A Wind Turbine Nacelle Fire Early Warning System Based on Image Recognition

Abstract

This application relates to the technical field of wind turbines, and particularly to a wind turbine nacelle fire warning system based on image recognition. It includes: a central control unit for establishing a smoke image model; a monitoring unit for setting multiple monitoring sub-areas, and monitoring sub-modules are arranged in the monitoring sub-areas; the monitoring unit is used to collect monitoring data in each monitoring sub-area, and the monitoring data includes equipment image data and environmental image data; a smoke unit for judging whether there is smoke inside the nacelle; multiple monitoring sub-areas are established according to the equipment parameters inside the wind turbine nacelle, and by monitoring in real time whether there is smoke inside the nacelle, when there is no smoke, potential fire risk sources are timely discovered according to the equipment state models in each monitoring sub-area, and early warning and maintenance are carried out. When there is smoke in the nacelle, the possible spreading area and speed of the smoke are predicted, so as to provide data support for fire fighting decisions and reduce the cost losses caused by fires.

Description

Technical Field

[0001] This application relates to the technical field of wind turbines, and particularly to a fire warning system for a wind turbine nacelle based on image recognition. Background Art

[0002] The main components such as the frequency converter, transformer, and controller of the wind turbine are all located inside the wind turbine nacelle. These components contain a large number of electrical components. Once high temperature, electrical short circuit, and arcing occur, it is extremely easy to cause a fire. At the same time, the hydraulic oil of the hydraulic station, gearbox of the wind turbine, lubricating oil of the oil pump station, nacelle cover, fairing, and blades all belong to combustibles. Once the electrical components of the components such as the frequency converter, transformer, and controller in the wind turbine nacelle catch fire, it is extremely easy to cause a fire in the entire nacelle and even the hub blades.

[0003] With the wide application of wind power generation technology, the fire safety of the wind turbine nacelle has become a crucial issue. There are many deficiencies in the existing wind turbine nacelle fire protection systems: ordinary smoke and temperature sensors are prone to false alarms or malfunctions in harsh environments such as large altitude temperature differences, high humidity, and strong wind and sand, and their reliability is relatively low. There is a lack of effective visual monitoring means, making it difficult to accurately locate the fire source position and the development of the fire situation at the initial stage of the fire, which is not conducive to taking targeted fire protection measures in a timely manner. Summary of the Invention

[0004] The purpose of this application is: To solve the above technical problems, this application provides a fire warning system for a wind turbine nacelle based on image recognition, aiming to improve the fire warning efficiency and processing speed for the inside of the wind turbine nacelle and ensure the safe operation of the wind turbine.

[0005] In some embodiments of this application, multiple monitoring sub - regions are established according to the equipment parameters inside the wind turbine nacelle, and by adding smoke units, it is monitored in real - time whether there is smoke inside the nacelle. When there is no smoke, according to the equipment state models in each established monitoring sub - region, all the equipment inside the nacelle is monitored to determine whether there are abnormal situations such as overheating, smoking, component displacement, or damage of the equipment, timely discover potential fire risk sources, and give early warnings for maintenance to ensure the safe operation of the nacelle.

[0006] In some embodiments of this application, when there is smoke inside the nacelle, the monitoring unit continuously collects environmental images in each monitoring sub - region, performs real - time analysis and processing on the collected image sequence. Through the comparative analysis of consecutive multi - frame images, the diffusion direction and speed of the smoke are calculated, and combined with the spatial layout information of the nacelle, a smoke movement trend map is generated to predict the possible spreading area and speed of the smoke, thereby providing data support for fire protection decision - making, improving the fire handling efficiency inside the nacelle, and reducing the cost losses caused by the fire.

[0007] In some embodiments of the present application, a wind turbine nacelle fire warning system based on image recognition is provided, including:

[0008] A central control unit for establishing a smoke image model;

[0009] A monitoring unit for setting a plurality of monitoring sub - regions, and monitoring sub - modules are arranged in the monitoring sub - regions;

[0010] The monitoring unit is used to collect monitoring data in each monitoring sub - region, and the monitoring data includes equipment image data and environmental image data;

[0011] A smoke unit for judging whether there is smoke inside the nacelle;

[0012] The central control unit includes:

[0013] A first processing module for setting the working parameters of the monitoring unit;

[0014] An equipment module for establishing an equipment status model for each monitoring sub - region and generating an operation anomaly value for each monitoring sub - region according to a preset monitoring time node;

[0015] A second processing module for generating a fire - fighting decision according to the smoke image model and all environmental image data.

[0016] In some embodiments of the present application, the first processing module is further used for:

[0017] Obtaining the real - time smoke status inside the nacelle;

[0018] If there is no smoke, setting the monitoring unit to the first - level working mode;

[0019] The first - level working mode includes:

[0020] Setting a plurality of monitoring time nodes;

[0021] Obtaining the equipment image data of each monitoring sub - region according to the preset monitoring time node;

[0022] If there is smoke, setting the monitoring unit to the second - level working mode;

[0023] The second - level working mode includes:

[0024] Setting the image acquisition parameters of the monitoring unit;

[0025] Collecting the environmental image data of each monitoring sub - region;

[0026] Generating a feedback data packet for each monitoring sub - region.

[0027] In some embodiments of the present application, the device module is further configured to:

[0028] Establish a sequence of monitored sub - regions A, A=(a 1 , a 2 … a i … a n ), where a i is the i - th monitored sub - region; n is the number of monitored sub - regions;

[0029] Set a i as the target monitored sub - region;

[0030] Obtain the device parameters of the target monitored sub - region and generate a training data packet based on the device parameters;

[0031] Establish a device status model for the target monitored sub - region according to the training data packet;

[0032] Establish device status models for each monitored sub - region in sequence;

[0033] Establish a sequence of device status models B, B=(b 1 , b 2 … b i … b n ), where bi is the device status model of the i - th monitored sub - region.

[0034] In some embodiments of the present application, the device module is further configured to:

[0035] Set the device status model of the target monitored sub - region as the target device status model;

[0036] Obtain the device image data of the target monitored sub - region at the current monitoring time node;

[0037] Pre - process the device image data based on the target device status model;

[0038] Generate an abnormal operation value f of the target monitored sub - region at the current monitoring time node according to the processing result;

[0039] f = µ i * j i ;

[0040] Where θ is the number of device metrics of the target monitored sub - region; µ i is the influence factor of the i - th device metric; j i is the deviation evaluation value of the i - th device metric of the target monitored sub - region at the current monitoring time node;

[0041] Generate abnormal operation values of each monitored sub - region at the current monitoring time node in sequence;

[0042] Establish an abnormal operation value sequence F, F = (f 1 , f 2 … f i … f n ), where f i is the abnormal operation value of the i-th monitoring sub-region at the current monitoring time node;

[0043] Judge whether to generate a warning instruction at the current monitoring time node according to the abnormal operation value sequence F.

[0044] In some embodiments of the present application, judging whether to generate a warning instruction at the current monitoring time node includes:

[0045] Generate an operation risk value c according to the abnormal operation value sequence F;

[0046] c = e1 * Q1 * f i + e2 * Q2 * Y(i) * (f i - f');

[0047] Where, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; f' is the abnormal operation value threshold; Y(i) is a selection coefficient. If (f i - f') > 0, Y(i) = 1 / (f i - f'); if (f i - f') < 0, Y(i) = 0;

[0048] Preset the first operation risk value threshold C1;

[0049] If c > C1, generate a warning instruction at the current monitoring time node;

[0050] If c < C1, do not generate a warning instruction at the current monitoring time node b.

[0051] In some embodiments of the present application, the second processing module is further configured to:

[0052] If the monitoring unit is in the secondary working mode, obtain the feedback data packets of each monitoring sub-region;

[0053] Set a i as the sub-region to be diagnosed according to the monitoring sub-region sequence A;

[0054] Establish an environmental image sequence D according to the feedback data packet of the sub-region to be diagnosed, D = (d 1 , d 2 … d i … d m), where d i is the i-th frame environmental image of the sub-region to be diagnosed collected based on the time series; m is the amount of environmental image collection;

[0055] Generate the smoke analysis result of the sub-region to be diagnosed according to the smoke image model and the environmental image sequence D;

[0056] Generate the smoke analysis results of each monitoring sub-region in sequence;

[0057] Fuse all the smoke analysis results and generate a smoke movement trend chart according to the fusion result;

[0058] Generate a fire fighting decision according to the smoke movement trend chart.

[0059] In some embodiments of the present application, when generating the smoke analysis result of the diagnostic sub-region, it includes:

[0060] Set d i as the target environmental image in sequence according to the environmental image sequence D;

[0061] Obtain the smoke contour map k1 in the target environmental image;

[0062] Obtain the smoke contour map k2 of the previous frame environmental image based on the time series of the target environmental image;

[0063] Generate the smoke movement parameters in the target environmental image according to the comparison result of the smoke contour map k1 and the smoke contour map k2;

[0064] Generate the smoke movement parameters of each environmental image in sequence;

[0065] Generate the smoke diffusion direction and the smoke diffusion speed sequence V of the sub-region to be diagnosed according to all the smoke movement parameters;

[0066] The smoke diffusion speed sequence V, V=(v 1 , v 2 … v i … v m ), where v i is the smoke diffusion speed in the i-th frame environmental image of the sub-region to be diagnosed;

[0067] Generate the initial fire value g of the sub-region to be diagnosed according to the smoke diffusion speed sequence V;

[0068] Generate the smoke analysis result of the sub-region to be diagnosed according to the initial fire value g and the smoke diffusion direction.

[0069] In some embodiments of the present application, generating the initial fire value g of the sub-region to be diagnosed includes:

[0070] g = e3 * Q3 * vi / m]+e4*Q4*U;

[0071] Among them, e3 is a preset third weight coefficient; e4 is a preset fourth weight coefficient; Q3 is a preset third fixed coefficient; Q4 is a preset fourth fixed coefficient; U is a first-level reference value generated based on the smoke diffusion speed sequence V.

[0072] In some embodiments of the present application, the second processing module is further configured to:

[0073] Establish an association model according to the position parameters of all monitoring sub-regions;

[0074] Obtain the smoke diffusion direction of each monitoring sub-region;

[0075] Generate a smoke movement trend diagram according to the association model and the smoke diffusion directions of all monitoring sub-regions;

[0076] Generate compensation coefficients for each monitoring sub-region according to the smoke movement trend diagram;

[0077] Establish a compensation coefficient sequence R, R = (r 1 , r 2 …r i …r n ), where r i is the compensation coefficient of the i-th monitoring sub-region;

[0078] Generate the first-level fire values for each monitoring sub-region according to the compensation coefficient sequence R;

[0079] Generate a fire fighting decision according to all the first-level fire values.

[0080] In some embodiments of the present application, when generating a fire fighting decision according to all the first-level fire values, it includes:

[0081] Establish a first-level fire value sequence H = (h 1 , h 2 …h i …h n ), where h i is the first-level fire value of the i-th monitoring sub-region;

[0082] h i = r i * g i， Among them, g i is the initial fire value of the i-th monitoring sub-region;

[0083] Set the working parameters of the fire fighting equipment in each monitoring sub-region in sequence according to the first-level fire value sequence H.

[0084] Compared with the prior art, the beneficial effects of the wind turbine nacelle fire warning system based on image recognition in the embodiments of the present application are as follows:

[0085] Multiple monitoring sub-regions are established according to the equipment parameters inside the wind turbine nacelle, and by adding smoke units, it is possible to monitor in real time whether there is smoke inside the nacelle. When there is no smoke, according to the equipment status models in each established monitoring sub-region, all the equipment inside the nacelle is monitored to determine whether there are abnormal situations such as overheating, smoking, component displacement or damage of the equipment, so as to timely discover potential fire risk sources and give early warnings for maintenance to ensure the safe operation of the nacelle.

[0086] When there is smoke inside the nacelle, the monitoring unit continuously collects environmental images in each monitoring sub-region, performs real-time analysis and processing on the collected image sequence, calculates the diffusion direction and speed of the smoke through the comparative analysis of consecutive multi-frame images, and combines the spatial layout information of the nacelle to generate a smoke movement trend diagram, predicting the possible spreading area and speed of the smoke, thereby providing data support for fire fighting decision-making, improving the fire handling efficiency inside the nacelle, and reducing the cost losses caused by the fire. Description of the Drawings

[0087] Figure 1 It is a schematic structural diagram of a wind turbine nacelle fire warning system based on image recognition in the preferred embodiment of the embodiments of the present application. Detailed Embodiments

[0088] The following will further describe in detail the specific embodiments of the present application in conjunction with the drawings and embodiments. The following embodiments are used to illustrate the present application, but are not used to limit the scope of the present application.

[0089] In the description of the present application, it should be understood that the orientation or positional relationship indicated by the terms "center", "upper", "lower", "front", "rear", "left", "right", "vertical", "horizontal", "top", "bottom", "inner", "outer", etc. is based on the orientation or positional relationship shown in the drawings, and is only for the convenience of describing the present application 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 should not be construed as a limitation to the present application.

[0090] The terms "first" and "second" are only used for descriptive purposes, and cannot be construed as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, the features defined with "first" and "second" may explicitly or implicitly include one or more of such features. In the description of the present application, unless otherwise stated, the meaning of "a plurality" is two or more.

[0091] In the description of the present application, it should be noted that unless otherwise clearly specified and limited, the terms "installation", "connection", and "coupling" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be directly connected or indirectly connected through an intermediate medium, and it can be the communication inside two components. For those of ordinary skill in the art, the specific meanings of the above terms in the present application can be understood according to specific situations.

[0092] As Figure 1 shown, a preferred embodiment of an image recognition-based wind turbine nacelle fire warning system according to an embodiment of the present application includes:

[0093] A central control unit for establishing a smoke image model;

[0094] A monitoring unit for setting a plurality of monitoring sub-regions, and monitoring sub-modules are arranged in the monitoring sub-regions;

[0095] The monitoring unit is used to collect monitoring data in each monitoring sub-region, and the monitoring data includes equipment image data and environmental image data;

[0096] A smoke unit for determining whether there is smoke inside the nacelle;

[0097] The central control unit includes:

[0098] A first processing module for setting the working parameters of the monitoring unit;

[0099] An equipment module for establishing an equipment status model for each monitoring sub-region and generating an operation anomaly value for each monitoring sub-region according to a preset monitoring time node;

[0100] A second processing module for generating a fire fighting decision according to the smoke image model and all environmental image data.

[0101] Specifically, the smoke unit is preferably a plurality of smoke sensors, and the smoke sensors are arranged in each monitoring sub-region for real-time monitoring of whether there is smoke inside the nacelle.

[0102] Specifically, the monitoring sub-module is preferably an infrared imaging device, which can collect image data and temperature data in each monitoring sub-region.

[0103] Specifically, a plurality of monitoring sub-regions are established according to the spatial layout parameters inside the nacelle and the position parameters of the main components such as the wind turbine frequency converter, transformer, and controller, and monitoring sub-modules and fire fighting equipment are arranged in each monitoring sub-region.

[0104] Specifically, the first processing module is further used for:

[0105] Obtain the real-time smoke status inside the cabin;

[0106] If there is no smoke, set the monitoring unit to the first-level working mode;

[0107] The first-level working mode includes:

[0108] Set multiple monitoring time nodes;

[0109] Obtain the device image data of each monitoring sub-region according to the preset monitoring time nodes;

[0110] If there is smoke, set the monitoring unit to the second-level working mode;

[0111] The second-level working mode includes:

[0112] Set the image acquisition parameters of the monitoring unit;

[0113] Collect the environmental image data of each monitoring sub-region;

[0114] Generate feedback data packets for each monitoring sub-region.

[0115] Specifically, the first-level working mode means that there is no smoke inside the current cabin. At this time, by monitoring the operating status of each device, potential fire hazards are warned and repaired in a timely manner to prevent fires inside the cabin and ensure the safe operation of the cabin.

[0116] Specifically, the second-level working mode means that there is smoke inside the current cabin. It is necessary to analyze each monitoring sub-region to generate the diffusion direction and speed of the smoke, as well as the area and speed where the smoke may spread, so as to formulate precise fire-fighting decisions and improve the fire handling efficiency inside the cabin.

[0117] In the preferred embodiment of this application, the device module is further used for:

[0118] Establish a monitoring sub-region sequence A, A=(a 1 ,a 2 …a i …a n ), where a i is the i-th monitoring sub-region; n is the number of monitoring sub-regions;

[0119] Set a i as the target monitoring sub-region according to the monitoring sub-region sequence A;

[0120] Obtain the device parameters of the target monitoring sub-region and generate a training data packet according to the device parameters;

[0121] Establish a device status model for the target monitoring sub-region according to the training data packet;

[0122] Establish the device status models for each monitoring sub-region in sequence;

[0123] Establish a sequence of device status models B, B = (b 1 , b 2 … b i … b n ), where bi is the device status model of the i-th monitoring sub-region.

[0124] Specifically, by analyzing the devices in the target monitoring sub-region, construct the appearance model of the devices in the normal operating state, thereby generating the corresponding device status model.

[0125] Specifically, the device module is also used for:

[0126] Set the device status model of the target monitoring sub-region as the target device status model;

[0127] Obtain the device image data of the target monitoring sub-region at the current monitoring time node;

[0128] Preprocess the device image data based on the target device status model;

[0129] Generate the abnormal operation value f of the target monitoring sub-region at the current monitoring time node according to the processing result;

[0130] f = µ i * j i ;

[0131] Where θ is the number of device indicators of the target monitoring sub-region; µ i is the influence factor of the i-th device indicator; j i is the deviation evaluation value of the i-th device indicator of the target monitoring sub-region at the current monitoring time node;

[0132] Generate the abnormal operation values of each monitoring sub-region at the current monitoring time node in sequence;

[0133] Establish a sequence of abnormal operation values F, F = (f 1 , f 2 … f i … f n ), where f i is the abnormal operation value of the i-th monitoring sub-region at the current monitoring time node;

[0134] Judge whether to generate a warning instruction at the current monitoring time node according to the sequence of abnormal operation values F.

[0135] Specifically, the device indicators include, but are not limited to, the device operating temperature and the device operating appearance. The greater the deviation from the evaluation value, the greater the deviation between the current operating state of the device and the normal operating state.

[0136] Specifically, the collected device image data includes conventional images and infrared images. By analyzing the conventional images, deviation evaluation values of each device indicator associated with the device appearance are generated. By analyzing the infrared images, deviation evaluation values of each device indicator related to the device operating temperature are generated.

[0137] Specifically, determining whether a warning instruction is generated at the current monitoring time node includes:

[0138] Generating an operation risk value c according to the abnormal operation value sequence F;

[0139] c = e1 * Q1 * f i + e2 * Q2 * Y(i) * (f i - f');

[0140] Wherein, e1 is a preset first weight coefficient; e2 is a preset second weight coefficient; Q1 is a preset first fixed coefficient; Q2 is a preset second fixed coefficient; f' is an abnormal operation value threshold; Y(i) is a selection coefficient. If (f i - f') > 0, Y(i) = 1 / (f i - f'); if (f i - f') < 0, Y(i) = 0;

[0141] Presetting a first operation risk value threshold C1;

[0142] If c > C1, a warning instruction is generated at the current monitoring time node;

[0143] If c < C1, no warning instruction is generated at the current monitoring time node b.

[0144] Specifically, all parameters in the model are normalized by presetting the first fixed coefficient and the second fixed coefficient, so that each parameter in the model is within the same value range.

[0145] Specifically, the abnormal operation value threshold can be set according to historical parameters. The greater the operation risk value, the greater the possibility of potential fire risks (such as local overheating caused by electrical equipment short - circuit, high temperature generated by mechanical component friction, etc.) inside the engine room at the current feedback time node.

[0146] Specifically, corresponding maintenance strategies are formulated according to the warning instructions, so as to timely eliminate potential fire risks inside the engine room.

[0147] It can be understood that in the above embodiments, multiple monitoring sub - regions are established according to the equipment parameters inside the nacelle of the wind turbine, and by adding smoke units, it is possible to monitor in real time whether there is smoke inside the nacelle. When there is no smoke, according to the equipment state models in each established monitoring sub - region, all the equipment inside the nacelle is monitored to determine whether there are abnormal situations such as overheating, smoking, component displacement or damage of the equipment, timely discover potential fire risk sources, and give early warnings for maintenance to ensure the safe operation of the nacelle.

[0148] In the preferred embodiment of the present application, the second processing module is further configured to:

[0149] When the monitoring unit is in the secondary working mode, obtain the feedback data packets of each monitoring sub - region;

[0150] Set a according to the monitoring sub - region sequence A i as the sub - region to be diagnosed;

[0151] Establish an environmental image sequence D according to the feedback data packet of the sub - region to be diagnosed, D=(d 1 , d 2 …d i …d m ), where d i is the i - th frame of the environmental image of the sub - region to be diagnosed collected based on the time series; m is the amount of environmental image collection;

[0152] Generate a smoke analysis result of the sub - region to be diagnosed according to the smoke image model and the environmental image sequence D;

[0153] Generate the smoke analysis results of each monitoring sub - region in turn;

[0154] Perform a fusion process on all the smoke analysis results, and generate a smoke movement trend chart according to the fusion result;

[0155] Generate a fire - fighting decision according to the smoke movement trend chart.

[0156] Specifically, when the smoke unit detects that there is smoke inside the nacelle, collect multiple frames of environmental images of each monitoring sub - region, generate a corresponding image sequence based on the time series, and determine the type of the combustible by extracting the color of the smoke (the colors of the smoke generated by the combustion of different substances are different), and calculate the diffusion direction and speed of the smoke through the comparative analysis of consecutive multiple frames of images.

[0157] Specifically, when generating the smoke analysis result of the diagnostic sub - region, it includes:

[0158] Set d i as the target environmental image in turn according to the environmental image sequence D;

[0159] Obtain the smoke contour map k1 in the target environment image;

[0160] Obtain the smoke contour map k2 of the previous frame of the environment image based on the time series of the target environment image;

[0161] Generate the smoke motion parameters in the target environment image according to the comparison result of the smoke contour map k1 and the smoke contour map k2;

[0162] Generate the smoke motion parameters of each environment image in sequence;

[0163] Generate the smoke diffusion direction and the smoke diffusion speed sequence V of the sub-region to be diagnosed according to all the smoke motion parameters;

[0164] The smoke diffusion speed sequence V, V = (v 1 , v 2 … v i … v m ), where v i is the smoke diffusion speed in the i-th frame of the environment image of the sub-region to be diagnosed;

[0165] Generate the initial fire value g of the sub-region to be diagnosed according to the smoke diffusion speed sequence V;

[0166] Generate the smoke analysis result of the sub-region to be diagnosed according to the initial fire value g and the smoke diffusion direction.

[0167] Specifically, if all the data in the smoke diffusion speed sequence V are 0, it indicates that there is no smoke in the current sub-region to be diagnosed, that is, the ignition point is not in the current sub-region to be diagnosed.

[0168] Specifically, generating the initial fire value g of the sub-region to be diagnosed includes:

[0169] g = e3 * Q3 * v i / m] + e4 * Q4 * U;

[0170] Among them, e3 is the preset third weight coefficient; e4 is the preset fourth weight coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; U is the first-level reference value generated based on the smoke diffusion speed sequence V.

[0171] Specifically, the first-level reference value is set according to the change state of the smoke diffusion speed. The larger the first-level reference value corresponding to the increasing state of the smoke diffusion speed, and the smaller the first-level reference value corresponding to the stable state or the decreasing state, that is, the larger the first-level reference value indicates that the current smoke diffusion speed is getting faster and faster.

[0172] Specifically, by presetting the third fixed coefficient and the fourth fixed coefficient, all parameters in the model are normalized, so that each parameter in the model is within the same value range.

[0173] Specifically, the larger the initial fire value is, the more serious the fire situation is in the corresponding monitored sub-region. It is necessary to increase the working efficiency of the corresponding fire-fighting equipment, timely control the spread of the fire, and reduce the overall fire loss.

[0174] It can be understood that in the above embodiments, when there is smoke in the cabin, the monitoring unit continuously collects the environmental images in each monitored sub-region, performs real-time analysis and processing on the collected image sequence, calculates the diffusion direction and speed of the smoke through the comparative analysis of consecutive multi-frame images, combines the spatial layout information of the cabin, generates a smoke movement trend map, predicts the possible spread area and speed of the smoke, so as to provide data support for fire-fighting decision-making.

[0175] In the preferred embodiment of the present application, the second processing module is further configured to:

[0176] Establish an association model according to the position parameters of all monitored sub-regions;

[0177] Obtain the smoke diffusion direction of each monitored sub-region;

[0178] Generate a smoke movement trend map according to the association model and the smoke diffusion direction of all monitored sub-regions;

[0179] Generate a compensation coefficient for each monitored sub-region according to the smoke movement trend map;

[0180] Establish a compensation coefficient sequence R, R = (r 1 , r 2 …r i …r n ), where r i is the compensation coefficient of the i-th monitored sub-region;

[0181] Generate a primary fire value for each monitored sub-region according to the compensation coefficient sequence R;

[0182] Generate a fire-fighting decision according to all primary fire values.

[0183] Specifically, according to whether there is an adjacent relationship between each monitored sub-region and the gas flow trend inside the cabin, a smoke diffusion model between each monitored sub-region is set, so as to generate a corresponding association model.

[0184] Specifically, according to the smoke movement trend map, it can be analyzed which monitored sub-regions the smoke may spread to, so as to set the corresponding compensation coefficient. The larger the compensation coefficient is, the greater the possibility that the smoke diffuses in the corresponding monitored sub-region.

[0185] Specifically, when generating a fire control decision based on all first-level fire values, it includes:

[0186] Establish a sequence of first-level fire values H=(h 1 ,h 2 …h i …h n ), where h i is the first-level fire value of the i-th monitoring sub-region;

[0187] h i =r i *g i， where g i is the initial fire value of the i-th monitoring sub-region;

[0188] Set the working parameters of the fire-fighting equipment in each monitoring sub-region in sequence according to the sequence of first-level fire values H.

[0189] Specifically, set a threshold for the first-level fire value. When h i is greater than the threshold of the first-level fire value, start the fire-fighting equipment in the i-th monitoring sub-region, and set the spraying speed of the fire-fighting equipment according to h i ; set the spraying angle of the fire-fighting equipment in the i-th monitoring sub-region according to the smoke movement trend diagram.

[0190] It can be understood that in the above embodiments, when it is determined that a fire has occurred, according to the first-level fire values of each monitoring sub-region, the corresponding fire-fighting equipment (such as dry powder fire extinguishers, gas fire extinguishing devices, etc.) is started, and the optimal spraying speed and spraying angle are determined, so as to improve the efficiency of fire handling in the engine room and reduce fire losses.

[0191] According to the first concept of the present application, multiple monitoring sub-regions are established based on the equipment parameters inside the wind turbine nacelle, and by adding smoke units, it is monitored in real time whether there is smoke inside the nacelle. When there is no smoke, according to the equipment state models in each established monitoring sub-region, all the equipment inside the nacelle is monitored to determine whether there are abnormal situations such as abnormal heating, smoking, component displacement or damage of the equipment, timely discover potential fire risk sources, and give early warning for maintenance to ensure the safe operation of the nacelle.

[0192] According to the second concept of the present application, when there is smoke in the nacelle, the monitoring unit continuously collects the environmental images in each monitoring sub-region, performs real-time analysis and processing on the collected image sequence, calculates the diffusion direction and speed of the smoke through the comparative analysis of consecutive multi-frame images, combines the spatial layout information of the nacelle, generates a smoke movement trend diagram, and predicts the possible spreading area and speed of the smoke, so as to provide data support for fire control decisions, improve the efficiency of fire handling inside the nacelle, and reduce the cost losses caused by fires.

[0193] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the technical principle of the present application, several improvements and substitutions can be made, and these improvements and substitutions should also be regarded as the protection scope of the present application.

Claims

1. A wind turbine cabin fire warning system based on image recognition, characterized in that: Including: A central control unit for establishing a smoke image model; A monitoring unit for setting a plurality of monitoring sub - regions, and monitoring sub - modules are arranged in the monitoring sub - regions; The monitoring unit is used to collect monitoring data in each monitoring sub - region, and the monitoring data includes equipment image data and environmental image data; A smoke unit for judging whether there is smoke inside the engine room; The central control unit includes: A first processing module for setting the working parameters of the monitoring unit; An equipment module for establishing the equipment state models of each monitoring sub - region and generating the abnormal operation values of each monitoring sub - region according to the preset monitoring time nodes; A second processing module for generating a fire - fighting decision according to the smoke image model and all environmental image data; The first processing module is further used for: Obtaining the real - time smoke state inside the engine room; If there is no smoke, setting the monitoring unit to the first - level working mode; The first - level working mode includes: Setting a plurality of monitoring time nodes; Obtaining the equipment image data of each monitoring sub - region according to the preset monitoring time nodes; If there is smoke, setting the monitoring unit to the second - level working mode; The second - level working mode includes: Setting the image acquisition parameters of the monitoring unit; Collecting the environmental image data of each monitoring sub - region; Generating feedback data packets for each monitoring sub - region; The equipment module is further used for: Establish a monitoring sub-area sequence A, A=(a1, a2…a i …a n ), where a i is the ith monitoring sub-area; n is the number of monitoring sub-areas; Set a according to the monitoring sub-area sequence A i Monitor sub-areas for the target; Obtaining the equipment parameters of the target monitoring sub - region and generating a training data packet according to the equipment parameters; Establishing the equipment state model of the target monitoring sub - region according to the training data packet; Sequentially establishing the equipment state models of each monitoring sub - region; Establish the equipment state model series B, B=(b1, b2…b i …b n ), where bi is the equipment status model of the i-th monitoring sub-area; The equipment module is further used for: Setting the equipment state model of the target monitoring sub - region as the target equipment state model; Obtaining the equipment image data of the target monitoring sub - region at the current monitoring time node; Pre - processing the equipment image data based on the target equipment state model; Generating the abnormal operation value f of the target monitoring sub - region at the current monitoring time node according to the processing result; f= µ i *j i ]; Among them, θ is the number of equipment indicators in the target monitoring sub-area; µ i is the influencing factor of the i-th equipment index; j i is the deviation evaluation value of the i-th equipment indicator in the target monitoring sub-area at the current monitoring time node; Sequentially generating the abnormal operation values of each monitoring sub - region at the current monitoring time node; Establish abnormal operation value series F, F=(f1,f2…f i …f n ), where f i is the abnormal operation value of the i-th monitoring sub-area at the current monitoring time node; Judging whether to generate a warning instruction at the current monitoring time node according to the abnormal operation value sequence F; Judging whether to generate a warning instruction at the current monitoring time node includes: Generating an operation risk value c according to the abnormal operation value sequence F; c=e1*Q1* f i ]+e2*Q2* Y(i)*(f i -f')]; Among them, e1 is the preset first weight coefficient; e2 is the preset second weight coefficient; Q1 is the preset first fixed coefficient; Q2 is the preset second fixed coefficient; f' is the abnormal operation value threshold; Y(i) is the selection coefficient, if (f i -f')>0,Y(i)=1 / (f i -f'); if (f i -f')<0,Y(i)=0; Presetting a first operation risk value threshold C1; If c > C1, generating a warning instruction at the current monitoring time node; If c < C1, not generating a warning instruction at the current monitoring time node b.

2. The wind turbine cabin fire warning system based on image recognition according to claim 1, characterized in that: The second processing module is further used for: When the monitoring unit is in the second - level working mode, obtaining the feedback data packets of each monitoring sub - region; Set a according to the monitoring sub-area sequence A i is the sub-region to be diagnosed; According to the feedback data packet of the sub-area to be diagnosed, the environmental image sequence D is established, D=(d1, d2…d i …d m ), where d i is the i-th frame of the environment image of the sub-region to be diagnosed acquired based on the time series; m is the amount of environment image acquisition; Generating the smoke analysis result of the sub - region to be diagnosed according to the smoke image model and the environmental image sequence D; Sequentially generating the smoke analysis results of each monitoring sub - region; Performing a fusion process on all the smoke analysis results and generating a smoke movement trend chart according to the fusion result; Generating a fire - fighting decision according to the smoke movement trend chart.

3. The wind turbine cabin fire warning system based on image recognition according to claim 2, characterized in that: When generating the smoke analysis result of the diagnostic sub - region, it includes: According to the environmental image sequence D, set d i is the target environment image; Obtaining the smoke contour map k1 in the target environmental image; Obtaining the smoke contour map k2 of the previous frame of the environmental image based on the time series of the target environmental image; Generate smoke motion parameters in the target environment image according to the comparison results of the smoke contour image k1 and the smoke contour image k2; Generate smoke motion parameters for each environment image in sequence; Generate a series V of smoke diffusion direction and smoke diffusion speed in the sub-region to be diagnosed according to all smoke motion parameters; Smoke diffusion speed series V, V = (v1, v2…v i …v m ), where v i is the smoke diffusion speed in the i-th frame of the environment image of the sub-area to be diagnosed; Generate the initial fire value g of the sub-area to be diagnosed according to the smoke diffusion speed series V; The smoke analysis results of the sub-area to be diagnosed are generated according to the initial fire value g and the smoke diffusion direction.

4. The wind turbine cabin fire warning system based on image recognition as claimed in claim 3, characterized in that: Generate the initial fire value g of the sub-area to be diagnosed, including: g=e3*Q3* v i / m]+e4*Q4*U; Among them, e3 is the preset third weight coefficient; e4 is the preset fourth weight coefficient; Q3 is the preset third fixed coefficient; Q4 is the preset fourth fixed coefficient; U is the primary reference value generated based on the smoke diffusion speed series V.

5. The wind turbine cabin fire warning system based on image recognition according to claim 4, characterized in that: The second processing module is also used for: Establish a correlation model based on the location parameters of all monitoring sub-areas; Obtain the smoke diffusion direction of each monitoring sub-area; Generate a smoke movement trend graph based on the correlation model and the smoke diffusion direction of all monitored sub-areas; Generate compensation coefficients for each monitoring sub-area based on the smoke movement trend graph; Establish compensation coefficient series R, R=(r1, r2…r i …r n ), where r i is the compensation coefficient of the i-th monitoring sub-area; Generate the first-level fire value of each monitoring sub-area according to the compensation coefficient sequence R; Generate firefighting decisions based on all first-level fire values.

6. The wind turbine cabin fire warning system based on image recognition according to claim 5, characterized in that: When generating firefighting decisions based on all primary fire values, including: Establish the first-level fire value series H=(h1,h2…h i …h n ), where h i is the first-level fire value of the i-th monitoring sub-area; h i =r i *g i， Among them, g i is the initial fire value of the ith monitoring sub-area; According to the first-level fire value series H, the working parameters of the fire-fighting equipment in each monitoring sub-area are set in turn.

Images