Distributed photovoltaic fire early warning method

Abstract

The invention discloses a distributed photovoltaic fire early warning method, and relates to the technical field of fire early warning, and the technical scheme is characterized in that the method comprises the steps: collecting the current, voltage and infrared temperature matrix of a photovoltaic module through a detection module; the data collected by the detection module is analyzed based on the space-time correlation analysis module, the dynamic weight coefficient of the current is output in real time, and a weighted feature vector is generated through dynamic fusion; the position P of the photovoltaic module is obtained through the positioning module, and a space-time correlation matrix between the photovoltaic modules is calculated through the space-time correlation analysis module; analyzing the weighted feature vector and the space-time incidence matrix based on an LSTM network, and outputting a risk score R (t) of the photovoltaic module; and triggering a grading analysis response according to the risk score. According to the invention, through multi-source detection, dynamic analysis and grading correspondence, efficient and accurate photovoltaic fire early warning is realized.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of fire warning, and more specifically, to a distributed photovoltaic fire warning method. Background Art

[0002] With the wide application of renewable energy, the solar photovoltaic power generation system, as a common way to utilize solar energy resources, has developed rapidly in recent years and is widely used in the field of clean energy supply. However, during the long-term operation of the photovoltaic power generation system, it may face various potential safety hazards. Among them, when the photovoltaic power generation system generates electricity, it also generates heat, which may cause a fire. The fire will cause serious losses to the power generation system and even threaten personal safety. Therefore, an efficient and reliable photovoltaic fire warning method is very important.

[0003] However, there are still deficiencies in the existing photovoltaic fire warning methods. The existing monitoring methods are very single, detecting only certain parameters in one aspect, lacking accuracy and comprehensiveness, resulting in inaccurate detection. Moreover, the existing photovoltaic fire warning generally uses traditional temperature sensors. The traditional temperature sensors have a low sampling rate and rely on cloud processing for data upload, resulting in low spatio-temporal correlation between photovoltaic modules, and thus a long delay time for fire warning. Usually, the warning starts after a period of time when a fire breaks out, which increases the harm and losses caused by the fire.

[0004] Therefore, a new solution needs to be proposed to solve this problem. Summary of the Invention

[0005] The purpose of the embodiments of the present invention is to provide a distributed photovoltaic fire warning method to solve the above problems.

[0006] The above technical purpose of the embodiments of the present invention is achieved through the following technical solutions:

[0007] In a first aspect, a distributed photovoltaic fire warning method is characterized in that the fire warning method includes the following steps:

[0008] Collect the current, voltage, and infrared temperature matrix of the photovoltaic modules through a detection module;

[0009] Analyze the data collected by the detection module based on a spatio-temporal correlation analysis module, and output in real time the dynamic weight coefficient α(t) of the current, the dynamic weight coefficient β(t) of the voltage, and the dynamic weight coefficient γ(t) of the infrared temperature matrix, and dynamically fuse to generate a weighted feature vector;

[0010] Locate the photovoltaic modules through a positioning component to obtain the position P of the photovoltaic modules, and calculate the spatio-temporal correlation matrix between the photovoltaic modules through the spatio-temporal correlation analysis module;

[0011] Analyze the weighted feature vector and the spatio-temporal correlation matrix based on the LSTM network, and output the risk score R(t) of the photovoltaic module;

[0012] Trigger a hierarchical analysis response according to the risk score. When 0.6 < R(t) ≤ 0.8, trigger a secondary alarm, reduce the load of the photovoltaic module power, and push the alarm information at the same time; when R(t) > 0.8, trigger a primary alarm, cut off the current, and start the fire extinguishing device at the same time.

[0013] The present invention is further configured that: the spatio-temporal correlation analysis module analyzes the data collected by the detection module, including normalizing the collected current and voltage;

[0014]

[0015] Among them, I norm (t) represents the data after the current is normalized, and V norm (t) represents the data after the voltage is normalized;

[0016] Compress the infrared temperature matrix to obtain T PCA (t).

[0017] The present invention is further configured that: the calculation methods of the dynamic weight coefficient α(t) of the output current, the dynamic weight coefficient β(t) of the voltage, and the dynamic weight coefficient γ(t) of the infrared temperature matrix are as follows:

[0018]

[0019]

[0020] Among them, the numerator represents the change rates of the current, voltage, and infrared temperature matrix, and the denominator is the sum of the change rates of the current, voltage, and infrared temperature matrix.

[0021] The present invention is further configured that: the calculation method of the dynamically fused weighted feature vector is as follows:

[0022] X(t) = α(t)·I norm (t) + β(t)·V norm (t) + γ(t)·T PCA (t)

[0023] Among them, the X(t) represents the weighted feature vector.

[0024] The present invention is further configured that: the spatio-temporal correlation analysis module calculates the spatio-temporal correlation matrix between photovoltaic modules, including:

[0025] The positioning of photovoltaic modules A and B is obtained as P A and P B , then the calculation formula of the spatio-temporal correlation matrix is:

[0026]

[0027] where M AB represents the spatio-temporal correlation matrix between photovoltaic modules A and B, d0 is the thermal diffusion characteristic distance, and the calculation formula of the d0 is:

[0028] d0 = k·υ(t)

[0029] where k is the material thermal conductivity of the photovoltaic module, and υ(t) is the real-time wind speed.

[0030] The present invention is further configured that: the LSTM network includes a fully connected output layer. By introducing an attention mechanism, the LSTM network captures the evolution laws of weighted feature vectors and spatio-temporal correlation matrices, generates a risk score for the photovoltaic module through the fully connected output layer, and aggregates the risk scores of all photovoltaic modules to generate an overall risk assessment.

[0031] In a second aspect, a computer-readable storage medium is provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a computer or a processor, the method described in any one of the above claims is implemented.

[0032] In a third aspect, a computer program product is provided. The computer program product includes a computer program, and when the computer program is executed by a computer or a processor, the computer or the processor is caused to execute the method described in any one of the above claims.

[0033] In summary, the present invention has the following beneficial effects:

[0034] 1. The current, voltage, and infrared temperature matrix of the photovoltaic module are simultaneously collected by the detection module, and the collected data are dynamically fused to generate a weighted feature vector. The weighted feature vector converts the parameters such as current, voltage, and temperature collected by different sensors into a unified mathematical expression, which is convenient for analyzing and comparing these collected data. By inputting the weighted feature vector into the LSTM network, the state of the photovoltaic module can be analyzed. By collecting and analyzing multiple types of data, the detection of the photovoltaic module is more accurate and comprehensive, and the state of the photovoltaic module can be detected more accurately:

[0035] 2. Position each photovoltaic module through the positioning component to obtain the position P of each photovoltaic module, and then generate the spatio-temporal correlation matrix between each photovoltaic module through the spatio-temporal correlation analysis module, enhancing the spatio-temporal correlation between each photovoltaic module. Furthermore, the fault spread path can be predicted through the heat diffusion and electrical coupling relationship between the photovoltaic modules. When the local anomaly has not exceeded the standard, an alarm can be triggered in advance based on the correlation, thus being able to remind the staff in advance, reducing the occurrence of fires and the hazards and losses caused by fires. The present invention realizes efficient and accurate photovoltaic fire warning through multi-source detection, dynamic analysis, and hierarchical response. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 is a schematic structural diagram of a distributed photovoltaic fire warning method of the present invention; DETAILED DESCRIPTION OF THE EMBODIMENTS

[0037] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work belong to the scope of protection of the present invention.

[0038] In a feasible embodiment, please refer to Figure 1 as shown, a distributed photovoltaic fire warning method, the fire warning method includes the following steps:

[0039] Step 101: Collect the current, voltage, and infrared temperature matrix of the photovoltaic module through the detection module;

[0040] Step 102: Analyze the data collected by the detection module based on the spatio-temporal correlation analysis module, and output the dynamic weight coefficient α(t) of the current, the dynamic weight coefficient β(t) of the voltage, and the dynamic weight coefficient γ(t) of the infrared temperature matrix in real time, and dynamically fuse to generate a weighted feature vector;

[0041] Step 103: Locate the photovoltaic module through the positioning component to obtain the position P of the photovoltaic module, and calculate the spatio-temporal correlation matrix between the photovoltaic modules through the spatio-temporal correlation analysis module;

[0042] Step 104: Analyze the weighted feature vector and the spatio-temporal correlation matrix based on the LSTM network, and output the risk score R(t) of the photovoltaic module;

[0043] Step 105: Trigger hierarchical analysis response according to the risk score. When 0.6 < R(t) ≤ 0.8, trigger a secondary alarm, reduce the power load of the photovoltaic module, and push the alarm information at the same time. When R(t) > 0.8, trigger a primary alarm, cut off the current, and start the fire extinguishing device at the same time.

[0044] Specifically, the detection module performs multi-source acquisition on the current, voltage, and infrared temperature matrix of the photovoltaic module, and then analyzes the collected data through the spatio-temporal correlation analysis module to obtain the dynamic weight coefficients α(t), β(t), and γ(t) of the current, voltage, and infrared temperature matrix respectively. Then, the spatio-temporal correlation analysis module dynamically fuses to generate a weighted feature vector, which will be updated in real time as time changes. By collecting and analyzing various data, the detection of the photovoltaic module is more accurate and comprehensive, and the state of the photovoltaic module can be detected more accurately.

[0045] Further, please refer to Figure 1 As shown, the photovoltaic module is positioned by the positioning component to obtain the position P of each photovoltaic module, and then the spatio-temporal correlation matrix between each photovoltaic module is calculated through the spatio-temporal correlation analysis model, so as to correlate each photovoltaic module, enhancing the spatio-temporal correlation between each photovoltaic module. Analyze the weighted feature vector and the spatio-temporal correlation matrix through the LSTM network, and then it is possible to predict the fault spread path based on the heat diffusion and electrical coupling relationship between the photovoltaic modules. When the local anomaly has not exceeded the standard, an alarm is triggered in advance based on the correlation, which can remind the staff in advance and reduce the occurrence of fires and the hazards and losses caused by fires.

[0046] Further, after analyzing the weighted feature vector and the spatio-temporal correlation matrix, the LSTM network will output the risk score R(t) of the photovoltaic module. Different values of R(t) will trigger different reactions. The present invention realizes efficient and accurate photovoltaic fire warning through multi-source detection, dynamic analysis, and hierarchical response.

[0047] Specifically, please refer to Figure 1 As shown, in step 101, the detection module includes Hall sensors for collecting current and voltage, such as Allegro ACS712. Allegro ACS712 has the advantages of high precision, low noise, and fast response. The sampling rate of Allegro ACS712 ≥ 1 kHz. The infrared temperature matrix is generated by an infrared thermal imager, such as FLIR A315. FLIR A315 can identify those high-temperature problems that are easily overlooked and has built-in logic devices. The resolution of FLIR A315 is 0.1 degree Celsius.

[0048] Further, please refer to Figure 1As shown, in step 102, the spatio-temporal correlation analysis module analyzes the data collected by the detection module, including normalizing the collected current and voltage;

[0049]

[0050] Among them, I norm (t) represents the data after normalizing the current, and V norm (t) represents the data after normalizing the voltage;

[0051] By normalizing the current and voltage to the interval [0, 1], the dimensional difference between the current and voltage can be eliminated, which can accelerate the convergence speed of optimization algorithms such as gradient descent, enabling the spatio-temporal correlation analysis module to find the optimal solution faster, thus accelerating the training speed;

[0052] Specifically, compress the infrared temperature matrix to obtain T PCA (t), reduce the dimension of the infrared temperature matrix, extract the main components, and reduce the amount of calculation.

[0053] The calculation methods of the dynamic weight coefficient α(t) of the output current, the dynamic weight coefficient β(t) of the voltage, and the dynamic weight coefficient γ(t) of the infrared temperature matrix are as follows:

[0054]

[0055] Among them, the numerator represents the change rates of the current, voltage, and infrared temperature matrix, and the denominator is the sum of the change rates of the current, voltage, and infrared temperature matrix. The weight coefficient changes dynamically with time, making it more accurate;

[0056] The calculation method of dynamically fusing to generate a weighted feature vector is as follows:

[0057] X(t) = α(t)·I norm (t) + β(t)·V norm (t) + γ(t)·T PCA (t)

[0058] Among them, X(t) represents the weighted feature vector. The generated weighted feature vector will be uploaded to the cloud to trigger spatio-temporal correlation analysis. By dynamically fusing to generate a weighted feature vector, the detection sensitivity for photovoltaic module anomalies can be improved. Adjust the weight according to the parameter change rate, so that sudden anomalies such as arcs and local overheating can be quickly captured. The dynamically fused weighted feature vector can optimize through real-time weight adjustment, multi-data complementarity, and edge computing, solving the defects of traditional methods in terms of sensitivity, false alarm rate, and adaptability, being able to accurately capture early faults, and being able to detect and solve problems as early as possible.

[0059] Furthermore, the spatio-temporal correlation analysis module calculates the spatio-temporal correlation matrix between photovoltaic modules, including:

[0060] The positions of photovoltaic modules A and B are located at P A and P B , then the calculation formula of the spatio-temporal correlation matrix is:

[0061]

[0062] where M AB represents the spatio-temporal correlation matrix between photovoltaic modules A and B, X A is the weighted feature vector of photovoltaic module A, X B is the weighted feature vector of photovoltaic module B, d0 is the thermal diffusion characteristic distance, and the calculation formula of d0 is:

[0063] d0 = k·υ(t)

[0064] where k is the material thermal conductivity of the photovoltaic module, υ(t) is the real-time wind speed. The spatio-temporal correlation matrix can capture the correlation between photovoltaic modules, avoid isolated misjudgment, accurately locate the fault source through the spatio-temporal correlation matrix, and predict the propagation path at the same time, and respond in advance.

[0065] Specifically, the position of the photovoltaic module is located through the positioning component, which can be located by GPS. d0 is the thermal diffusion characteristic distance, which represents the effective range of heat transfer on the surface of the photovoltaic module. By calculating d0, the thermal diffusion distance under different wind speeds can be obtained, and early warning can be adjusted in time.

[0066] Furthermore, the LSTM network includes a fully connected output layer. By introducing a spatio-temporal attention mechanism, the LSTM network captures the evolution laws of the weighted feature vector and the spatio-temporal correlation matrix, generates a risk score for the photovoltaic module through the fully connected output layer, aggregates the risk scores of all photovoltaic modules, generates an overall risk assessment, and triggers a hierarchical analysis response according to the risk score at the same time. When 0.6 < R(t) ≤ 0.8, a secondary alarm is triggered, the power of the photovoltaic module is reduced, and alarm information is pushed at the same time; when R(t) > 0.8, a primary alarm is triggered, the current is cut off, and the fire extinguishing device is started at the same time. The LSTM network is an improved recurrent neural network, which has a powerful long-distance dependence modeling ability, has wide application adaptability, and has a powerful prediction ability. The core advantage of the LSTM network lies in its gating structure and cell state design, making it one of the preferred models for processing sequence data, especially outstanding in scenarios that require long-term memory.

[0067] Specifically, LSTM combines the weighted feature vector with the spatio-temporal correlation rectangle to generate an enhanced feature Its calculation formula is as follows:

[0068]

[0069] N is the total number of components, and then the normalized score is output through the activation function:

[0070]

[0071] Among them, σ is the Sigmoid function, wA is the trainable weight parameter, which can be -1, 0 or 1, and b is the bias term.

[0072] Specifically, for example, when the temperature of the faulty photovoltaic module A rises from 40 °C to 120 °C and the required time is 60 seconds, the temperature of the adjacent photovoltaic module B will be affected by heat diffusion and rise from 40 °C to 65 °C. Then, when the time is 0 second, it is calculated that M at this time AB is 0.2, RA(t) is 0.3, RB(t) is 0.2, and the system has no response at this time; when 20 seconds have passed, it is calculated that M at this time AB is 0.6, RA(t) is 0.7, RB(t) is 0.4, and the secondary warning is started at this time to reduce the operation of the faulty photovoltaic module A; when 40 seconds have passed, it is calculated that M at this time AB is 0.9, RA(t) is 0.95, RB(t) is 0.7. At this time, the secondary warning of the adjacent photovoltaic module B is started to reduce the operation of the faulty photovoltaic module B, and the primary alarm of the faulty photovoltaic module A is started to cut off the circuit of photovoltaic module A; from this analysis, it can be seen that the traditional method does not have a spatio-temporal correlation matrix. Therefore, when 40 seconds have passed, only photovoltaic module A issues an alarm and does not predict the risk of photovoltaic module B. However, in the present invention, after 20 seconds have passed, through the spatio-temporal correlation matrix M AB Early warning is carried out to strive for 20 key seconds for emergency disposal, so as to effectively reduce the losses caused by the fire.

[0073] The embodiment of the present application also provides a computer-readable storage medium. The computer-readable storage medium can be any available medium that can be stored by a computing device or a data storage device such as a data center containing one or more available media. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a magnetic tape), an optical medium (for example, a DVD), or a semiconductor medium (for example, a solid-state drive), etc. The computer-readable storage medium contains instructions that instruct the computing device to execute the foregoing time synchronization method.

[0074] An embodiment of the present application further provides a computer program product including instructions. The computer program product can be software or a program product including instructions that can run on a computing device or be stored in any available medium. When the computer program product runs on a computer device, the computing device is caused to execute the foregoing time synchronization method.

[0075] In the description of this specification, the descriptions referring to the terms "one embodiment", "example", "specific example", etc. mean that the specific features, structures, materials or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0076] The preferred embodiments of the present invention disclosed above are only used to help explain the present invention. The preferred embodiments do not describe all the details in detail, nor do they limit the invention to the specific embodiments described. Obviously, many modifications and variations can be made according to the content of this specification. These embodiments are selected and specifically described in this specification in order to better explain the principles and practical applications of the present invention, so that those skilled in the art can understand and utilize the present invention well. The present invention is only limited by the claims and their full scope and equivalents.

Claims

1. A distributed photovoltaic fire early warning method, characterized in that: The fire warning method includes the following steps: Collect the current, voltage, and infrared temperature matrix of the photovoltaic module through a detection module; Based on a spatio-temporal correlation analysis module, analyze the data collected by the detection module, and output the dynamic weight coefficient α(t) of the current, the dynamic weight coefficient β(t) of the voltage, the dynamic weight coefficient γ(t) of the infrared temperature matrix in real time, and dynamically fuse to generate a weighted feature vector; Locate the photovoltaic module through a positioning component to obtain the position P of the photovoltaic module, and calculate the spatio-temporal correlation matrix between the photovoltaic modules through the spatio-temporal correlation analysis module; Analyze the weighted feature vector and the spatio-temporal correlation matrix based on the LSTM network, and output the risk score R(t) of the photovoltaic module; Trigger a hierarchical analysis response according to the risk score. When 0.6 < R(t) ≤ 0.8, trigger a secondary alarm, reduce the load of the photovoltaic module power, and push the alarm information at the same time. When R(t) > 0.8, trigger a primary alarm, cut off the current, and start the fire extinguishing device at the same time.

2. A distributed photovoltaic fire early warning method according to claim 1, characterized in that: The spatio-temporal correlation analysis module analyzes the data collected by the detection module, including normalizing the collected current and voltage; Among them, I norm (t) represents the data after current normalization, V norm (t) represents the data after voltage normalization; The infrared temperature matrix is ​​compressed to obtain T PCA (t).

3. A distributed photovoltaic fire early warning method according to claim 2, characterized in that: The calculation methods of the output dynamic weight coefficient α(t) of the current, the dynamic weight coefficient β(t) of the voltage, and the dynamic weight coefficient γ(t) of the infrared temperature matrix are as follows: Among them, the numerator represents the change rate of the current, voltage, and infrared temperature matrix, and the denominator is the sum of the change rates of the current, voltage, and infrared temperature matrix.

4. A distributed photovoltaic fire early warning method according to claim 3, characterized in that: The calculation method of dynamically fusing to generate a weighted feature vector is as follows: X(t)=α(t)·I norm (t)+β(t)·V norm (t)+γ(t)·T PCA (t) Among them, X(t) represents the weighted feature vector.

5. A distributed photovoltaic fire warning method according to claim 4, wherein: The spatio-temporal correlation analysis module calculates the spatio-temporal correlation matrix between the photovoltaic modules, including: The positioning of photovoltaic components A and B is P A and P B , then the calculation formula of the spatiotemporal correlation matrix is: Among them, M AB represents the spatiotemporal correlation matrix between photovoltaic modules A and B, d0 is the characteristic distance of thermal diffusion, and the calculation formula of d0 is: d0 = k·υ(t) Among them, k is the material thermal conductivity of the photovoltaic module, and υ(t) is the real-time wind speed.

6. A distributed photovoltaic fire early warning method according to claim 1, characterized in that: The LSTM network includes a fully connected output layer. The LSTM network captures the evolution law of the weighted feature vector and the spatio-temporal correlation matrix by introducing a spatio-temporal attention mechanism, generates a risk score for the photovoltaic module through the fully connected output layer, and aggregates the risk scores of all photovoltaic modules by the LSTM network to generate an overall risk assessment.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a computer or a processor, it implements the method described in any one of claims 1-6 above.

8. A computer program product, characterized in that: The computer program product includes a computer program, and when the computer program is executed by a computer or a processor, it causes the computer or the processor to execute the method described in any one of claims 1-6.

Images