A Multi-Feature Fusion Photovoltaic Fire Multi-Level Early Warning System and Method

Abstract

This invention belongs to the field of photovoltaic DC power distribution technology and discloses a multi-feature fusion photovoltaic fire multi-level early warning system and method. The method is based on fault arc detection using frequency domain and mode decomposition: it performs frequency domain fault feature extraction, ICEEMDAN fault feature extraction, and construction of a multi-dimensional feature space for DC arc faults, obtaining a two-dimensional feature plane of arc faults characterized by frequency domain and mode decomposition energy entropy; it sorts multiple fire hazard factors from low to high importance, including voltage, current, insulation condition, temperature, arc risk, and smoke alarm, to measure the probability of fire occurrence; it obtains a fire risk credibility measurement model with fixed weights; and it divides the credibility level values ​​of the fire risk credibility measurement model according to non-equidistant measurement distances, resulting in a three-level photovoltaic fire early warning system. This invention combines fault arc detection to identify fire risk levels, enabling accurate judgment of fire risk levels.

Description

Technical Field

[0001] This invention belongs to the field of photovoltaic DC power distribution technology, and particularly relates to a multi-feature fusion photovoltaic fire multi-level early warning system and method. Background Technology

[0002] With the continuous expansion of photovoltaic (PV) power generation systems, PV DC distribution systems, as a crucial component of these systems, face a series of technical challenges. Within PV DC distribution cabinets, the risk of fire gradually increases due to the prolonged operation of electrical equipment and interference from internal and external factors such as environmental changes. Furthermore, the compact and complex space within these cabinets makes protection and isolation difficult in the event of a fire. In DC distribution cabinets lacking arc detection and protection devices, large-scale electrical fires are often caused by faulty arcs. Because DC arcs are difficult to extinguish, their occurrence can lead to serious fire accidents, thus serving as a warning sign of an impending fire. As a type of electrical fire, PV fires not only cause the shutdown of PV power generation systems but also pose serious threats to personal safety and the environment. Therefore, there is an urgent need to establish an effective and reliable multi-level early warning system for PV fires.

[0003] Current photovoltaic fire early warning systems only detect fire characteristic quantities such as smoke and temperature, as well as basic electrical quantities such as voltage and current. Existing technologies do not involve multi-level detection of photovoltaic fires that incorporates fault arc detection technology.

[0004] Based on the above analysis, the problems and defects of the existing technology are as follows: the existing technology does not combine multi-information fusion fault arc detection technology to identify fire risk levels, and the judgment of fire risk levels is complex and inaccurate. Summary of the Invention

[0005] To overcome the problems existing in related technologies, the present invention discloses an embodiment of a multi-feature fusion photovoltaic fire multi-level early warning system and method.

[0006] The technical solution is as follows: a multi-feature fusion method for multi-level early warning of photovoltaic fires. This method integrates the evaluation of multi-dimensional features with arc behavior, treats arc behavior as a fire risk factor, classifies and assesses fire risks, and obtains the potential hazard of arc events and the fire risk level. The method includes the following steps:

[0007] S1, Fault arc detection based on frequency domain and mode decomposition: frequency domain fault feature extraction, ICEEMDAN fault feature extraction and DC arc fault multidimensional feature space construction are performed to obtain a two-dimensional feature plane of arc fault characterized by frequency domain and mode decomposition energy entropy.

[0008] S2, based on the obtained two-dimensional feature plane of arc fault, sorted by importance from low to high, select fire risk factors such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm, measure the probability of fire occurrence, and obtain a fire risk credibility measurement model with fixed weights.

[0009] S3, according to the non-equidistant measurement distance, divide the fire risk credibility measurement model into credibility level values, and obtain the photovoltaic fire three-level early warning system.

[0010] In step S1, frequency domain fault feature extraction includes: performing a fast Fourier transform on the time-domain bus current signal; selecting a frequency band between 10kHz and 100kHz for arc detection; and calculating the sum of squares of the current amplitudes at all points corresponding to this frequency band as the energy of that frequency band, expressed as:

[0011]

[0012] In the formula, E arc I represents the energy of the frequency band. Li Let be the amplitude of the i-th current signal within the frequency band.

[0013] In step S1, the ICEEMDAN fault feature extraction uses the ICEEMDAN algorithm to decompose the signal and extract fault features, including the following steps:

[0014] S1.1, add controllable noise to the original signal to construct a noisy signal M, the expression is:

[0015] s (ξ) =s+a1E1[ω n (ξ) ],ξ=1,2…M (2)

[0016] In the formula, s (ξ) Let ξ be the constructed signal, s be the original signal, a1 be the ratio of the standard deviation of the added white noise to the standard deviation of the first input signal, E1[] be the first modal component of EMD decomposition, and ω be the signal of the first mode. n (ξ) Added standard normal white noise;

[0017] S1.2, Calculate the first residual, the expression is:

[0018] γ1= <M[s (ζ) (3)

[0019] In the formula, γ1 is the first residual, M[] is the local mean function, and <> is the average value calculation;

[0020] S1.3, Calculate the first modal component, the expression is:

[0021] Fim1 =s-γ1 (4)

[0022] In the formula, F im1 This is the first modal component;

[0023] S1.4, Calculate the q-th modal component, the expression is:

[0024] γ q = <M{γ q-1 +a q E q [ω n (ξ) (5)

[0025] F imq =γ q-1 -γ q (6)

[0026] In the formula, γ q For the q-th residual, F imq For the q-th modal component, q≥2; γ q-1 For the (q-1)th residual, a q E is the ratio of the standard deviations of the q-th input signal. q [] represents the q-th modal component of the EMD decomposition;

[0027] S1.5, repeat step S1.4 until the iteration termination condition is met.

[0028] In step S1.4, the modal component F is obtained. imq Next, usable modal components are screened, and those with the lowest similarity to the modal signal under normal conditions are selected for analysis. Generally, it is advisable to select the 1 to 2 modal components with the lowest similarity, preferably below 0.15. The cosine similarity is used to calculate the similarity between each modal component of the normal signal and the noise signal. The formula for calculating the similarity is:

[0029]

[0030] In the formula, s j For the cosine similarity of the modal components, F j (t) is the j-th order modal signal, M j (t) represents a j-th order normal signal, where j is the order of the modal component and N is the number of data points in the signal.

[0031] Furthermore, after comparing and obtaining the modal component with the lowest similarity, the energy entropy H is defined to represent the probability of a certain information occurring. The energy entropy of this order signal is calculated as the energy entropy feature of the arc fault signal. The formula for calculating the energy entropy H is:

[0032]

[0033] In the formula, H j Let ρ be the energy entropy of the j-th order modal component. j Let be the proportion of the signal energy of the j-th order modal component to the total signal energy, and m be the total order of the modal components.

[0034] In step S1, the construction of the multidimensional feature space for DC arc faults includes:

[0035] A two-dimensional feature plane for arc faults, characterized by frequency domain and mode decomposition energy entropy, is constructed. The arc fault feature plane is divided into three regions: normal, interference, and fault. The position of the current signal in the plane is determined by the frequency domain distance and the distance from the origin based on the mode decomposition energy entropy. The formula for calculating the distance to the feature origin O is as follows:

[0036] d 频域 =|E arc -O|,d 模态 =|H j -O| (9)

[0037] In the formula, d 频域 E represents the distance of the current signal along its characteristic direction in the frequency domain relative to the origin O. arc The energy represented by the current signal, O is the characteristic origin, d 模态 H represents the distance of the current signal along the characteristic direction of mode decomposition relative to the characteristic origin O. j Let d be the energy entropy. 频域 and d 模态 The energy E corresponding to the current signal respectively arc The energy entropy H corresponding to the j-th modal component of the current signal j Decide.

[0038] Furthermore, the construction of the two-dimensional feature plane of the arc fault characterized by frequency domain and mode decomposition energy entropy includes:

[0039] Define F m and F n These are the judgment thresholds for the frequency domain energy feature direction under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the frequency domain energy feature direction;

[0040] Define H m and H n These are the judgment thresholds for the characteristic directions of mode decomposition energy entropy under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the characteristic directions of mode decomposition energy entropy.

[0041] ΔF and ΔH are defined as the distances from the normal region to the arc fault region in the direction of frequency domain features and mode decomposition energy entropy features, respectively, and are respectively represented as the side length of the interference region along the feature direction. The smaller the values ​​of ΔF and ΔH, the more sensitive the judgment of arc faults, but the worse the reliability. Conversely, the larger the interference region, the higher the reliability of the judgment of arc faults, but the worse the sensitivity.

[0042] In step S2, the expression for the fire risk credibility measurement model is:

[0043]

[0044] In the formula, T is the credible assessment value of fire risk, and y i Let y be the attribute risk value of the i-th fire risk attribute. i ∈{2,5,7,9,10};a i Let a be the weight value of the i-th fire risk attribute, where 0 ≤ a i ≤1, In the case of y1 to y6, the six fire hazard factors are voltage, current, insulation condition, temperature, arc risk, and smoke alarm, respectively.

[0045] The weights of each fire hazard factor were calculated using the Analytic Hierarchy Process (AHP); the importance of factors was measured using a 9-level scale; a relationship matrix was established based on the importance scale; and the weighted coefficients of each factor were obtained through matrix operations.

[0046] In the positive-negative judgment matrix, a 1-9 scale is used to determine the contribution of any i-th and j-th attributes to the software's trustworthiness, and the contribution is assigned a value of a. ij The importance of the i-th attribute to the j-th attribute is divided into nine levels: {9, 7, 5, 3, 1, 1 / 3, 1 / 5, 1 / 7, 1 / 9}, with 9 being the most important and 1 / 9 being the least important. The judgment matrix C is listed.

[0047] The consistency of the judgment matrix C is checked using the following formula:

[0048]

[0049]

[0050] In the formula, CI is the consistency index, CR is the consistency ratio under random conditions of the judgment matrix, and λ max is the largest eigenvalue of the positively reciprocal matrix, n is the matrix order, and RI is the average random consistency index of the judgment matrix.

[0051] The matrix order is determined to be 6, and RI = 1.26 is set. CR is obtained according to Formula 12. When CR < 0.1, the weights are reasonable. If CR is greater than or equal to 0.1, the matrix is ​​readjusted until CR is less than 0.1. Finally, a fire risk credibility measurement model T with fixed weights is obtained.

[0052] In step S3, in the photovoltaic fire three-level early warning system, level one is the lowest level and level three is the highest level. The higher the fire risk confidence level, the smaller the measurement interval of that level.

[0053] Another objective of this invention is to provide a multi-feature fusion photovoltaic fire multi-level early warning system, which implements the multi-feature fusion photovoltaic fire multi-level early warning method. The system includes:

[0054] The fault arc detection module based on frequency domain and mode decomposition performs frequency domain fault feature extraction, ICEEMDAN fault feature extraction, and construction of a multi-dimensional feature space for DC arc faults to obtain a two-dimensional feature plane of arc faults characterized by frequency domain and mode decomposition energy entropy.

[0055] The photovoltaic fire three-level early warning system establishment module is used to select fire hazard factors such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm based on the acquired two-dimensional feature plane of arc fault, sort them from low to high importance, measure the probability of fire occurrence, and obtain a fire risk credibility measurement model with fixed weights; and divide the credibility level value of the fire risk credibility measurement model according to the non-equidistant measurement distance to obtain the photovoltaic fire three-level early warning system.

[0056] Combining all the above technical solutions, the advantages and positive effects of this invention are as follows: This invention fills the gap in the detection system for power distribution fires that are usually caused by electric arcs in photovoltaic DC distribution cabinets. It accurately and quickly judges electric arc faults that are prone to cause fires through multi-feature fusion technology. At the same time, it quantifies fire risk through a risk credibility measurement model and scientifically defines fire risk levels, so as to realize fire classification early warning and prevention control. Attached Figure Description

[0057] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this disclosure and, together with the description, serve to explain the principles of this disclosure;

[0058] Figure 1 This is a flowchart of the multi-feature fusion photovoltaic fire multi-level early warning method provided in the embodiments of the present invention;

[0059] Figure 2 This is a two-dimensional schematic diagram of fault arc detection provided in an embodiment of the present invention;

[0060] Figure 3This is a schematic diagram of the fire risk level classification axis provided in an embodiment of the present invention;

[0061] Figure 4 This is a schematic diagram of a multi-feature fusion photovoltaic fire multi-level early warning system provided in an embodiment of the present invention;

[0062] In the figure: 1. Fault arc detection module based on frequency domain and mode decomposition; 2. Photovoltaic fire three-level early warning system establishment module. Detailed Implementation

[0063] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Many specific details are set forth in the following description to provide a thorough understanding of the present invention. However, the present invention can be practiced in many other ways different from those described herein, and those skilled in the art can make similar modifications without departing from the spirit of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0064] The innovative aspect of the multi-feature fusion photovoltaic fire multi-level early warning system and method provided in this invention lies in its innovative integration of multi-dimensional feature assessment with arc behavior, further considering arc behavior as a fire risk factor to achieve fire risk classification and assessment. This method enables a more comprehensive understanding of the potential hazards of arc events, thereby improving the accuracy and reliability of fire risk classification. The application of this technology is expected to have a significant impact on the field of fire safety, providing more effective fire prevention and management methods.

[0065] Example 1, such as Figure 1 As shown, the multi-feature fusion-based photovoltaic fire multi-level early warning method integrates the assessment of multi-dimensional features with arc behavior, treating arc behavior as a fire risk factor, classifying and assessing fire risks, and obtaining the potential hazard of arc events and fire risk levels. The method includes the following steps:

[0066] S1, Fault arc detection based on frequency domain and mode decomposition: frequency domain fault feature extraction, ICEEMDAN fault feature extraction and DC arc fault multidimensional feature space construction are performed to obtain a two-dimensional feature plane of arc fault characterized by frequency domain and mode decomposition energy entropy.

[0067] S2, based on the obtained two-dimensional feature plane of arc fault, sorted by importance from low to high, select fire risk factors such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm, measure the probability of fire occurrence, and obtain a fire risk credibility measurement model with fixed weights.

[0068] S3, according to the non-equidistant measurement distance, divide the fire risk credibility measurement model into credibility level values, and obtain the photovoltaic fire three-level early warning system.

[0069] It is understandable that the credibility level value is determined independently based on the actual situation, but the non-equidistant value depends on historical fire data: the higher the fire risk level, the smaller its proportion.

[0070] In step S1 of this embodiment of the invention, frequency domain fault feature extraction includes:

[0071] A Fast Fourier Transform (FFT) is performed on the time-domain bus current signal. Due to the geometry of wiring in a typical PV system, the noise current density above 200kHz varies significantly with frequency. For this reason, a general frequency band between 10kHz and 100kHz is chosen for arc detection. The sum of the squares of the current amplitudes at all points corresponding to this frequency band is calculated as the energy of that frequency band, as shown in the following formula:

[0072]

[0073] In the formula, E arc I represents the energy of the frequency band. Li Let be the amplitude of the i-th current signal within the frequency band.

[0074] Formula (1) can be used to quantify the fault characteristics of the frequency band in the frequency domain.

[0075] In step S1 of this embodiment of the invention, the ICEEMDAN fault feature extraction uses the ICEEMDAN algorithm to decompose the signal and extract fault features, specifically including the following steps:

[0076] S1.1, add controllable noise to the original signal to construct a noisy signal M, the expression is:

[0077] s (ξ) =s+a1E1pω n (ξ) [,ξ=1,2…M (2)

[0078] In the formula, s (ξ) Let ξ be the constructed signal, s be the original signal, a1 be the ratio of the standard deviation of the added white noise to the standard deviation of the first input signal, E1[] be the first modal component of EMD decomposition, and ω be the signal of the first mode. n (ξ) Added standard normal white noise;

[0079] S1.2, Calculate the first residual, the expression is:

[0080] γ1= <M[s (ξ) (3)

[0081] In the formula, γ1 is the first residual, M[] is the local mean function, and <> is the average value calculation;

[0082] S1.3, Calculate the first modal component, the expression is:

[0083] F im1 =s-γ1 (4)

[0084] In the formula, F im1 This is the first modal component;

[0085] S1.4, Calculate the q-th modal component, the expression is:

[0086] γ q = <M{γ q-1 +a q E q [ω n (ξ) (5)

[0087] F imq =γ q-1 -γ q (6)

[0088] In the formula, γ q For the w-th residual, F imq For the q-th modal component, w≥2; γ q-1 For the (q-1)th residual, a q E is the ratio of the standard deviations of the q-th input signal. q [] represents the q-th modal component of the EMD decomposition;

[0089] S1.5, repeat step S1.4 until the iteration termination condition is met.

[0090] After obtaining the modal component F imq Next, usable modal components are screened, and those with the lowest similarity to the modal signal under normal conditions are selected for analysis. Generally, it is advisable to select the 1 to 2 modal components with the lowest similarity, preferably below 0.15. The cosine similarity is used to calculate the similarity between each modal component of the normal signal and the noise signal. The formula for calculating the similarity is:

[0091]

[0092] In the formula, s j For the cosine similarity of the modal components, F j (t) is the j-th order modal signal, M j (t) represents a j-th order normal signal, where j is the order of the modal component and N is the number of data points in the signal.

[0093] After comparing and obtaining the modal component with the lowest similarity, the energy entropy H is defined to characterize the probability of a certain information appearing. Therefore, the energy entropy of this order signal can be calculated as the energy entropy feature of the arc fault signal. The formula for calculating the energy entropy H is:

[0094]

[0095] In the formula, H j Let ρ be the energy entropy of the j-th order modal component. j Let be the proportion of the signal energy of the j-th order modal component to the total signal energy, and m be the total order of the modal components.

[0096] Using formulas (2)-(8), feature information can be extracted by utilizing information in the time-frequency domain. The original signal is decomposed into multiple modal components using formulas (2)-(6). Among these modal components, some components contain more fault feature information. Formula (7) is used to filter out these components with more information. Finally, energy entropy is defined using formula (8) to characterize the probability of fault information occurring.

[0097] In step S1 of this embodiment of the invention, the construction of the multidimensional feature space for DC arc faults includes:

[0098] In step S1, the construction of the multidimensional feature space for DC arc faults includes:

[0099] A two-dimensional feature plane for arc faults is constructed, characterized by frequency domain and mode decomposition energy entropy. The arc fault feature plane is divided into three regions: normal, interference, and fault. The position of the current signal in the plane is determined by the frequency domain distance and the distance from the origin based on the mode decomposition energy entropy. The formula for calculating the distance to the feature origin O is as follows:

[0100] d 频域 =|E arc -O|,d 模态 =|H j -O|(9)

[0101] In the formula, d 频域 E represents the distance of the current signal along its characteristic direction in the frequency domain relative to the origin O. arc The energy represented by the current signal, O is the characteristic origin, d 模态 H represents the distance of the current signal along the characteristic direction of mode decomposition relative to the characteristic origin O. j Let d be the energy entropy. 频域 and d 模态 The energy E corresponding to the current signal respectively arc The energy entropy H corresponding to the j-th modal component of the current signal j Decide.

[0102] A two-dimensional feature plane for arc faults, characterized by frequency domain and mode decomposition energy entropy, is constructed, including:

[0103] Define F m and F n These are the judgment thresholds for the frequency domain energy feature direction under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the frequency domain energy feature direction;

[0104] Define H m and H n These are the judgment thresholds for the characteristic directions of mode decomposition energy entropy under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the characteristic directions of mode decomposition energy entropy.

[0105] Let ΔF and ΔH be the distances from the normal region to the arc fault region along the frequency domain characteristic and mode decomposition energy entropy characteristic directions, respectively. They are also represented as the side lengths of the interference region along the characteristic directions. Smaller values ​​of ΔF and ΔH result in higher sensitivity but lower reliability in arc fault detection; conversely, larger interference regions lead to higher reliability but lower sensitivity in arc fault detection. Figure 2 Two-dimensional schematic diagram of fault arc detection.

[0106] In this embodiment of the invention, the expression for the fire risk credibility measurement model is:

[0107]

[0108] In the formula, T is the credible assessment value of fire risk, and y i Let y be the attribute risk value of the i-th fire risk attribute. i ∈{2,5,7,9,10};a i Let a be the weight value of the i-th fire risk attribute, where 0 ≤ a i ≤1, In the case of y1 to y6, they correspond to the six fire hazard factors, namely voltage, current, insulation condition, temperature, arc risk, and smoke alarm, respectively.

[0109] The weights of each fire hazard factor were calculated using the Analytic Hierarchy Process (AHP); the importance of factors was measured using a 9-level scale; a relationship matrix was established based on the importance scale; and the weighted coefficients of each factor were obtained through matrix operations.

[0110] In the positive-negative judgment matrix, a 1-9 scale is used to determine the contribution of any i-th and j-th attributes to the software's trustworthiness, and the contribution is assigned a value of a. ijThe importance of the i-th attribute to the j-th attribute is divided into nine levels: {9, 7, 5, 3, 1, 1 / 3, 1 / 5, 1 / 7, 1 / 9}, with 9 being the most important and 1 / 9 being the least important. The judgment matrix C is listed.

[0111] The consistency of the judgment matrix C is checked using the following formula:

[0112]

[0113]

[0114] In the formula, CI is the consistency index, CR is the consistency ratio under random conditions of the judgment matrix, and λ max Let n be the largest eigenvalue of the positively reciprocal matrix, n be the matrix order, and RI be the average random consistency index of the matrix. The table below lists the RI values ​​for orders 1-9.

[0115] Table 1. RI values ​​corresponding to orders 1-9

[0116]

[0117]

[0118] The present invention determines that the matrix order is 6, takes RI = 1.26, and calculates CR according to formula 12; when CR < 0.1, the weights are reasonable; if it is greater than or equal to 0.1, the matrix is ​​readjusted until CR is less than 0.1; finally, a fire risk credibility measurement model T with fixed weights is obtained.

[0119] In step S3 of this embodiment of the invention, the fire risk level inside the cabinet is determined to be divided into three levels, with level one being the lowest and level three being the highest. The reliability level values ​​of the fire risk reliability measurement model are divided according to non-equidistant measurement distances; the higher the fire risk reliability level, the smaller the measurement interval for that level. The fire level intervals are as follows: Figure 3 Fire risk level classification axis diagram. From this, a three-level early warning system for photovoltaic fires can be obtained.

[0120] As can be seen from the above embodiments, the solution of the present invention can achieve low-cost, accurate and reliable detection of photovoltaic DC power distribution fires. It completes fire risk classification by using conventional fire hazard characteristics such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm, thereby reducing system costs and improving its reliability.

[0121] The technical solution of this invention fills the technical gap in the field of photovoltaic DC power distribution fire detection, which combines multi-dimensional information fusion for fault arc detection and quantitative fire risk and level classification.

[0122] This invention accurately and reliably quantifies fire risk by combining fault arc detection technology, and then classifies fire hazard levels, thus solving the problem of inaccurate and unscientific fire hazard level classification in this field.

[0123] This invention effectively overcomes data bias and algorithmic bias. In the construction of a multi-level fire early warning system, this method uses the Analytic Hierarchy Process (AHP) to calculate the weights of each fire hazard factor, minimizing technical bias caused by personal subjective factors.

[0124] Example 2, as Figure 4 As shown, the multi-feature fusion photovoltaic fire multi-level early warning system provided in this embodiment of the invention includes:

[0125] The fault arc detection module 1 based on frequency domain and mode decomposition is used to extract frequency domain fault features, extract ICEEMDAN fault features, and construct a multi-dimensional feature space for DC arc faults, thereby obtaining a two-dimensional feature plane of arc faults characterized by frequency domain and mode decomposition energy entropy.

[0126] Module 2 of the photovoltaic fire three-level early warning system is used to select multiple fire risk factors, such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm, based on the acquired two-dimensional feature plane of arc faults, sorted by importance from low to high, to measure the probability of fire occurrence; and obtain a fire risk credibility measurement model with fixed weights.

[0127] And the credibility level value used to classify the fire risk credibility measurement model according to non-equidistant measurement distance, to obtain the photovoltaic fire three-level early warning system.

[0128] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0129] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments.

[0130] This invention also provides a computer device comprising: at least one processor, a memory, and a computer program stored in the memory and executable on the at least one processor, wherein the processor executes the computer program to implement the steps in any of the above method embodiments.

[0131] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps described in the various method embodiments above.

[0132] This invention also provides an information data processing terminal, which, when executed on an electronic device, provides a user input interface to implement the steps described in the above method embodiments. The information data processing terminal is not limited to mobile phones, computers, or switches.

[0133] This invention also provides a server that, when executed on an electronic device, provides a user input interface to implement the steps described in the above method embodiments.

[0134] This invention provides a computer program product that, when run on an electronic device, enables the electronic device to implement the steps described in the various method embodiments.

[0135] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of this application can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying the computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks.

[0136] To further illustrate the effects of the embodiments of the present invention, the following experiment was conducted: A photovoltaic DC power distribution system was constructed, including a DC constant voltage source, a circuit breaker, an arc generator, a current sensor, an inverter, and an RL load. Following actual engineering connections, the arc generator was connected in series at the DC bus, i.e., the output terminal of the DC constant voltage source. The industrial control computer housing the photovoltaic fire multi-level early warning system was connected to both ends of the inverter, and voltage and current information were measured. The experiment employed an arc-drawing test method, where the two poles of the arc generator were separated at a fixed speed to generate an arc. To simulate a fire, a PVC sheath was placed over the two poles of the arc generator, and an arc was drawn. The alarm level of the photovoltaic fire multi-level early warning system was observed at different times after the arc began. The experimental results showed that the alarm level should increase as time progressed.

[0137] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any modifications, equivalent substitutions and improvements made by those skilled in the art within the scope of the technology disclosed in the present invention and within the spirit and principles of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A multi-feature fusion method for multi-level early warning of photovoltaic fires, characterized in that, This method integrates the assessment of multidimensional features with arc behavior, treating arc behavior as a fire risk factor to classify and assess fire risks, and obtain the potential hazards and fire risk levels of arc events. The method includes the following steps: S1, Fault arc detection based on frequency domain and mode decomposition: frequency domain fault feature extraction, ICEEMDAN fault feature extraction and DC arc fault multidimensional feature space construction are performed to obtain a two-dimensional feature plane of arc fault characterized by frequency domain and mode decomposition energy entropy. S2, based on the obtained two-dimensional feature plane of arc fault, sorted by importance from low to high, select fire risk factors such as voltage, current, insulation condition, temperature, arc risk, and smoke alarm, measure the probability of fire occurrence, and obtain a fire risk credibility measurement model with fixed weights. S3, according to the non-equidistant measurement distance, divide the fire risk credibility measurement model into credibility level values, and obtain the photovoltaic fire three-level early warning system.

2. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 1, characterized in that, In step S1, frequency domain fault feature extraction includes: performing a fast Fourier transform on the time-domain bus current signal; selecting a frequency band between 10kHz and 100kHz for arc detection; and calculating the sum of squares of the current amplitudes at all points corresponding to this frequency band as the energy of that frequency band, expressed as: In the formula, The energy represented by the frequency band, The first in the frequency band The amplitude of the current signal.

3. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 1, characterized in that, In step S1, the ICEEMDAN fault feature extraction uses the ICEEMDAN algorithm to decompose the signal and extract fault features, including the following steps: S1.1, adding controllable noise to the original signal to construct... The expression for a noisy signal is: In the formula, For the construction of the first One signal, The original signal, To include the ratio of the standard deviation of white noise to the standard deviation of the first input signal, This is the first modal component of EMD decomposition. Added standard normal white noise; S1.2, calculate the first residual, the expression is: In the formula, For the first residual, It is a local mean function. Calculated as an average; S1.3, Calculate the first modal component, the expression is: In the formula, This is the first modal component; S1.4, calculate the first... There are one modal component, expressed as: In the formula, For the first One residual, For the first One modal component, ; For the first One residual, For the first The ratio of the standard deviations of the input signals. The first EMD decomposition One modal component; S1.5, repeat step S1.4 until the iteration termination condition is met.

4. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 3, characterized in that, In step S1.4, the modal components are obtained. Next, select usable modal components and analyze the components with the lowest similarity to the modal signal under normal conditions. It is advisable to select the 1 to 2 modal components with the lowest similarity, below 0.

15. The cosine similarity is used to calculate the similarity between each modal component of the normal signal and the noise signal. The formula for calculating the similarity is: In the formula, Cosine similarity of modal components. for First-order modal signal, for Normal signal, Let be the order of the modal components. This refers to the number of data points in the signal.

5. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 4, characterized in that, After comparing and identifying the modal component with the lowest similarity, the energy entropy is defined. Characterizing the probability of the occurrence of certain information, the energy entropy of the signal at that order is calculated as the energy entropy characteristic of the arc fault signal. The calculation formula is: In the formula, for The energy entropy of the first modal component For the first The proportion of first-order mode component signal energy to the total signal energy. This represents the total order of the modal components.

6. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 1, characterized in that, In step S1, the construction of the multidimensional feature space for DC arc faults includes: A two-dimensional feature plane for arc faults is constructed, characterized by frequency domain and mode decomposition energy entropy. The arc fault feature plane is divided into three regions: normal, interference, and fault. The position of the current signal in the plane is determined by the frequency domain distance and the distance from the origin based on the mode decomposition energy entropy. The feature origin is defined as follows. The distance calculation formula is: In the formula, For the current signal in the frequency domain with respect to the characteristic origin distance, The energy represented by the current signal. As the characteristic origin, For the current signal in the mode decomposition characteristic direction with respect to the characteristic origin distance, For energy entropy, and The energy represented by the current signal With the current signal at the first Energy entropy corresponding to the first modal component Decide.

7. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 6, characterized in that, The construction of the two-dimensional feature plane of arc faults characterized by frequency domain and mode decomposition energy entropy includes: definition and These are the judgment thresholds for the frequency domain energy feature direction under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the frequency domain energy feature direction; definition and These are the judgment thresholds for the characteristic directions of mode decomposition energy entropy under normal and fault conditions, respectively, and are respectively represented as the boundaries of the normal region and the fault region in the characteristic directions of mode decomposition energy entropy. definition and These represent the distances from the normal region to the arc fault region along the frequency domain feature and mode decomposition energy entropy feature directions, respectively, and are respectively expressed as the side lengths of the interference region along the feature directions. and The smaller the value, the more sensitive the judgment of arc faults, but the worse the reliability. Conversely, the larger the interference area, the higher the reliability of the judgment of arc faults, but the worse the sensitivity.

8. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 7, characterized in that, In step S2, the expression for the fire risk credibility measurement model is: In the formula, This is a reliable assessment value for fire risk. For the first The attribute risk value of each fire risk attribute. ; For the first The weight value of each fire risk attribute, when satisfying In the case of, then to These correspond to six fire hazard factors in order: voltage, current, insulation condition, temperature, arc risk, and smoke alarm. The weights of each fire hazard factor were calculated using the Analytic Hierarchy Process (AHP); the importance of factors was measured using a 9-level scale; a relationship matrix was established based on the importance scale; and the weighted coefficients of each factor were obtained through matrix operations. In the positive-negative judgment matrix, a 19-scale is used to judge any number of... The first attribute and the first Each attribute contributes to the trustworthiness of the software, and the contribution value is assigned a value. According to the first The attribute for the first The importance of each attribute is categorized into nine levels: {9, 7, 5, 3, 1, 1 / 3, 1 / 5, 1 / 7, 1 / 9}, with 9 being the most important and 1 / 9 being the least important. A judgment matrix is ​​then created. ; The judgment matrix is ​​determined by the following formula. Consistency check: In the formula, As a consistency indicator, To determine the consistency ratio in the case of matrix randomness, It is the largest eigenvalue of the positively reciprocal matrix. Let be the order of the matrix. To determine the average random consistency index of a matrix; If the order of the matrix is ​​6, then take... According to formula 12, the result is obtained ;when When the weights are equal to or greater than 0.1, the matrix is ​​readjusted until... Less than 0.1; finally, a fire risk credibility measurement model with fixed weights is obtained. .

9. The multi-feature fusion method for multi-level early warning of photovoltaic fires according to claim 1, characterized in that, In step S3, in the photovoltaic fire three-level early warning system, level one is the lowest level and level three is the highest level. The higher the fire risk confidence level, the smaller the measurement interval of that level.

10. A multi-feature fusion photovoltaic fire multi-level early warning system, characterized in that, The system implements the multi-feature fusion photovoltaic fire multi-level early warning method according to any one of claims 1-9, and the system includes: The fault arc detection module based on frequency domain and mode decomposition (1) performs fault arc detection based on frequency domain and mode decomposition: frequency domain fault feature extraction, ICEEMDAN fault feature extraction and DC arc fault multidimensional feature space construction to obtain a two-dimensional feature plane of arc fault characterized by frequency domain and mode decomposition energy entropy. The photovoltaic fire three-level early warning system establishment module (2) is used to select fire risk factors such as voltage, current, insulation status, temperature, arc risk, and smoke alarm based on the obtained two-dimensional feature plane of arc fault, sort them from low to high importance, measure the probability of fire occurrence, obtain a fire risk credibility measurement model with fixed weights, and divide the credibility level value of the fire risk credibility measurement model according to the non-equidistant measurement distance to obtain the photovoltaic fire three-level early warning system.

Images