Fire extinguishing system of water area photovoltaic power station
Through multi-sensor collaborative detection and accurate fire route planning, the problems of fault monitoring and fire rescue in water photovoltaic power plants have been solved, and accurate monitoring and efficient handling of water photovoltaic power plants faults have been achieved, which has improved the fire safety level and emergency response capabilities.
Patent Information
- Application Number
- CN202510756383.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-06
- Publication Date
- 2025-08-08
AI Technical Summary
The existing photovoltaic power station fire protection systems are difficult to accurately capture early failure signals in water environments, and cannot cope with the different needs of combustion of electrical and non-electrical equipment. The fire protection route planning does not fully consider the priority of fault points and rescue paths, resulting in inefficient rescue efficiency, resulting in economic losses and ecological environment damage.
Multi-sensor collaborative detection is adopted to obtain arc, cable temperature and multi-spectral data. By weighted summing the priority and route distance between the fault point, scientifically plan the fire route, and execute gas fire extinguishing or fine water mist fire extinguishing instructions based on the fault point attributes to achieve accurate monitoring and efficient disposal.
Accurate monitoring, hierarchical early warning and efficient handling of water photovoltaic power station faults has been achieved, fire safety level and emergency response capabilities have been improved, false alarms and excessive responses have been reduced, response time has been shortened, and fire identification in complex water environments have been adapted.
Smart Images
Figure CN120437527A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of fire protection of photovoltaic power stations, and in particular to a fire protection system of a photovoltaic power station in a water area. Background Art
[0002] Currently, existing PV power plant firefighting technologies are primarily designed for land-based scenarios and lack specific design considerations for the specific characteristics of aquatic environments and electrical faults. For example, conventional temperature monitoring methods struggle to accurately detect early fault signals in complex aquatic environments, and a single firefighting method cannot address the diverse needs of electrical and non-electrical equipment fires.
[0003] Furthermore, traditional firefighting route planning approaches fail to fully consider fault point priority and rescue path optimization, leading to inefficient rescue efforts and the inability to promptly control fires, resulting in significant economic losses and ecological damage. Therefore, developing an intelligent, precise firefighting system suitable for water-based photovoltaic power plants is urgently needed to ensure their safe and stable operation. Summary of the Invention
[0004] The purpose of this invention is to separately collect arc and cable temperature data, as well as multispectral data, and pre-process it into data packets. It also prioritizes fault points and scientifically plans firefighting routes by weighted summation of fault point priorities and distances to the primary fault point. Furthermore, it precisely executes gas or water mist extinguishing commands based on the fault point's attributes. This firefighting system enables precise monitoring, graded early warning, and efficient handling of faults in photovoltaic power plants located in waters, significantly improving the power plant's fire safety and emergency response capabilities.
[0005] In order to achieve the above-mentioned object, the present invention provides a fire protection system for a water photovoltaic power station, comprising: The monitoring unit arranges first-type monitoring nodes and second-type monitoring nodes based on the water photovoltaic power station; obtains arc data and cable temperature data based on the first-type monitoring nodes; and obtains multispectral data based on the second-type monitoring nodes; The control unit obtains the fault point location information and the first-level warning level of the current fault point based on the first-type monitoring node, and combines the verification data obtained by the second-type monitoring node to modify the first-level warning level of the current fault point to generate a second-level warning level; Execution unit: Generates multi-level firefighting instructions based on the second-level warning level.
[0006] In some embodiments of the present invention, the monitoring unit is further configured to: Arrange arc detection sensors and distributed temperature optical fibers based on the first type of monitoring nodes; Obtain arc data and cable temperature data at all first-category monitoring nodes; The obtained arc data and cable temperature data are pre-processed to generate the arc data packet D1 of the water photovoltaic power station, D1={d 11 ,d 12 …d 1i …d 1n} and cable temperature data packet D2, D2={d 21 ,d 22 …d 2i …d 2n}; Arrange multispectral flame detectors based on the second type of monitoring nodes to obtain multispectral data; Preprocess the multispectral data to generate a multispectral data packet D3, D3={d 31 ,d 32 …d 3i …d 3m}; Among them, d 1i is the arc data reference value at the i-th first-class monitoring node, d 2i is the reference value of cable temperature data at the i-th first-class monitoring node, d 3i is the multispectral data reference value at the i-th second-category monitoring node, n is the number of first-category monitoring nodes, and m is the number of second-category monitoring nodes.
[0007] In some embodiments of the present invention, obtaining the fault point location information and the first-level warning level of the current fault point further includes: Grouping the first type of monitoring nodes to generate multiple groups of location monitoring points; Set the monitoring period of the location monitoring point based on historical data; Based on the monitoring period, each group of position monitoring points is monitored in turn to obtain the arc data packet D1 and the cable temperature data packet D2 of each group of position monitoring points; Build a fault prediction model; The fault area is generated by combining the arc data packet D1, cable temperature data packet D2 and warning level prediction model of each group of location monitoring points; Determine whether a fault occurs in the first type of monitoring node in the fault area based on the obtained arc data packet D1 and cable temperature data packet D2 of the first type of monitoring node in the fault area; Determine the fault location information and the first-level warning level of the current fault point based on the first-class monitoring nodes in the fault area; In some embodiments of the present invention, when constructing a first-level warning level prediction model, the following is also included: Obtain historical monitoring data of the first type of monitoring nodes; The historical monitoring data includes the warning information of the first type of monitoring nodes and the corresponding arc data reference values and cable temperature data reference values; Classifying the first-level warning levels of the first-category monitoring nodes based on the warning information to generate multiple first-level warning levels; And obtain the arc data reference value da and cable temperature data reference value db corresponding to each first-level warning level multiple times; Generate sample data of the current first-level warning level based on the arc data reference value da and the cable temperature data reference value db corresponding to the current first-level warning level; Construct a first-level warning level prediction model based on all first-level warning level sample data: p1=wa*da+wb*db; p1 is the predicted value of the first-level warning level; wa is the weight of the arc data reference value da, and wb is the weight of the cable temperature data reference value db; Generate corresponding level intervals of first-level warning level prediction values based on multiple first-level warning levels; Each level interval corresponds to a first-level warning level.
[0008] In some embodiments of the present invention, when generating a fault area, the process further includes: Obtain arc data packet D1 and cable temperature data packet D2 of the current group position monitoring point; Determine the first-level warning level of all monitoring points in the current group based on the first-level warning level prediction model; Set the judgment window, each judgment window includes n1 position monitoring points, n1 <n; The judgment window is translated by moving one position monitoring point at a time until all position monitoring points of the current group are traversed; Generate a first-level warning level change value for the current judgment window based on the first-level warning level difference between the first-position monitoring point and the last-position monitoring point in the current judgment window; Setting a first preset value of the first level warning level change value based on historical first level warning level change values; If the first-level warning level change value of the position monitoring point in the current judgment window is greater than the first preset value, the current judgment window area is a fault area.
[0009] In some embodiments of the present invention, when determining the fault point location information and the first-level warning level of the current fault point, the following steps are further included: Obtain the first-level warning level of all first-class monitoring nodes in the fault area; Based on the first-level warning levels of all first-level monitoring nodes, the first-level warning levels are compared to select the first-level monitoring node with the highest first-level warning level; The location information of the first-class monitoring node with the highest first-class warning level is obtained to generate the fault point location information.
[0010] In some embodiments of the present invention, when generating the second level warning level, the following steps are further included: Build a burning point identification model based on the acquired historical multispectral data; Combine the multispectral data package D3 and the burning point identification model to obtain the entire burning prediction area; Determine whether the current combustion prediction area includes the generated fault point location information. If it does, do not modify the first warning level to generate the second warning level. If not included, the first-level warning level of the current fault point will be lowered to generate the second-level warning level.
[0011] In some embodiments of the present invention, when constructing the combustion point recognition model, the following steps are further included: Generate infrared wave data and ultraviolet wave data based on the acquired historical multispectral data packets; Combining infrared wave data and ultraviolet wave data to generate flame characteristics; Obtain historical combustion states and corresponding flame characteristics to construct a combustion sample set; The combustion point recognition model is constructed by combining all combustion sample sets.
[0012] In some embodiments of the present invention, the execution unit is further configured to: Generate a fire map based on the secondary warning level of each fault point of the photovoltaic power station in the current water area; The fire map sets the processing priority of each fault point and generates the priority ranking of the fault points; Generate fire routes based on the priority ranking of fault points; Execute firefighting instructions on fault points in sequence based on firefighting strategies; Monitor the attributes of fault points on the fire route and select corresponding fire instructions based on the attributes of the fault points; If the fault point is a burning electrical device, a gas fire extinguishing instruction is generated; If the fault point is not caused by burning electrical equipment, a fine water mist fire extinguishing instruction will be generated.
[0013] In some embodiments of the present invention, the generating of the fire fighting route further includes: Based on the priority sorting of the fault points, the fault point with the highest priority is selected to generate the first fault point; Obtain the route distance L from the remaining fault points to the first fault point; The processing priority of the fault point and the route distance L from the remaining fault points to the first fault point are weighted and summed to generate the processing order value of the fault point; The firefighting route is generated by sorting the fault points based on their processing order values.
[0014] Compared with the prior art, the fire protection system for a photovoltaic power station in a water area provided by the embodiment of the present invention has the following advantages: Multi-dimensional fire monitoring and high-precision positioning based on multi-sensor collaborative detection; real-time capture of electrical fault characteristics (such as arcing, cable overheating) to achieve early warning of electrical fires.
[0015] The second type of monitoring nodes analyzes flame characteristics through infrared / ultraviolet bands, generates multi-spectral data packets, verifies the existence of open flames, and reduces false alarms.
[0016] Through group monitoring, sliding window algorithm and comparison of warning level change values (Δp1), abnormal areas (such as local overheating of cables or arc concentration areas) can be quickly identified.
[0017] By combining the fault prediction model and the highest warning level node positioning, the coordinates of the fault point can be accurately determined, and the error rate is significantly reduced.
[0018] The linear model based on electrical parameters (arc, temperature) divides the level intervals and preliminarily assesses the risk.
[0019] Multi-spectral verification is introduced. If no flame characteristics are detected (such as false alarms or non-open flame failures), the warning level is automatically lowered to avoid excessive response. If combustion is confirmed, the warning level is maintained or upgraded to improve reliability.
[0020] The burning point recognition model is trained with historical flame characteristics to enhance its ability to identify fires in complex water environments (such as water surface reflections and fog).
[0021] Generate the optimal path based on fire map priority and weighted distance to shorten response time.
[0022] Supports parallel processing logic for multiple fault points, prioritizes high-risk points, and adapts to water environments: multi-spectral anti-interference design and waterproof sensor deployment are suitable for harsh conditions such as high humidity and salt spray on the water surface. BRIEF DESCRIPTION OF THE DRAWINGS
[0023] Figure 1 This is a structural diagram of a fire protection system of a water photovoltaic power station provided by an embodiment of the present invention. DETAILED DESCRIPTION
[0024] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following embodiments are used to illustrate the present invention but are not intended to limit the scope of the present invention.
[0025] In the description of the present invention, it should be understood that the terms "center", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating the orientation or position relationship, are based on the orientation or position relationship shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention.
[0026] The terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the quantity of the technical features being referred to. Thus, a feature specified as "first" or "second" may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise specified, "plurality" means two or more.
[0027] In the description of the present invention, it should be noted that, unless otherwise expressly specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense. For example, they may refer to fixed, detachable, or integral connections; mechanical or electrical connections; direct or indirect connections through an intermediate medium; and internal communication between two components. Those skilled in the art will understand the specific meanings of the above terms in the present invention based on the specific circumstances.
[0028] Example 1: The embodiment of the present invention provides a fire protection system for a photovoltaic power station in a water area, such as Figure 1 As shown, including: The monitoring unit arranges first-type monitoring nodes and second-type monitoring nodes based on the water photovoltaic power station; obtains arc data and cable temperature data based on the first-type monitoring nodes; and obtains multispectral data based on the second-type monitoring nodes; The control unit obtains the fault point location information and the first-level warning level of the current fault point based on the first-type monitoring node, and combines the verification data obtained by the second-type monitoring node to modify the first-level warning level of the current fault point to generate a second-level warning level; Execution unit: Generates multi-level firefighting instructions based on the second-level warning level.
[0029] In embodiment 2, the monitoring unit is further configured to: Arrange arc detection sensors and distributed temperature optical fibers based on the first type of monitoring nodes; Obtain arc data and cable temperature data at all first-category monitoring nodes; The obtained arc data and cable temperature data are pre-processed to generate the arc data packet D1 of the water photovoltaic power station, D1={d11 ,d 12 …d 1i …d 1n} and cable temperature data packet D2, D2={d 21 ,d 22 …d 2i …d 2n}; Arrange multispectral flame detectors based on the second type of monitoring nodes to obtain multispectral data; Preprocess the multispectral data to generate a multispectral data packet D3, D3={d 31 ,d 32 …d 3i …d 3m}; Among them, d 1i is the arc data reference value at the i-th first-class monitoring node, d 2i is the reference value of cable temperature data at the i-th first-class monitoring node, d 3i is the multispectral data reference value at the i-th second-category monitoring node, n is the number of first-category monitoring nodes, and m is the number of second-category monitoring nodes.
[0030] In this embodiment, the multi-spectral flame detector simultaneously monitors spectral signals in multiple wavelength ranges, typically covering infrared, visible light, and ultraviolet bands. By comparing the signal strength ratios of different bands, it can distinguish between real flames and interference sources. This includes spectral intensity, wavelength characteristics, flicker frequency, spatial distribution, etc. It is mainly deployed in high-risk fire areas such as inverter stations, distribution boxes, and cable junctions. Arc data processing (D1) includes: real-time acquisition of arc characteristic parameters; filtering out environmental interference and background noise; identifying arc characteristic parameters such as peak value, frequency, duration, etc. to generate standardized d1i reference values; Cable temperature data processing (D2) includes: real-time monitoring of cable temperature distribution along the cable; correction based on ambient temperature influences; locating temperature anomalies and gradient changes; and generating standardized d2i temperature reference values. Multispectral data processing (D3) includes: synchronous acquisition of signals of different wavelengths; analysis of the proportional relationship of signal intensities in each band; feature extraction: extraction of flame characteristic parameters to generate standardized D3i multispectral reference values; In-depth analysis of data packet structure: D1={d 11 ,d 12 ...d 1i ...d 1n}: Each d 1i Is a vector containing multiple arc characteristic parameters, such as {intensity, frequency, waveform characteristics, duration} D2={d21 ,d 22 ...d 2i ...d 2n}: Each d 2i Contains parameters such as {maximum temperature, average temperature, temperature gradient, temperature change rate} D3={d 31 ,d 32 ...d 3i ...d 3m}: Each d 3i Contains multi-dimensional parameters such as {infrared signal intensity, visible light signal, ultraviolet signal, ratio of each band} This comprehensive monitoring system design enables water photovoltaic power stations to achieve multi-level, all-round safety risk monitoring, providing basic data support for subsequent anomaly detection, fault warning and automatic fire protection systems, and improving system safety and operation and maintenance efficiency.
[0031] In embodiment 3, when obtaining the fault point location information and the first-level warning level of the current fault point, the following steps are further included: Grouping the first type of monitoring nodes to generate multiple groups of location monitoring points; Set the monitoring period of the location monitoring point based on historical data; Based on the monitoring period, each group of position monitoring points is monitored in turn to obtain the arc data packet D1 and the cable temperature data packet D2 of each group of position monitoring points; Build a fault prediction model; The fault area is generated by combining the arc data packet D1, cable temperature data packet D2 and warning level prediction model of each group of location monitoring points; Determine whether a fault occurs in the first type of monitoring node in the fault area based on the obtained arc data packet D1 and cable temperature data packet D2 of the first type of monitoring node in the fault area; Determine the fault location information and the first-level warning level of the current fault point based on the first-class monitoring nodes in the fault area; In this embodiment, all the first-class monitoring nodes are sampled and grouped at equal intervals, and (a1, a 1+e 、a 1+2e …) as a group, (a2, a 2+e 、a 2+2e ...) is a group of ...; e is the sampling interval, which should be adjusted according to actual needs.
[0032] Analyze historical fault frequency distribution, shorten monitoring cycles during high-risk periods, establish equipment aging curves, increase monitoring frequency during the aging phase, adjust monitoring cycles based on meteorological data (e.g., shorten cycles during thunderstorms), and dynamically adjust cycles based on load factor, temperature change rate, etc. In Example 4, when constructing a first-level warning level prediction model, the following steps are also included: Obtain historical monitoring data of the first type of monitoring nodes; The historical monitoring data includes the warning information of the first type of monitoring nodes and the corresponding arc data reference values and cable temperature data reference values; Classifying the first-level warning levels of the first-category monitoring nodes based on the warning information to generate multiple first-level warning levels; And obtain the arc data reference value da and cable temperature data reference value db corresponding to each first-level warning level multiple times; Generate sample data of the current first-level warning level based on the arc data reference value da and the cable temperature data reference value db corresponding to the current first-level warning level; Construct a first-level warning level prediction model based on all first-level warning level sample data: p1=wa*da+wb*db; p1 is the predicted value of the first-level warning level; wa is the weight of the arc data reference value da, and wb is the weight of the cable temperature data reference value db; Generate corresponding level intervals of first-level warning level prediction values based on multiple first-level warning levels; Each level interval corresponds to a first-level warning level.
[0033] In this embodiment, statistics such as mean, standard deviation, peak, skewness, and kurtosis are extracted; autocorrelation coefficients, trend components, periodic characteristics, and rates of change are calculated; spectral features such as main frequency components and frequency band energy distribution are extracted through FFT; relief, information gain, or model-based feature importance ranking is used; algorithms such as SMOTE are used to generate synthetic samples for minority class samples; random undersampling or cluster undersampling is performed on majority class samples; and different weights are assigned to different class samples by combining oversampling and undersampling to balance class influences. By fitting historical data, minimizing the square error between the predicted value and the actual level, iteratively adjusting the weights, minimizing the loss function, using k-fold cross validation to determine the optimal weight combination, introducing L1 or L2 regularization terms to avoid overfitting and improve generalization ability, and determining the weight wa of the weighted arc data reference value da and the weight wb of the cable temperature data reference value db.
[0034] In embodiment 5, when generating a fault area, the method further includes: Obtain arc data packet D1 and cable temperature data packet D2 of the current group position monitoring point; Determine the first-level warning level of all monitoring points in the current group based on the first-level warning level prediction model; Set the judgment window, each judgment window includes n1 position monitoring points, n1 <n; The judgment window is translated by moving one position monitoring point at a time until all position monitoring points of the current group are traversed; Generate a first-level warning level change value for the current judgment window based on the first-level warning level difference between the first-position monitoring point and the last-position monitoring point in the current judgment window; Setting a first preset value of the first level warning level change value based on historical first level warning level change values; If the first-level warning level change value of the position monitoring point in the current judgment window is greater than the first preset value, the current judgment window area is a fault area.
[0035] In this embodiment, the monitoring points are grouped: the cable loop of a photovoltaic power station in a certain water area is divided into one monitoring group, which includes 10 first-category monitoring nodes (n=10), numbered N1 to N10.
[0036] Determine the window size: Set the sliding window to contain 5 consecutive monitoring points (n1=5).
[0037] Historical data benchmark: Based on historical statistics, the first preset value of the first-level warning level change value (Δp1) is set to 2.0 (that is, a fault judgment is triggered when the difference in warning levels between the first and last nodes in the window exceeds 2.0).
[0038] If Δp1>2.0, the current window area is marked as a fault candidate area.
[0039] Otherwise, if Δp1<2.0, ignore it and continue sliding the window.
[0040] In Example 6, when determining the fault point location information and the first-level warning level of the current fault point, the following is further included: Obtain the first-level warning level of all first-class monitoring nodes in the fault area; Based on the first-level warning levels of all first-level monitoring nodes, the first-level warning levels are compared to select the first-level monitoring node with the highest first-level warning level; The location information of the first-class monitoring node with the highest first-class warning level is obtained to generate the fault point location information.
[0041] In this embodiment, a data request command is sent to all first-category monitoring nodes within the fault area via a data communication network, such as Ethernet or a wireless network (Wi-Fi, 4G / 5G, etc.). Upon receiving the command, the monitoring node retrieves pre-calculated Level 1 warning level data from its own storage module. This data is calculated by the monitoring node based on real-time sensor data (such as temperature, pressure, current, and voltage) combined with a built-in warning algorithm model. The monitoring node encapsulates the Level 1 warning level data and transmits it back to the data processing center via the network in a specified data format (such as JSON or XML). The data processing center receives and stores this data for subsequent analysis.
[0042] After receiving the first-level warning level data for all first-category monitoring nodes, the data processing center uses a sorting algorithm (such as quick sort or merge sort) to sort this data in descending order. Taking quick sort as an example, a baseline value is first selected, and all data is divided into two parts: one larger than the baseline value and the other smaller than the baseline value. These two parts are then recursively sorted, ultimately resulting in a list arranged from high to low by first-level warning level. After sorting is complete, the first-category monitoring node corresponding to the first element in the list is the node with the highest first-level warning level. Alternatively, a screening algorithm can be used to set an initial maximum warning level and corresponding node, then iterate through all node data. If a higher warning level is found, the maximum warning level and corresponding node are updated. The node obtained after the traversal is the target node.
[0043] After identifying the first-category monitoring node with the highest Level 1 warning level, the data processing center sends a location information request command to that node. Upon receiving the command, the node obtains its current location coordinates from its own positioning module or retrieves location information from a pre-set location parameter table. This information is also packaged in the agreed-upon data format and transmitted back to the data processing center. The data processing center receives this location information and records it as the fault point location. It also determines the Level 1 warning level of the node as the current fault point's Level 1 warning level, creating a complete fault information record for subsequent fault handling and warning issuance.
[0044] In Example 7, when generating the second level warning level, the following steps are also included: Build a burning point identification model based on the acquired historical multispectral data; Combine the multispectral data package D3 and the burning point identification model to obtain the entire burning prediction area; Determine whether the current combustion prediction area includes the generated fault point location information. If it does, do not modify the first warning level to generate the second warning level. If not included, the first-level warning level of the current fault point will be lowered to generate the second-level warning level.
[0045] In this embodiment, historical multispectral data is retrieved from a database. This data covers spectral information from different time periods and regions, including data from multiple bands such as near-infrared and short-wave infrared. A deep learning algorithm (such as a convolutional neural network (CNN)) or a machine learning algorithm (such as a support vector machine (SVM)) is used as the model framework. The historical multispectral data is divided into a training set, a validation set, and a test set. The training set is used to optimize model parameters, the validation set assists in adjusting hyperparameters, and the test set evaluates the model's generalization ability. Through continuous iterative training, the model is able to accurately identify combustion characteristics contained in the multispectral data, thus constructing a combustion point recognition model.
[0046] Obtain the current multispectral data packet D3, which contains real-time multispectral imagery of the fault area. Input D3 into the established combustion point recognition model. The model analyzes each pixel or area in the data packet and, based on its trained combustion feature recognition capabilities, determines whether the area presents a combustion risk. Areas with combustion risk are marked, and all marked areas are integrated to obtain the total predicted combustion area.
[0047] The fault point location information is spatially matched with all predicted combustion areas. If the fault point location information is within a predicted combustion area, it indicates that the fault point has a high combustion risk. The current level 1 warning level is maintained and directly used as the level 2 warning level. If the fault point location information is not within any predicted combustion area, it indicates that the fault point has a low combustion risk. The current level 1 warning level of the fault point is lowered according to the preset rules (such as lowering one warning level). The lowered warning level is the level 2 warning level.
[0048] Example 8, when building a combustion point identification model, further includes: Generate infrared wave data and ultraviolet wave data based on the acquired historical multispectral data packets; Combining infrared wave data and ultraviolet wave data to generate flame characteristics; Obtain historical combustion states and corresponding flame characteristics to construct a combustion sample set; The combustion point recognition model is constructed by combining all combustion sample sets.
[0049] In this embodiment, a specific band is extracted from a historical multispectral data packet: Infrared wave data: Extracts the 3-5μm mid-infrared (MIR) and 8-14μm thermal infrared (TIR) bands, which are highly sensitive to high-temperature targets (such as flames); Ultraviolet wave data: Extract the 0.2-0.4μm ultraviolet band, especially the solar-blind ultraviolet (240-280nm) which is sensitive to free radical radiation such as CH and OH in flames; Spectral characteristics: Calculate the radiation intensity ratio of the infrared / ultraviolet band (such as MIR / TIR) and construct the flame feature vector; Spatiotemporal characteristics: Analyze the dynamic characteristics of infrared / ultraviolet hotspots in continuous frames, such as movement trajectory and area change rate; Texture features: Extract the gray-level co-occurrence matrix (GLCM) features of the hotspot area to describe the texture characteristics of the flame; Collect historical fire records and mark the corresponding multispectral data timestamps; A combination of expert annotation and automatic recognition is used to mark different combustion stages (smoldering, open flame, and extinguished); Establish a mapping relationship between the extracted flame feature vector and the marked combustion state Construct a set of sample pairs in the form of {flame feature vector → combustion state label}; Model framework selection: Use 3D convolutional neural network (CNN) to process spatiotemporal sequence data for multi-feature fusion classification; Divide into training set (70%), validation set (15%), and test set (15%); Batch normalization (BN) and Dropout techniques are used to prevent overfitting; Use Adam optimizer to minimize the cross entropy loss function; Adjust hyperparameters such as learning rate and BatchSize through the validation set.
[0050] In embodiment 9, the execution unit is further configured to: Generate a fire map based on the secondary warning level of each fault point of the photovoltaic power station in the current water area; The fire map sets the processing priority of each fault point and generates the priority ranking of the fault points; Generate fire routes based on the priority ranking of fault points; Execute firefighting instructions on fault points in sequence based on firefighting strategies; Monitor the attributes of fault points on the fire route and select corresponding fire instructions based on the attributes of the fault points; If the fault point is a burning electrical device, a gas fire extinguishing instruction is generated; If the fault point is not caused by burning electrical equipment, a fine water mist fire extinguishing instruction will be generated.
[0051] In this embodiment, the data of the secondary warning level, location coordinates, equipment type, etc. of each fault point are normalized and unified in dimension to facilitate subsequent analysis.
[0052] An R-Tree data structure is used to establish a spatial index of fault points, enabling efficient spatial queries and range retrieval. Fault points are mapped to thermal zones of varying colors (yellow, orange, and red) based on the warning level, visually displaying the distribution of fire risks. The precise location of each fault point is marked on the heat map, along with key information such as device type and warning level.
[0053] If the fault point is a burning electrical device, select the gas fire extinguishing strategy, calculate the required CO2 / HFC-227ea dosage based on the device volume, set the injection time to 60 seconds, and add pre-conditions such as "disconnect power" and "evacuate personnel." If the fault point is not the burning electrical equipment, choose the fine water mist fire extinguishing strategy, calculate the required water pressure (50-100 bar) based on the burning area, select the appropriate nozzle type (solid cone / hollow cone), set the spray time to 120 seconds, calculate the coverage area and plan the spray angle.
[0054] In embodiment 10, when generating a fire route, the method further includes: Based on the priority sorting of the fault points, the fault point with the highest priority is selected to generate the first fault point; Obtain the route distance L from the remaining fault points to the first fault point; The processing priority of the fault point and the route distance L from the remaining fault points to the first fault point are weighted and summed to generate the processing order value of the fault point; The firefighting route is generated by sorting the fault points based on their processing order values.
[0055] In this embodiment, core fault points requiring priority are screened from all fault points (e.g., those with the largest fires and the most trapped personnel). Priority scoring indicators (e.g., fire severity, casualties, property value, etc.) are assigned to each fault point, and a weighted calculation is performed to determine a comprehensive priority value. Priority values are sorted in descending order, and the fault point with the highest value is selected as the first fault point (denoted as F1).
[0056] Calculate route distance: The distance L between the remaining fault points and F1 quantifies the spatial correlation between each fault point and the core fault point and evaluates the scheduling cost.
[0057] Use geographic information system (GIS) or path planning algorithm (such as Dijkstra algorithm) to calculate the shortest route distance Li from the remaining fault points Fi (i=2,3,...,n) to F1.
[0058] Li may include factors such as actual road distance, travel time, or road complexity.
[0059] Formula definition: Si=α*Pi+(1−α)*Li where: Si: the processing order value of the fault point Fi (the smaller the value, the higher the priority, and the more it needs to be processed first); Pi: Normalized priority value of the fault point Fi (the larger the value, the higher the urgency); Li: normalized route distance from fault point Fi to F1 (the larger the value, the longer the distance); α: Weight coefficient (0≤α≤1), used to adjust the relative importance of priority and distance (for example, in a fire scenario, α=0.7 can be set to emphasize priority).
[0060] Based on the comprehensive evaluation results, the optimal driving path is planned to minimize the total processing time. All fault points Si are sorted in ascending order to obtain the processing order [F1, Fa, Fb, ..., Fn].
[0061] Combined with path planning algorithms (such as TSP traveling salesman problem optimization), the shortest firefighting route covering all fault points is generated.
[0062] Finally, it should be noted that it is apparent that those skilled in the art may make various modifications and variations to the present invention without departing from the spirit and scope of the present invention. Thus, to the extent such modifications and variations fall within the scope of the present invention and its equivalents, the present invention is intended to include such modifications and variations.
[0063] The above description is merely a preferred embodiment of the present invention and is not intended to limit the scope of protection of the present invention.
Claims
1. A fire protection system for a water photovoltaic power station, characterized in that: include: Monitoring units, based on the water photovoltaic power station, are respectively arranged with the first type of monitoring nodes and the second type of monitoring nodes; Acquire arc data and cable temperature data based on the first type of monitoring nodes; Acquire multispectral data based on the second type of monitoring nodes; The control unit obtains the fault point location information and the first-level warning level of the current fault point based on the first-type monitoring node, and combines the verification data obtained by the second-type monitoring node to modify the first-level warning level of the current fault point to generate a second-level warning level; Execution unit: Generates multi-level firefighting instructions based on the second-level warning level.
2. The fire protection system of the water photovoltaic power station according to claim 1, characterized in that: The monitoring unit is further configured to: Arrange arc detection sensors and distributed temperature optical fibers based on the first type of monitoring nodes; Obtain arc data and cable temperature data at all first-category monitoring nodes; The obtained arc data and cable temperature data are pre-processed to generate the arc data packet D1 of the water photovoltaic power station, D1={d 11 ,d 12 …d 1i …d 1n } and cable temperature data packet D2, D2={d 21 ,d 22 …d 2i …d 2n }; Arrange multispectral flame detectors based on the second type of monitoring nodes to obtain multispectral data; Preprocess the multispectral data to generate a multispectral data packet D3, D3={d 31 ,d 32 …d 3i …d 3m }; Among them, d 1i is the arc data reference value at the i-th first-class monitoring node, d 2i is the reference value of cable temperature data at the i-th first-class monitoring node, d 3i is the multispectral data reference value at the i-th second-category monitoring node, n is the number of first-category monitoring nodes, and m is the number of second-category monitoring nodes.
3. The fire protection system of the water photovoltaic power station according to claim 2, characterized in that: The acquisition of the fault point location information and the first-level warning level of the current fault point also includes: Grouping the first type of monitoring nodes to generate multiple groups of location monitoring points; Set the monitoring period of the location monitoring point based on historical data; Based on the monitoring period, each group of position monitoring points is monitored in turn to obtain the arc data packet D1 and the cable temperature data packet D2 of each group of position monitoring points; Build a fault prediction model; The fault area is generated by combining the arc data packet D1, cable temperature data packet D2 and warning level prediction model of each group of location monitoring points; Determine whether a fault occurs in the first type of monitoring node in the fault area based on the obtained arc data packet D1 and cable temperature data packet D2 of the first type of monitoring node in the fault area; The first-class monitoring nodes in the fault area determine the fault point location information and the first-level warning level of the current fault point.
4. The fire protection system of the water photovoltaic power station according to claim 3, characterized in that: When constructing the first-level warning level prediction model, it also includes: Obtain historical monitoring data of the first type of monitoring nodes; The historical monitoring data includes the warning information of the first type of monitoring nodes and the corresponding arc data reference values and cable temperature data reference values; Classifying the first-level warning levels of the first-category monitoring nodes based on the warning information to generate multiple first-level warning levels; And obtain the arc data reference value da and cable temperature data reference value db corresponding to each first-level warning level multiple times; Generate sample data of the current first-level warning level based on the arc data reference value da and the cable temperature data reference value db corresponding to the current first-level warning level; Construct a first-level warning level prediction model based on all first-level warning level sample data: p1=wa*da+wb*db; p1 is the predicted value of the first-level warning level; wa is the weight of the arc data reference value da, and wb is the weight of the cable temperature data reference value db; Generate corresponding level intervals of first-level warning level prediction values based on multiple first-level warning levels; Each level interval corresponds to a first-level warning level.
5. The fire protection system of the water photovoltaic power station according to claim 4, characterized in that: The generating of the fault area further includes: Obtain arc data packet D1 and cable temperature data packet D2 of the current group position monitoring point; Determine the first-level warning level of all monitoring points in the current group based on the first-level warning level prediction model; Set the judgment window, each judgment window includes n1 position monitoring points, n1 <n; The judgment window is translated by moving one position monitoring point at a time until all position monitoring points of the current group are traversed; Generate a first-level warning level change value for the current judgment window based on the first-level warning level difference between the first-position monitoring point and the last-position monitoring point in the current judgment window; Setting a first preset value of the first level warning level change value based on historical first level warning level change values; If the first-level warning level change value of the position monitoring point in the current judgment window is greater than the first preset value, the current judgment window area is a fault area.
6. The fire protection system of the water photovoltaic power station according to claim 5, characterized in that: The determination of the fault point location information and the first-level warning level of the current fault point also includes: Obtain the first-level warning level of all first-class monitoring nodes in the fault area; Based on the first-level warning levels of all first-level monitoring nodes, the first-level warning levels are compared to select the first-level monitoring node with the highest first-level warning level; The location information of the first-class monitoring node with the highest first-class warning level is obtained to generate the fault point location information.
7. The fire protection system of the water photovoltaic power station according to claim 6, characterized in that: When generating the second-level warning level, it also includes: Build a burning point identification model based on the acquired historical multispectral data; Combine the multispectral data package D3 and the burning point identification model to obtain the entire burning prediction area; Determine whether the current combustion prediction area includes the generated fault point location information. If it does, do not modify the first warning level to generate the second warning level. If not included, the first-level warning level of the current fault point will be lowered to generate the second-level warning level.
8. The fire protection system of the water photovoltaic power station according to claim 7, characterized in that: When constructing the combustion point identification model, it also includes: Generate infrared wave data and ultraviolet wave data based on the acquired historical multispectral data packets; Combining infrared wave data and ultraviolet wave data to generate flame characteristics; Obtain historical combustion states and corresponding flame characteristics to construct a combustion sample set; The combustion point recognition model is constructed by combining all combustion sample sets.
9. The fire protection system of the water photovoltaic power station according to claim 8, characterized in that: The execution unit is further configured to: Generate a fire map based on the secondary warning level of each fault point of the photovoltaic power station in the current water area; The fire map sets the processing priority of each fault point and generates the priority ranking of the fault points; Generate fire routes based on the priority ranking of fault points; Execute firefighting instructions on fault points in sequence based on firefighting strategies; Monitor the attributes of fault points on the fire route and select corresponding fire instructions based on the attributes of the fault points; If the fault point is a burning electrical device, a gas fire extinguishing instruction is generated; If the fault point is not caused by burning electrical equipment, a fine water mist fire extinguishing instruction will be generated.
10. The fire protection system of the water photovoltaic power station according to claim 9, characterized in that: The generation of the fire route also includes: Based on the priority sorting of the fault points, the fault point with the highest priority is selected to generate the first fault point; Obtain the route distance L from the remaining fault points to the first fault point; The processing priority of the fault point and the route distance L from the remaining fault points to the first fault point are weighted and summed to generate the processing order value of the fault point; The firefighting route is generated by sorting the fault points based on their processing order values.