, forget, and output gates. This allows it to selectively "remember" important information from the historical sequence, such as the normal temperature baseline from one hour ago, and "forget" unimportant information when analyzing data points at the current moment. Therefore, the LSTM module can accurately identify dynamic patterns with obvious fire characteristics, such as "temperature rising sharply from 25°C to 90°C within 10 seconds," and can distinguish them from normal solar warming patterns, such as "temperature rising slowly from 25°C to 45°C within 2 hours." Finally, the LSTM module compresses the dynamic features of this time series and outputs it as a temporal feature vector.

[0050] SA4. The visual feature vector and the temporal feature vector are concatenated and weighted to generate a fused feature value, which serves as the suspected fire feature.

[0051] Specifically, the process involves obtaining visual feature vectors generated by CNN and temporal feature vectors generated by LSTM. First, the visual and temporal feature vectors are concatenated to form a longer fusion vector. Then, a set of pre-trained weights is used to perform a weighted summation on each component of the fusion vector. The pre-trained weights reflect the contribution of different features to the fire assessment. For example, the model may discover during training that "sharp temperature rise" (corresponding to the first component of the temporal vector) is a stronger fire indicator than "faint smoke" (corresponding to a component of the visual vector), thus giving "faint smoke" a higher weight. This intelligently fuses multi-dimensional and multi-modal evidence into a single value representing the overall confidence level.

[0052] SA5. The fused feature values ​​are compared with the preset fire level threshold table to determine the fire level; the pixel coordinates of the fire point in the image are mapped to the digital twin model through a coordinate transformation algorithm to obtain three-dimensional position information, which is used as the fire point position information; and the fire point development trend information is obtained by calculating the rate of change of position information at continuous time points.

[0053] SA6 outputs fire location information, fire severity level, and fire development trend information to obtain structured fire situation information.

[0054] S103. Based on the structured fire situation information, match and activate the preset emergency response plan, and generate linkage control commands in combination with the real-time environmental parameters of the target photovoltaic power station.

[0055] The preset emergency response plan refers to a collection of standardized handling procedures and resource scheduling schemes stored in advance for different fire levels, different locations, and different equipment types; matching and activation means that based on the parameters of the current structured fire situation information (such as the location being in the inverter room and the level being level 2), the most suitable set of plans is automatically searched for and activated from the preset emergency response plans.

[0056] In some implementations, step S103 is triggered immediately after the structured fire situation information is generated in the previous step. First, the current structured fire situation information is used as an index to search in the preset emergency response plan. After the corresponding plan is activated, the fixed instructions in the plan are not issued directly, but a dynamic optimization stage is entered. By retrieving real-time wind direction and speed data, if it is found that the current wind direction will accelerate the spread of the fire to adjacent components, an instruction may be added on the basis of activating the plan to dispatch another drone to spray flame retardant on the predicted spread path. Finally, a series of specific linkage control instructions that combine the plan and real-time parameters and have been dynamically optimized are generated.

[0057] In some specific implementations, S103 specifically includes: SB1. Using the fire location information in the structured fire situation information as an index, retrieve all equipment objects within a preset range around the fire point in the pre-created digital twin model of the target photovoltaic power station, and filter out active fire-fighting drones that are in an available state.

[0058] An index is a data key used for quick querying and locating, similar to a primary key in a database. A preset range is a predefined three-dimensional spatial region centered on the three-dimensional coordinates of the fire point, such as a sphere with a radius of 500 meters or a cube with a side length of 500 meters, used to define the boundaries of resource retrieval. A device object is a digital mapping entity of various devices in the physical world in a digital twin model, including its static attributes (such as model and ID) and dynamic status (such as location, battery level, and working status). A status of "available" indicates that the active firefighting drone is currently in a comprehensive state where it can immediately accept and execute tasks, typically including multiple conditions such as battery level above the safety threshold, not performing other tasks, no fault alarms, and sufficient fire extinguishing agent on board.

[0059] Specifically, the three-dimensional coordinates of the fire points will be extracted from the structured fire situation information. As the core parameter of the query command, a spatial range query request is initiated to the digital twin model. The logic of this request is: "Please return the three-dimensional coordinates of the fire point." The system first generates a list of all equipment objects within a sphere centered at a location with a radius of 500 meters. Then, the digital twin model returns a list that may contain digital proxies of objects such as photovoltaic panels, combiner boxes, fixed cameras, and several active firefighting drones. Next, the returned list is filtered. Specifically, each drone object is iterated over, and its real-time updated status data is read, checking multiple attributes such as "battery level," "mission status," "health status," and "payload status." Only when a drone simultaneously meets all the "available" conditions will it be retained in the final candidate list. Finally, one or more drones meeting the criteria are selected as the preferred objects for performing this mission.

[0060] SB2. Using the preset ballistic calculation module, the location of the selected active fire-fighting drone, the location of the fire point, and the wind speed and direction data in the real-time environmental parameters are used as input to calculate one or more flight trajectories of the fire extinguishing agent that include the lead time of the projection.

[0061] The preset ballistics calculation module refers to a dedicated algorithm program that embeds a fluid dynamics and parabolic motion model; real-time environmental parameters refer to environmental data collected and uploaded in real time by the micro weather station deployed in the photovoltaic power station or by the sensors carried by the drone itself; the projection lead is a targeting correction calculated to offset the influence of environmental factors such as gravity and wind on the flight trajectory of the extinguishing agent, that is, the point that the drone actually aims at when projecting is not the fire point itself, but a spatial point upwind of the fire point; the flight trajectory of the extinguishing agent refers to the precise three-dimensional spatial trajectory of the extinguishing agent (such as dry powder or foam) as it flies through the air towards the fire source after leaving the drone.

[0062] Specifically, the location of the selected active firefighting drone, the location of the fire point, and the wind speed and direction data from the real-time environmental parameters are input into a preset ballistic calculation module. For example, the drone's current real-time 3D coordinates (starting point), the fire point's 3D coordinates (target point), and the real-time wind speed and direction vector (e.g., "northwest wind, 3 m / s"). The ballistic calculation module runs a physical simulation model, first simulating the basic parabolic trajectory of the extinguishing agent under no wind and gravity, then superimposing the wind force as a continuously applied lateral vector force onto the running physical simulation model to calculate the drift of the extinguishing agent. To accurately hit the target point, the preset ballistic calculation module reverse-engineers the "projection point" that the drone needs to aim at at the moment of projection; the spatial difference between this projection point and the target point is the "projection lead." Finally, the preset ballistic calculation module outputs a complete 3D curve from the projection point to the target point, i.e., the flight trajectory, which precisely describes how the extinguishing agent "falls with the wind" and covers the fire source.

[0063] SB3. Using the preset path planning module, the starting position of the active firefighting drone, the projection point position calculated by the ballistic calculation module, and the obstacle data in the digital twin model are used as inputs to calculate a three-dimensional flight path.

[0064] The preset path planning module refers to a software component that integrates a pathfinding algorithm such as A* (A-Star) or RRT (Rapid Expanding Random Tree); the projection point location refers to the precise three-dimensional spatial coordinates that the UAV needs to fly to in order to perform the optimal projection, calculated by the previous step SB2; obstacle data refers to the geometric models and spatial location information of all fixed and temporary obstacles stored in the digital twin model, such as photovoltaic arrays, utility poles, high-voltage lines, buildings, etc.; the three-dimensional flight path refers to a route consisting of a series of continuous three-dimensional spatial waypoints that ensures the UAV can fly safely and efficiently from the starting point to the destination.

[0065] Specifically, the pre-defined path planning module receives three key inputs: the drone's current GPS coordinates as the starting point, the projection point coordinates calculated in step SB2 as the ending point, and a 3D obstacle avoidance map derived from the digital twin model. This 3D obstacle avoidance map precisely marks the spatial occupancy of all obstacles. The path planning algorithm (e.g., A*) immediately searches in 3D space for a path from the starting point to the ending point that minimizes the total cost (usually the shortest distance), while ensuring that no point on the path collides with the spatial model of obstacles. The algorithm generates a series of continuous waypoints, such as... , , ..., The generated waypoints are connected to form the final three-dimensional flight path, which may include a series of actions such as climbing, level flight, turning, and descending to avoid all obstacles.

[0066] SB4. Combine and encapsulate the selected active firefighting drone ID, the calculated three-dimensional flight path, the calculated flight trajectory, and the preset position parameters of the fixed monitoring equipment group used for coordinated illumination into a data packet, which serves as a linkage control command.

[0067] An active firefighting drone ID is a unique identification code assigned to each drone for targeted command delivery. Preset position parameters, for PTZ (potentially tilted-zoom) fixed monitoring equipment, are a set of precise control parameters, typically including horizontal rotation angle (Pan), vertical pitch angle (Tilt), and lens zoom, used to instruct the camera to automatically turn and focus on a specific location. Combining and encapsulating refers to integrating data of different types and targets into a structured, single data transmission unit according to a predefined communication protocol format. A data packet is the basic unit of transmission in a digital communication network, containing address information, control information, and actual data.

[0068] Specifically, the process involves obtaining the ID of the selected active firefighting drone, such as "UAV-FIRE-01"; then, incorporating the long sequence of three-dimensional flight path waypoint coordinates generated in step SB3 and the flight trajectory parameters calculated in step SB2 (which may include projection angle, initial velocity, etc.) into the data packet; simultaneously, based on the fire point coordinates and the positions of surrounding fixed cameras, calculating the camera that can provide the best observation angle for this firefighting operation (such as "CAM-PV-A07") and its preset position parameters (such as Pan: 185°, Tilt: 45°, Zoom: 12x); finally, assembling all information, including the selected active firefighting drone ID, flight path, trajectory parameters, camera ID, and its preset position parameters, into a structured data packet according to the agreed format; the assembled data packet serves as the final generated linkage control command, containing all the command information required for this multi-device collaborative task.

[0069] S104. Issue the linkage control command to one or more designated equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups. The active fire-fighting drones are used to strike the fire source, and the fixed monitoring equipment groups are used for illumination. (See attached...) Figure 5 As shown, attached Figure 5 The diagram below shows the use of the active firefighting drone in Example 1.

[0070] Equipment execution unit refers to terminal equipment capable of receiving and physically executing control commands; active fire-fighting drones specifically refer to drones equipped with fire extinguishing devices (such as dry powder / water-based fire extinguishing bombs, high-pressure water guns, etc.) and possessing autonomous flight and precise delivery capabilities; fire source strike refers to the act of drones delivering fire extinguishing agents to the core area of ​​a fire source to directly extinguish the flames; illumination refers to the use of visible light cameras or dedicated lighting equipment in fixed monitoring equipment groups to turn on high-intensity light sources (such as strobe lights, lasers, etc.) to continuously illuminate or indicate the fire source area or the drone's operating area.

[0071] In some implementations, a pre-set control network distributes linkage control commands to designated devices. Upon receiving the command, the active firefighting drone automatically plans its flight path to the target coordinates and delivers fire extinguishing bombs. Simultaneously, the fixed camera closest to the fire receives the illumination command, adjusts its gimbal to aim at the fire source, and activates a high-intensity strobe mode. This illumination not only provides visual assistance to the drone at night or in dense smoke environments, enhancing its strike accuracy, but also provides clear images for subsequent effect assessment.

[0072] As can be seen, by first aggregating multi-source heterogeneous sensing data from fixed and mobile equipment groups, a comprehensive data foundation is provided for fire identification. On this basis, the fire identification model is invoked for feature comparison, enabling the system to accurately identify fire features from complex backgrounds. After identifying the fire, the system combines a digital twin model to generate structured fire situation information including location, level, and trend, and generates and issues linkage control commands to execution units such as drones. The entire process forms a seamless closed loop, changing the current situation of disconnect between alarm and fire protection, and improving emergency response efficiency.

[0073] The above embodiments provide a general overview of the most basic solution of this application. The following steps focus on the key steps in the general process of the previous embodiments, and elaborate on the more specific, in-depth and optimized implementation methods for each key step.

[0074] In some implementations, S201-S203 are included before S101, as follows: S201. Fixed monitoring equipment group performs scanning monitoring; specifically, fixed monitoring equipment group deployed in the target photovoltaic power station is used to perform scanning monitoring to obtain preliminary scanning monitoring data.

[0075] S202. During the scanning and monitoring process, when abnormal temperature data is detected, the fixed monitoring equipment group will report the abnormal signal and coordinate information. Specifically, based on the previous scanning and detection data, it will be determined whether there is a temperature abnormality area in the target photovoltaic power station, and the coordinate information of the temperature abnormality area will be determined.

[0076] Among them, abnormal temperature data refers to the temperature value of a certain point or area collected during the scanning process being significantly higher than the background temperature of the surrounding environment or higher than the historical normal temperature threshold of the device itself; abnormal signal and coordinate information refers to a data packet that includes the alarm type (e.g., abnormal temperature), alarm level, and the precise location coordinates of the abnormal point in the power plant map or 3D model.

[0077] S203. Dispatch the mobile inspection equipment group to fly to the abnormal point to obtain multi-source heterogeneous sensing data; specifically, if there is a temperature abnormality area in the target photovoltaic power station, then based on the coordinate information of the temperature abnormality area, dispatch the mobile inspection equipment in the target photovoltaic power station to move to the temperature abnormality area and start the collection of multi-source heterogeneous sensing data.

[0078] In some implementations, after acquiring an anomaly signal with precise coordinates, the location of the anomaly is assessed, and the closest and best-performing drone or inspection robot is selected as the response unit from among the standby drones or inspection robots. Subsequently, a flight mission command containing the target coordinates is generated and sent to the drone. After receiving the command, the drone will automatically take off and, based on its built-in path planning algorithm, quickly fly to or near the anomaly. Then, using its onboard high-definition zoom camera, thermal imager, and other equipment, it will conduct detailed reconnaissance of the anomaly from multiple angles and at close range, and transmit the collected higher-definition and richer multi-source heterogeneous sensing data back in real time.

[0079] It is evident that S201-S203 constructs a collaborative workflow of wide-area initial screening and close-range detailed investigation; its fixed reporting coordinates and precise mobile tracking feature interaction form an efficient discovery-confirmation closed loop, solving the problems of blind searching and time delay in related inspection modes, and improving the efficiency of early fire hazard discovery and confirmation.

[0080] In some implementations, steps S301-S303 are included before step S102, as follows: S301. Using mobile inspection equipment and fixed monitoring equipment under different conditions, collect full-scene images and video data of the photovoltaic power station, including simulated fire under normal operating conditions, to obtain the original sample database.

[0081] Among them, different conditions refer to the various environments and operating conditions that photovoltaic power plants may encounter, including but not limited to different seasons, different times, different weather, and various interference factors; simulated fire refers to the early smoke and fire characteristics of different scales and forms simulated using tools such as smoke cakes and alcohol lamps under safe and controllable conditions; full-scene image and video data refers to a diverse and high-fidelity visual data set collected by dynamic and static equipment under the above-mentioned conditions, including normal operating conditions, simulated fire, and various interference items; the original sample database is a database that stores raw, unlabeled data.

[0082] S302. Output the original sample database so that the full-scene image and video data are labeled.

[0083] Annotation refers to the process by which professionals use annotation tools to precisely outline the location of targets such as fire, smoke, reflections, and shadows in each frame of the original image or video, and to assign the correct category label (such as flame, smoke, interference item - reflection, etc.).

[0084] S303. Using incremental learning and transfer learning strategies, the original sample database is input into a predetermined initial fire identification model for iterative training to obtain a pre-constructed fire identification model.

[0085] Incremental learning is a training strategy that allows a model to continuously learn new data and categories without forgetting what it has already learned, without having to retrain the entire model from scratch. Transfer learning is another strategy that uses a mature model pre-trained on a large general dataset (such as ImageNet) as a starting point, and then fine-tunes the model with specific data (i.e., labeled samples) for the specific application scenario, so that it can quickly adapt to new tasks.

[0086] In some implementations, during model training, transfer learning is first employed. A powerful pre-trained visual model is selected and fine-tuned using S302-annotated samples to quickly enable it to recognize fire and smoke features in photovoltaic scenarios, forming a basic model. Then, after model deployment, an incremental learning strategy is adopted. When new samples are collected on-site that the model misidentifies (such as a new type of reflective interference or a special smoke pattern), the new samples are simply annotated, and the existing model is incrementally trained. The model can then learn new knowledge without discarding previous training results, achieving continuous iteration and self-evolution of the model.

[0087] As can be seen, S301-S303 constructs the original sample database by collecting data from moving and static devices in all scenarios. This ensures that the training set of the model can fully cover various complex working conditions that may be encountered in actual applications. It uses incremental learning and transfer learning strategies for iterative training, enabling the model to quickly acquire basic capabilities based on relevant technologies and achieve continuous high-precision recognition in a multi-source heterogeneous sensing system.

[0088] In some embodiments, step S102 may further include SC1-SC5, as follows: SC1. Locate the fire point in the digital twin model to obtain location information. Specifically, based on fire information, locate the fire point in the pre-created digital twin model of the target photovoltaic power station to obtain the fire point location information.

[0089] Here, the fire point refers to the specific location of the fire identified by the fire identification model in sensor data (such as video images); positioning refers to the process of converting the fire point from a two-dimensional sensor data coordinate system to a three-dimensional physical world geographic coordinate system; the location information not only includes latitude, longitude, and altitude three-dimensional coordinates, but may also include the specific equipment asset number corresponding to the coordinates.

[0090] In some implementations, the fire point information output by the fire identification model (e.g., pixel coordinates (1024, 768) from the video stream of camera C-05) is used as input; the precise three-dimensional spatial position, orientation, focal length, and other intrinsic and extrinsic parameters of camera C-05 are pre-stored in the digital twin model; using the pre-stored intrinsic and extrinsic parameters and the three-dimensional geometric model of the power station built into the model, the precise three-dimensional coordinates of the pixel in the real physical space are calculated through inverse projection or triangulation algorithms, and associated with the specific equipment object where the coordinate point is located, thereby generating highly accurate location information with clear semantics.

[0091] SC2. Call the built-in geographic information system to obtain the spatial topology of the fire point; specifically, based on the fire point location information, combined with the built-in geographic information system in the pre-created digital twin model of the target photovoltaic power station, the spatial topology of the fire point is obtained.

[0092] The built-in Geographic Information System (GIS) refers to a layer or module integrated within the digital twin model. It not only stores the geographical locations of each device, but more importantly, it stores the physical, electrical, and functional connections between them. Spatial topology refers to adjacency and connectivity relationships that go beyond simple distance. For example, device A and device B are physically adjacent, combiner box C and inverter D belong to the same circuit electrically, and cable trench E runs through area F, etc.

[0093] In some implementations, after SC1 determines the precise location of the fire and the associated device, a query is performed in the GIS module centered on that device. For example, if the fire is located on a combiner box, the GIS query will return a series of relational information: 1) Physical adjacency: a list of photovoltaic modules directly connected to the combiner box; 2) Electrical connection: the DC cable to which the combiner box belongs and the inverter number connected upstream; 3) Functional association: the range of power generation units affected by the combiner box; 4) Environmental adjacency: the location of the nearest fire hydrant, the type of vegetation below, etc. The above information together constitutes the spatial topology network of the fire.

[0094] SC3. Based on the spatial topological relationship and the wind direction and speed data of the fire point, determine the development trend; specifically, based on the spatial topological relationship of the fire point and combined with the wind direction and speed data of the fire point, obtain the fire point development trend information.

[0095] Among them, wind direction and wind speed data refer to environmental parameters obtained in real time from micro-weather stations deployed in photovoltaic power plants; in some embodiments, the data detected by the detection device closest to the fire point in the photovoltaic power plant is regarded as the wind direction and wind speed data of the fire point.

[0096] In some implementations, step SC3 is performed based on the first two steps; specifically, the spatial topology obtained in SC2 is fused and analyzed with real-time meteorological data, that is, the main spread direction is determined by wind direction and the spread size is determined by wind speed.

[0097] SC4. The fire level is determined based on the proportion of fire information in the power plant monitoring data stream.

[0098] Among them, the proportion in the data stream is a quantitative indicator, which can refer to the proportion of the area occupied by flame pixels in the entire video screen, or the area of ​​the overheated area in the thermal image, or the multiple of the smoke concentration sensor reading relative to the alarm threshold, etc.; the fire level is a standardized classification, such as classifying it into Level 1 (initial), Level 2 (developing), and Level 3 (intense) according to industry standards, with each level corresponding to a different emergency response level.

[0099] In some implementations, once a fire is identified, the raw data stream that triggered the alarm is simultaneously subjected to quantitative analysis. For example, the fire identification model not only outputs the conclusion that a fire exists, but also the pixel area of ​​the flame region. This area value is compared with a preset level classification threshold: if the area is less than 1000 pixels, it is classified as a Level 1 fire; if it is between 1000 and 5000 pixels, it is classified as a Level 2 fire; and if it is greater than 5000 pixels, it is classified as a Level 3 fire. This level is ultimately output as a key field of the structured fire situation information.

[0100] SC5 outputs fire location information, fire severity level, and fire development trend information to obtain structured fire situation information.

[0101] It is evident that locating the fire point in the digital twin model endows the location information with precise three-dimensional semantics. Next, the spatial topological relationship of the fire point is obtained by calling GIS, and the potential chain reaction caused by the fire is revealed by analyzing the physical and electrical connections between devices. Then, the development trend is determined by combining dynamic data such as wind direction and wind speed, which is a prediction based on spatial analysis and superimposed with the time dimension. Finally, the fire level is determined according to the data proportion, providing an objective quantitative standard for the severity of the fire.

[0102] In some implementations, after step S104, steps S301-S302 are further included, as follows: After the S301 active firefighting drone arrives above the target area, it performs a secondary scan and locks onto the location of the fire source; and then executes the fire source attack command.

[0103] The target area refers to the initial fire point coordinates generated by the main control system based on S102, which have a certain error range; secondary scanning and locking refers to the process by which the UAV uses its own sensors (such as thermal imaging pods) to conduct close-range, high-precision autonomous search in the target area, and accurately identify and track the core of the fire source in real time.

[0104] In some implementations, once the drone arrives at the target area based on GPS coordinates issued by the main control system, it does not immediately drop bombs. Instead, it activates terminal guidance mode, hovering at an altitude of tens of meters while using its high-resolution thermal imaging camera to scan the area below, searching for the hottest hotspots. Once the core of the fire source is identified, the onboard AI target tracking algorithm locks onto it. Even if the fire point moves slightly or the drone is disturbed by wind, the locking system can adjust in real time to ensure that the target is always aimed at the most critical position of the fire source. Only after confirming a stable lock does the drone execute the command to drop fire extinguishing bombs.

[0105] S302. After executing the fire source attack command, real-time detection data is obtained and transmitted back. The real-time detection data is used to determine whether additional fire extinguishing operations are needed.

[0106] Among them, real-time detection data feedback means that the drone will not return immediately after dropping bombs, but will continue to hover at the scene for a period of time, continuously transmitting the post-attack effect images captured by the airborne camera back to the main control system in real time; supplementary firefighting operations refer to deciding whether to carry out a second or more rounds of firefighting operations based on the assessment of the attack effect, such as dropping a second fire extinguishing bomb or dispatching another drone.

[0107] In some implementations, after the UAV completes its bombing in S1041, it enters an effect assessment mode; it receives and analyzes the real-time images transmitted back by the UAV, and determines whether the temperature has dropped significantly by comparing the thermal images before and after the attack; it determines whether the open flames have been extinguished and whether the smoke has turned white (water vapor) by analyzing the visible light video; if it is determined that the fire has been completely extinguished, the UAV is instructed to return to base; however, if it is found that the temperature is still very high or there are signs of reignition, it will be determined that additional firefighting operations are needed, and a new instruction will be generated and issued immediately to direct the UAV to conduct a second attack or call for support.

[0108] It is evident that this mechanism of immediate post-strike assessment transforms firefighting operations from a one-off, open-loop process into a closed-loop process based on feedback, ensuring that every firefighting resource is used to its maximum effect and reducing the risk of fire reignition due to insufficient assessment.

[0109] A problem exists in practical use: the generation and execution of linkage control commands are not synchronized, causing the commands to be unable to best match the current fire situation; at the same time, these commands are not generated based on real-time on-site data, inevitably leading to deviations; however, if commands are generated and executed synchronously based on images, fire extinguishing may be delayed, which deviates from the original design intent; therefore, in some optimized implementations, the following design is adopted: The pre-created digital twin model of the target photovoltaic power station includes a mirror model and a projection model; the mirror model is a digital twin model that keeps data synchronized with the target photovoltaic power station, and the projection model is a projection model of the mirror model after accelerated time-stream processing.

[0110] The structured fire situation information includes first structured fire situation information and second structured fire situation information; wherein, fire information is used as a perturbation variable and a mirror model is used to generate the first structured fire situation information; fire information is used as a perturbation variable and a projection model is used to generate the second structured fire situation information.

[0111] Based on the first structured fire situation information, a preset emergency response plan is matched and activated, and a first linkage control command is generated in combination with the real-time environmental parameters of the target photovoltaic power station; based on the second structured fire situation information, a preset emergency response plan is matched and activated, and a second linkage control command is generated in combination with the predicted environmental parameters of the target photovoltaic power station.

[0112] A first linkage control command is sent to one or more equipment execution units. During the execution of the first linkage control command by the equipment execution units, the multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and mobile inspection equipment group deployed within the target photovoltaic power station are re-aggregated to obtain a power station monitoring update data stream. Based on the power station monitoring update data stream, updated first structured fire situation information is obtained. The updated first structured fire situation information is compared with second structured fire situation information to obtain a state deviation value. When the state deviation value is less than a preset state deviation threshold, a second linkage control command is sent to one or more equipment execution units. When the state deviation value is not less than the preset state deviation threshold, based on the updated first structured fire situation information, a preset emergency response plan is matched and activated, and combined with the real-time environmental parameters of the target photovoltaic power station, a third linkage control command is generated. The third linkage control command is sent to one or more equipment execution units to switch the equipment execution units from executing the first linkage control command to executing the third linkage control command.

[0113] At this point, process S102 specifically includes: SD1. Use fire information as a perturbation variable; use a digital twin model to generate no less than two structured fire situation information, including a first structured fire situation information mirror that keeps data synchronized with the photovoltaic power station, and a second structured fire situation information extrapolated by accelerating the time flow. Among them, the disturbance variable refers to an input factor introduced into the stable system model to observe changes in the system state, specifically referring to the identified fire situation; the mirror body refers to an operating mode of the digital twin model, in which all models, parameters and states are completely synchronized with the physical photovoltaic power station in real time, faithfully reflecting the current real state of the physical world; the first structured fire situation information is intelligence generated by the mirror body that describes the precise current state of the fire scene; the projection body is another operating mode of the digital twin model, which, based on the mirror body, uses algorithms (such as accelerated thermodynamics and fluid dynamics simulations) to fast forward the time flow at several times or even hundreds of times the speed to simulate and predict the future development and evolution of the fire situation; the second structured fire situation information is predictive intelligence generated by the projection body that describes the possible state of the fire scene at a future point in time (such as 5 minutes later).

[0114] In some implementations, after a fire is identified, step SD1 is performed as a more refined situation generation method; the fire situation (such as the ignition point and initial scale) is input into the digital twin model as initial conditions; at this time, the digital twin model will split into two and run in parallel: one as a mirror body, continues to receive real-time sensor data, accurately reproduces the changes of the current fire scene every second, and generates the first structured fire situation information representing the present; the other as a projection body, performs high-speed simulation calculations in virtual space based on physical laws and environmental parameters (such as wind speed and combustible material distribution), quickly projects the spread path and scale changes of the fire in the next few minutes or even tens of minutes, and generates the second structured fire situation information representing the future.

[0115] This step SD1 differs from the single situation generation method in previous embodiments. Instead, it overcomes the technical barrier of traditional emergency decision-making being entirely based on current information and lacking foresight by constructing a dual virtual entity of mirror and simulation.

[0116] Step S103 specifically includes: SD2. Based on the first structured fire situation information, match and activate the preset emergency response plan, and generate the first linkage control command in combination with real-time environmental parameters; Among them, the first linkage control command refers to a tactical command that is specifically designed to respond to the current real-time fire situation and needs to be executed immediately.

[0117] In some implementations, steps SD2 and SD3 are executed in parallel or sequentially, focusing on addressing the immediate problem. Specifically, the system receives first structured fire situation information generated by the mirror image, reflecting the actual situation of the current fire scene. Based on the first structured fire situation information, the system quickly matches and activates the most suitable emergency response plan from the preset emergency response plans. For example, based on the precise location of the current fire point and the level II fire severity, the system activates the targeted elimination plan. Subsequently, the system fine-tunes the plan instructions by combining the current real-time environmental parameters such as wind speed and direction to ensure the accuracy of the attack. Finally, a set of first linkage control instructions aimed at immediately extinguishing the currently visible fire source is generated and prepared for execution.

[0118] SD3. Based on the second structured fire situation information, match and activate the preset emergency response plan, and generate the second linkage control command in combination with the predicted environmental parameters.

[0119] Among them, the predicted environmental parameters refer to the simulated predicted values ​​of environmental parameters (such as changes in wind direction and temperature rise) by the system model in the fire simulation process; the second linkage control command refers to the strategic command that is formulated in advance to deal with possible changes in the fire situation (such as fire spread and the emergence of new ignition points). It may not be executed immediately, but is used as a backup or follow-up command.

[0120] It is evident that the digital twin model constructs two parallel digital twins: a mirror image and a projection image. The mirror image, synchronized with the photovoltaic power station, generates a first structured fire situation information that accurately reflects the current state of the fire, and based on this, generates a first linkage control command for immediate execution, solving the immediate problem of what to do. Meanwhile, the projection image, which accelerates the timeline, simulates the development of the fire over a period of time, generating a second structured fire situation information that provides a predictive view, and based on this, generates a second linkage control command that looks to the future. The parallel generation of these two types of situation information and commands ensures that the system's decision-making is no longer a linear consideration at a single point in time. The mirror image ensures the realistic basis for action, while the projection image provides the foresight to act, enabling both precise suppression of the current fire and advance planning for future fires.

[0121] Step S104 specifically includes: SD4. Send the first linkage control command to one or more device execution units; Specifically, after generating the first linkage control command for the current real-time fire situation, it is immediately distributed to the corresponding equipment execution units via the communication network. For example, the command might dispatch an active firefighting drone to the fire site for the first strike. The purpose of this step SD4 is to respond quickly, to intervene in and control the fire situation as quickly as possible, and to buy time for subsequent assessment and adjustments.

[0122] SD5, the aggregation device execution unit executes the first linkage control command to update the multi-source heterogeneous sensing data, and obtains the power plant monitoring update data stream.

[0123] Among them, updating multi-source heterogeneous sensing data refers to the dynamic data collected by the sensors on the equipment during and after the execution of instructions, which reflects the latest changes in the fire scene; the power plant monitoring update data stream is a real-time data set that aggregates the latest data.

[0124] Specifically, drones carrying out firefighting missions transmit real-time video footage of the target area via their onboard cameras; fixed ground cameras also continuously monitor changes in the fire scene; the latest information data, including the effectiveness of firefighting and whether the fire is under control or expanding, is then aggregated to form a power plant monitoring update data stream that reflects the effectiveness of intervention measures. This power plant monitoring update data stream is the direct basis for assessing the effectiveness of the first directive.

[0125] SD6, and based on the power plant monitoring update data stream, combined with the digital twin model, update the first structured fire situation information to obtain the updated first structured fire situation information.

[0126] Specifically, the obtained power plant monitoring update data stream is input into the mirror body of the digital twin model; after the mirror body absorbs the latest real data from the power plant monitoring update data stream, it will immediately refresh its internal state, thereby updating the first structured fire situation information that reflects the latest and most realistic situation at the fire site.

[0127] SD7. Compare the updated first structured fire situation information with the second structured fire situation information at the corresponding time, and calculate the state deviation value. Among them, corresponding time refers to aligning real-world time with the accelerated elapsed time in the simulation to find a comparable point in time; state deviation value is a quantitative indicator used to measure the degree of difference between the actual development of the fire and the prediction of the model.

[0128] Specifically, assuming the simulation entity predicts the fire situation at time T0+5 minutes (second structured fire situation information); when the actual time reaches T0+5 minutes, the real fire situation updated by the mirror entity at this time, i.e. the updated first structured fire situation information, is compared with the simulation entity's original prediction in all aspects; the comparison can include multiple dimensions such as fire area, spread direction, and temperature peak, and finally a comprehensive state deviation value is obtained through weighted calculation.

[0129] SD8. When the state deviation value is less than the preset threshold, a second linkage control command is sent to one or more designated device execution units, causing the device execution unit to switch from the first linkage control command to the second linkage control command.

[0130] Specifically, if the calculated state deviation value is very small (e.g., less than 10%), it indicates that the actual development of the fire is entirely within the model's prediction, and the effect of the first round of attacks is also in line with expectations. This proves that the second linkage control command generated based on the simulation, such as intercepting the fire line that is about to spread, is correct and forward-looking. Therefore, the pre-generated second command is issued to the equipment execution unit, enabling the equipment execution unit to seamlessly switch from the current fire extinguishing task to the next fire control task, and always maintain strategic leadership over the fire.

[0131] SD9. When the state deviation value is not less than the preset threshold, based on the updated first structured fire situation information, the preset emergency response plan is matched and activated, and combined with real-time environmental parameters, a third linkage control command is generated. This third linkage control command is a completely new emergency command generated entirely based on the latest actual situation, used to replace the original plan.

[0132] Specifically, if the state deviation value is large (e.g., greater than 30%), it indicates that an unexpected change has occurred on site (which may be due to a sudden change in wind direction causing a change in the direction of fire spread, or the fire extinguishing effect being far less than expected). This means that the original prediction has failed, and the second linkage control command generated based on the prediction is no longer applicable. At this time, the second linkage control command should be abandoned decisively, and the process of "matching the plan - combining environmental parameters - generating the command" should be repeated based on the most realistic fire scene report, which is "updated first structured fire situation information", so as to generate a brand-new third linkage control command that is more suitable for the current emergency situation.

[0133] SD10: Send the third linkage control command to one or more designated device execution units; causing the device execution unit to switch from the first linkage control command to the third linkage control command.

[0134] As can be seen, by executing the first linkage control command to deal with the current fire situation, the updated real situation is compared with the predicted situation at the corresponding time, and the state deviation value is calculated. When the state deviation value is less than the threshold, it proves that the predicted path is accurate, and then the system decisively switches to the pre-planned second linkage control command. When the deviation value exceeds the threshold, it means that there has been an unexpected change in the fire scene, so the original plan is immediately abandoned, and a brand-new third linkage control command is generated based on the latest real situation, so as to maintain speed and accuracy in the uncertain fire scene environment.

[0135] The fire linkage control method for photovoltaic power plants described in Embodiment 1 provides a comprehensive data foundation for fire identification by aggregating multi-source heterogeneous sensing data from fixed and mobile equipment groups. Based on this, the fire identification model is invoked for feature comparison, enabling accurate identification of fire features from complex backgrounds. After fire identification, a structured fire situation information including location, level, and trend is generated by combining a digital twin model, and linkage control commands are generated and issued to equipment execution units such as drones. The entire process forms a seamless closed loop, changing the current situation of disconnect between alarm and fire protection and improving emergency response efficiency.

[0136] In this embodiment 1, a digital twin model is used to construct two parallel digital twins: a mirror twin and a projection twin. The mirror twin, which is synchronized with the photovoltaic power station, generates a first structured fire situation information that accurately reflects the current real state of the fire, and based on this, generates a first linkage control command for immediate execution, solving the problem of what to do in the present. At the same time, the projection twin, which accelerates the time flow, simulates the development of the fire over a period of time, generates a second structured fire situation information that predicts the future, and based on this, generates a second linkage control command that looks to the future. The parallel generation of the two types of situation information and commands makes the system's decision-making no longer a linear thinking at a single point in time. The mirror twin ensures the realistic basis of the action, while the projection twin gives the action foresight; it can not only accurately extinguish the current fire, but also plan for future fires in advance.

[0137] In this embodiment 1, the system responds to the current fire situation by executing the first linkage control command. The updated real-time situation is compared with the predicted situation at the corresponding time, and a state deviation value is calculated. When the deviation value is less than a threshold, it proves the predicted path is accurate, and the system decisively switches to the pre-planned second linkage control command. However, when the deviation value exceeds the threshold, it indicates an unexpected change in the fire situation, and the system immediately abandons the original plan and generates a new third linkage control command based on the latest real-time situation. This ensures both speed and accuracy in the face of uncertain fire conditions.

[0138] Example 2 As attached Figure 6 As shown in the figure, this embodiment 2 provides a fire linkage control system for photovoltaic power plants, including a data stream aggregation module, a fire situation identification module, a control command generation module, and a command sending and execution module.

[0139] The data stream aggregation module is used to aggregate multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups deployed in the target photovoltaic power station to obtain the power station monitoring data stream.

[0140] The fire situation identification module is used to extract features from the power plant monitoring data stream using a pre-built fire identification model and compare them with preset fire features to obtain fire information. Based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power plant, structured fire situation information is generated.

[0141] The control command generation module is used to match and activate the preset emergency response plan based on the structured fire situation information, and generate linkage control commands in combination with the real-time or predicted environmental parameters of the target photovoltaic power station.

[0142] The instruction sending and execution module is used to send linkage control instructions to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

[0143] Example 3 As attached Figure 7 As shown, this embodiment 3 provides an electronic device, including: a memory for storing a computer program; a processor for executing the computer program to implement the steps of a fire linkage control method for photovoltaic power plants; or, the processor executing the computer program to implement the functions of each module in the above-mentioned fire linkage control system for photovoltaic power plants.

[0144] For example, the computer program may be divided into one or more modules / units, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules / units may be a series of computer program instruction segments capable of performing a preset function, the instruction segments describing the execution process of the computer program in the electronic device.

[0145] The electronic device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The electronic device may include, but is not limited to, a processor and memory. Those skilled in the art will understand that the above are examples of electronic devices and do not constitute a limitation on the electronic device. It may include more components than described above, or combine certain components, or different components. For example, the electronic device may also include input / output devices, network access devices, buses, etc.

[0146] The processor can be a central processing unit, or other general-purpose processors, digital signal processors, application-specific integrated circuits, off-the-shelf programmable gate arrays or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor, or any conventional processor, etc. The processor is the control center of the electronic device, connecting various parts of the entire electronic device through various interfaces and lines.

[0147] The memory can be used to store the computer program and / or module. The processor implements various functions of the electronic device by running or executing the computer program and / or module stored in the memory and by calling the data stored in the memory.

[0148] The memory may primarily include a program storage area and a data storage area. The program storage area may store the operating system and at least one application program required for a given function (such as sound playback, image playback, etc.). The data storage area may store data created based on the use of the mobile phone (such as audio data, phonebook, etc.). Furthermore, the memory may include high-speed random access memory and non-volatile memory, such as hard disks, RAM, plug-in hard disks, smart memory cards, secure digital cards, flash memory cards, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0149] Example 4 This embodiment 4 also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the fire linkage control method for photovoltaic power plants.

[0150] If the modules / units integrated in the fire linkage control system for photovoltaic power plants are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium.

[0151] Based on this understanding, the present invention can implement all or part of the processes in the above-mentioned fire linkage control method for photovoltaic power plants, or it can be accomplished by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above-mentioned fire linkage control method for photovoltaic power plants. The computer program includes computer program code, which can be in the form of source code, object code, executable file, or preset intermediate form, etc.

[0152] The computer-readable storage medium may include any entity or device capable of carrying the computer program code, recording media, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory, random access memory, electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0153] Example 5 This embodiment 5 provides a computer product, which includes a computer program stored in a computer-readable storage medium. The processor of the electronic device reads the computer program from the computer-readable storage medium and executes the computer program, so that the electronic device can execute the fire linkage control method for photovoltaic power plants described in embodiment 1, which will not be described again here.

[0154] It should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when the program is executed, it can include the processes of the embodiments of the above methods.

[0155] The fire linkage control method for photovoltaic power plants described in this invention obtains a data stream by aggregating multi-source heterogeneous sensing data collected from fixed monitoring equipment groups and mobile inspection equipment groups deployed within the photovoltaic power plant; it then uses a fire identification model to extract features from the data stream and compares the extracted features with preset fire characteristics to identify the fire; combined with spatial information from a digital twin model, it generates structured fire situation information including location information, fire level, and development trend; based on the structured fire situation information, it matches and activates a preset emergency response plan, and generates linkage control commands based on real-time environmental parameters; finally, it issues the linkage control commands to one or more designated equipment execution units, which then execute the commands. The unit comprises an active firefighting drone and a fixed monitoring equipment group. The active firefighting drone is used to strike the fire source, while the fixed monitoring equipment group is used for illumination. By adopting the above technical solution, firstly, by aggregating multi-source heterogeneous sensing data from fixed and mobile equipment groups, a comprehensive data foundation is provided for fire identification. On this basis, a fire identification model is invoked for feature comparison to achieve accurate identification of fire characteristics from complex backgrounds. After identifying the fire, a structured fire situation information containing location, level, and trend is generated by combining a digital twin model. Linkage control commands are generated and issued to the drone and other execution units. The entire process forms a connected closed loop, changing the current situation of disconnect between alarm and firefighting and improving emergency response efficiency.

[0156] In this invention, a digital twin model is used to construct two parallel digital twins: a mirror twin and a projection twin. The mirror twin, synchronized with the photovoltaic power station, generates a first structured fire situation information that accurately reflects the current state of the fire, and based on this, generates a first linkage control command for immediate execution, solving the immediate problem of what to do. Simultaneously, the projection twin, with its accelerated timeline, simulates the fire's development over a future period, generating a second structured fire situation information that anticipates future events, and based on this, generates a second linkage control command focused on the future. This parallel generation of these two types of situation information and commands ensures that system decision-making is no longer a linear consideration at a single point in time. The mirror twin provides a realistic basis for action, while the projection twin provides foresight. This allows for both precise suppression of the current fire and proactive planning for future fires.

[0157] In this invention, the step of issuing a linkage control command to one or more designated device execution units specifically includes: sending a first linkage control command to one or more designated device execution units; aggregating the updated multi-source heterogeneous sensing data in the first linkage control command executed by the device execution units to obtain an updated data stream; updating the first structured fire situation information based on the updated data stream and a digital twin model; comparing the updated first structured fire situation information with the corresponding second structured fire situation information to calculate a state deviation value; and when the state deviation value is less than a preset threshold, sending a second linkage control command to one or more designated device execution units. The system executes the equipment execution unit; switches the equipment execution unit from the first linkage control command to the second linkage control command; when the state deviation value is not less than the preset threshold, it matches and activates the preset emergency response plan according to the updated first structured fire situation information, and generates a third linkage control command based on real-time environmental parameters; sends the third linkage control command to one or more designated equipment execution units; switches the equipment execution unit from the first linkage control command to the third linkage control command; responds to the current fire situation by executing the first linkage control command, compares the updated real situation with the predicted situation at the corresponding time, and calculates the state deviation value. When the deviation value is less than the threshold, it proves that the predicted path is accurate, and the system decisively switches to the pre-planned second linkage control command; when the deviation value exceeds the threshold, it means that there has been an unexpected change in the fire scene, and the system immediately abandons the original plan and generates a new third linkage control command based on the latest real situation. This ensures that the system maintains speed and accuracy in the uncertain fire scene environment.

[0158] In this invention, a fixed monitoring equipment group performs scanning monitoring. During the scanning monitoring process, when abnormal temperature data is detected, the fixed monitoring equipment group reports the abnormal signal and coordinate information. A mobile inspection equipment group is dispatched to fly to the abnormal point to obtain multi-source heterogeneous sensing data, thus constructing a workflow that combines wide-area initial screening with close-range detailed investigation. This interaction of fixed reporting coordinates and mobile precise follow-up forms an efficient closed loop of discovery and confirmation, solving the problems of blind searching and time delay in related inspection modes, and improving the efficiency of early fire hazard discovery and confirmation.

[0159] In this invention, mobile inspection equipment and fixed monitoring equipment are used to collect full-scene images and video data of a photovoltaic power station under different conditions, including simulated fires under normal operating conditions, to obtain an original sample database. The original sample database is output to annotate the full-scene images and video data. Incremental learning and transfer learning strategies are used to input the original sample database into the fire identification model for iterative training. By collecting data from moving and static equipment in the full scene to construct the original sample database, the data source ensures that the model's training set can comprehensively cover various complex operating conditions that may be encountered in actual applications. The incremental learning and transfer learning strategies are used for iterative training, enabling the model to quickly acquire basic capabilities based on relevant technologies, and achieving continuous high-precision identification in a multi-source heterogeneous sensing system.

[0160] In this invention, the location information of the fire point is obtained by locating it in a digital twin model; the spatial topology of the fire point is obtained by calling the built-in geographic information system; the development trend is determined based on the spatial topology and wind direction and speed data of the fire point; the fire severity level is determined based on the proportion of the fire in the data stream; by locating the fire point in the digital twin model, precise three-dimensional semantics of the location information are given; the spatial topology of the fire point is obtained by calling GIS, and the potential chain reactions caused by the fire are revealed by analyzing the physical and electrical connections between devices; then, the development trend is determined by combining dynamic data such as wind direction and speed, which is a prediction with a time dimension superimposed on the spatial analysis. Finally, the fire severity level is determined based on the data proportion, providing an objective quantitative standard for the severity of the fire.

[0161] In this invention, after the active firefighting drone arrives above the target area, it performs a secondary scan and locks onto the fire source location; then it executes a fire source strike command; after executing the fire source strike command, it acquires and transmits real-time detection data, which is used to determine whether supplementary firefighting operations are needed; this mechanism of immediate evaluation after strike transforms the firefighting operation from a one-off open-loop operation into a closed-loop process based on effect feedback. This ensures that every firefighting resource is used to its maximum effect and reduces the risk of fire reignition due to insufficient assessment.

[0162] The above embodiments are merely one of the implementation methods for achieving the technical solution of the present invention. The scope of protection claimed by the present invention is not limited to this embodiment, but also includes any variations, substitutions and other implementation methods that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention.

Claims

1. A fire linkage control method for photovoltaic power plants, characterized in that, include: The multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station are aggregated to obtain the power station monitoring data stream; Using a pre-built fire identification model, features are extracted from the power plant monitoring data stream and compared with preset fire features to obtain fire information; based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power plant, structured fire situation information is generated. Based on the structured fire situation information, the preset emergency response plan is matched and activated, and combined with the real-time or predicted environmental parameters of the target photovoltaic power station, a linkage control command is generated. Send linkage control commands to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

2. The fire linkage control method for photovoltaic power plants according to claim 1, characterized in that, Structured fire situation information includes fire location information, fire severity level, and fire development trend information; The process of generating structured fire situation information based on fire information and a pre-created digital twin model of the target photovoltaic power station is as follows: Based on fire information, the fire point is located in the pre-created digital twin model of the target photovoltaic power station to obtain the fire point location information; The fire severity level is determined based on the proportion of fire information in the power plant monitoring data stream; Based on the fire location information, and combined with the geographic information system built into the digital twin model of the pre-created target photovoltaic power station, the spatial topological relationship of the fire point is obtained. Based on the spatial topological relationship of the fire points and combined with the wind direction and speed data of the fire points, information on the development trend of the fire points is obtained. Output fire location information, fire severity level, and fire development trend information to obtain structured fire situation information.

3. The fire linkage control method for photovoltaic power plants according to claim 1, characterized in that, The pre-created digital twin model of the target photovoltaic power station includes a mirror model and a projection model; Among them, the mirror model is a digital twin model that keeps data synchronized with the target photovoltaic power station, and the derivation model is a derivation model of the mirror model after accelerated time stream processing; Structured fire situation information includes first structured fire situation information and second structured fire situation information; Specifically, fire information is used as a perturbation variable, and a mirror model is used to generate the first structured fire situation information; fire information is used as a perturbation variable, and a deductive model is used to generate the second structured fire situation information.

4. A fire linkage control method for photovoltaic power plants according to claim 3, characterized in that, The process of matching and activating a preset emergency response plan based on structured fire situation information, and generating linkage control commands by combining the real-time or preset environmental parameters of the target photovoltaic power station, is as follows: Based on the first structured fire situation information, the preset emergency response plan is matched and activated, and the first linkage control command is generated in combination with the real-time environmental parameters of the target photovoltaic power station. Based on the second structured fire situation information, the preset emergency response plan is matched and activated, and combined with the predicted environmental parameters of the target photovoltaic power station, a second linkage control command is generated.

5. A fire linkage control method for photovoltaic power plants according to claim 4, characterized in that, The process of sending linkage control commands to one or more device execution units is as follows: Send the first linkage control command to one or more device execution units; During the execution of the first linkage control command by the equipment execution unit, the multi-source heterogeneous sensing data collected by the fixed monitoring equipment group and the mobile inspection equipment group deployed in the target photovoltaic power station are re-aggregated to obtain the power station monitoring update data stream; Based on the power plant monitoring update data stream, the updated first structured fire situation information is obtained; The updated first structured fire situation information is compared with the second structured fire situation information to obtain the state deviation value. When the state deviation value is less than the preset state deviation threshold, a second linkage control command is sent to one or more device execution units; When the state deviation value is not less than the preset state deviation threshold, the preset emergency response plan is matched and activated based on the updated first structured fire situation information, and the third linkage control command is generated in combination with the real-time environmental parameters of the target photovoltaic power station. Send a third linkage control command to one or more device execution units to switch the device execution units from executing the first linkage control command to executing the third linkage control command.

6. A fire linkage control method for photovoltaic power plants according to claim 1, characterized in that, Before obtaining the power plant monitoring data stream, the process of aggregating multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups deployed within the target photovoltaic power plant also includes: Scanning monitoring is performed using a fixed monitoring equipment group deployed within the target photovoltaic power station to obtain preliminary scanning monitoring data; Based on the pre-scan detection data, determine whether there are abnormal temperature areas within the target photovoltaic power station, and determine the coordinate information of the abnormal temperature areas; If there is an abnormal temperature area within the target photovoltaic power station, the mobile inspection equipment within the target photovoltaic power station will be dispatched to the abnormal temperature area based on the coordinate information of the abnormal temperature area to begin multi-source heterogeneous sensing data collection.

7. A fire linkage control system for photovoltaic power plants, characterized in that, include: The data stream aggregation module is used to aggregate multi-source heterogeneous sensing data collected by fixed monitoring equipment groups and mobile inspection equipment groups deployed in the target photovoltaic power station to obtain the power station monitoring data stream; The fire situation identification module is used to extract features from the power plant monitoring data stream using a pre-built fire identification model and compare the features with preset fire features to obtain fire information; based on the fire information, combined with a pre-created digital twin model of the target photovoltaic power plant, structured fire situation information is generated. The control command generation module is used to match and activate the preset emergency response plan based on the structured fire situation information, and generate linkage control commands in combination with the real-time or predicted environmental parameters of the target photovoltaic power station. The instruction sending and execution module is used to send linkage control instructions to one or more equipment execution units; wherein, the equipment execution units include active fire-fighting drones and fixed monitoring equipment groups.

8. An electronic device, characterized in that, include: A processor is used to execute computer programs; A computer-readable storage medium storing a computer program, which, when executed by the processor, performs the fire linkage control method for a photovoltaic power station as described in any one of claims 1-6.

9. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the fire linkage control method for photovoltaic power plants as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the fire linkage control method for photovoltaic power plants as described in any one of claims 1-6.

Images