Intelligent monitoring and early warning evaluation model and device for photovoltaic combiner box

Abstract

The invention belongs to the technical field of photovoltaic combiner box monitoring and early warning, and particularly discloses a petrochemical enterprise photovoltaic combiner box intelligent monitoring and early warning evaluation model and device. The problems that an existing gas concentration detection sensor cannot achieve the monitoring and early warning functions and the safety evaluation result of the photovoltaic combiner box is unreliable are effectively solved. The construction of the petrochemical enterprise photovoltaic combiner box intelligent monitoring and early warning evaluation model comprises the following steps: selecting evaluation indexes, and establishing an intelligent monitoring and early warning index system; calculating by using an analytic hierarchy process to obtain an index weight of the state parameter; calculating to obtain a safety evaluation result of the photovoltaic combiner box; determining an initial weight and a threshold value of the BP neural network by a GA algorithm; and establishing an intelligent monitoring and early warning evaluation model based on a GA-BP neural network. Through the GA-BP neural network, the training of the intelligent monitoring early warning evaluation model is prevented from falling into local minimum, the global optimization capability can be improved, and the prediction accuracy of the evaluation model is improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of photovoltaic combiner box monitoring and early warning, and particularly relates to a petrochemical enterprise photovoltaic combiner box intelligent monitoring and early warning evaluation model and device. BACKGROUND

[0002] Petrochemical enterprises mainly operate oil, natural gas and their products as main operating products. In the process of production, processing and transportation, due to high temperature, high pressure and other environmental conditions, it is very easy to have combustible gas leakage. In view of this situation, petrochemical enterprises usually use various explosion-proof types for electrical equipment in the explosion danger zone to ensure the safe operation of electrical equipment.

[0004] The photovoltaic combiner box, as the element closest to the photovoltaic cell in the photovoltaic power station, is a wiring device that ensures the orderly connection and current collection function of photovoltaic modules. The device can ensure that the photovoltaic system is easy to turn on and off during maintenance and inspection, and reduce the range of power failure when the photovoltaic system fails.

[0005] Since the location of the newly built photovoltaic project in the petrochemical enterprise is generally not located in the explosion danger zone, there is no mandatory requirement to use a combiner box with explosion-proof function. However, according to the accident statistics of photovoltaic power stations in the past ten years, photovoltaic combiner box failure is one of the common causes of photovoltaic power station failure. Arcs will be generated due to contact point falling off, device aging, insulation rupture, and poor grounding. Once the concentration of combustible gas reaches a certain threshold and the minimum ignition energy is reached, a larger explosion will occur.

[0006] The current combustible gas monitoring method used by enterprises mainly monitors the gas concentration through sensors or detection equipment. When the concentration exceeds a certain threshold, an alarm will be issued. This method has certain limitations and can only provide the current gas concentration value, and cannot realize the monitoring and early warning function. In addition, the safety evaluation of the photovoltaic combiner box will be affected by multiple factors such as combustible gas concentration, oxygen concentration, minimum ignition energy, temperature and humidity. However, the current combustible gas monitoring often cannot take into account the influence of these factors. Based on the above situation, the application provides a photovoltaic combiner box intelligent monitoring and early warning evaluation model and device. SUMMARY

[0007] One object of the application is to provide a photovoltaic combiner box intelligent monitoring and early warning evaluation model, which effectively solves the problems that the existing gas concentration detection sensor cannot realize the monitoring and early warning function and the safety evaluation result of the photovoltaic combiner box is unreliable.

[0008] To solve the above technical problems, the technical solution adopted by the application is:

[0009] A photovoltaic combiner box intelligent monitoring and early warning evaluation model, comprising the following steps:

[0010] S1, select evaluation indexes, and establish an intelligent monitoring and early warning index system.

[0011] S2, calculate the index weight of the state variable by using the analytic hierarchy process.

[0012] S3, calculate the safety evaluation result of the photovoltaic combiner box.

[0013] S4, GA algorithm determines the initial weight and threshold of the BP neural network.

[0014] S5, establish an intelligent monitoring and early warning evaluation model based on GA-BP neural network.

[0015] Further, in step S1, the evaluation indexes include but are not limited to the following state variables: combustible gas concentration, combustible gas ratio, oxygen concentration, minimum ignition energy, photovoltaic combiner box humidity and photovoltaic combiner box temperature.

[0016] Further, in step S3, the historical data is normalized, the state grade comment set is constructed, and the running health state of the state variable is characterized by a semi-trapezoidal membership function. According to the weight and membership degree calculation result of each evaluation index, the safety evaluation result of the photovoltaic combiner box is obtained.

[0017] Further, in step S4, the GA algorithm takes the initial weight and threshold obtained from the safety evaluation result as the gene code of the genetic algorithm. According to the error of the BP neural network on the training set, the fitness of each individual is calculated, and selection operation, crossover operation and mutation operation are performed in turn. When the iteration number reaches a predetermined value, the genetic operation is stopped, and the best individual is taken as the initial weight and threshold of the BP neural network.

[0018] Further, in step S5, the weight and threshold optimized by the GA algorithm are taken as the initial weight and threshold of the BP neural network. The output values of all neurons are obtained through forward calculation of the BP neural network. According to the error between the predicted value and the true value, the gradient value of the loss function for each weight and threshold is calculated. All parameters in the BP neural network are updated using the gradient descent method. By continuously updating the weight and threshold, an intelligent monitoring and early warning evaluation model based on GA-BP neural network is finally constructed.

[0019] Further, in step S2, the nine-level scale method is used to compare each state variable at the same level two by two, and the judgment matrix A relative to the same upper element is obtained: A=(a ij )n ×n .

[0020] In the formula, a ija is a scale of the importance degree of the state variable i relative to the state variable j, a ji a is a scale of the importance degree of the state variable j relative to the state variable i, a ii a is a scale of the importance degree of the state variable i relative to the state variable i, wherein a ij > 0, a ij = 1 / a ji , a ii = 1.

[0021] The judgment matrix A is a positive reciprocal matrix, there is a maximum eigenvalue and the eigenvalue is unique, the corresponding eigenvector M is normalized, that is, the weight of each state variable is obtained:

[0022]

[0023] Further, in step S3, since the dimensions and orders of magnitude of the state variables are different, all the state variables participating in the evaluation are normalized to represent the safety of the photovoltaic combiner box by the deviation.

[0024] When the actual measurement value exceeds the normal operation limit value, the deviation value is 1.

[0025] When the actual measurement value is equal to the standard value, the deviation value is 0, and the state variable is in the optimal state.

[0026] Further, in step S3, the deviation function is established for the state variables of the more-is-better type and the less-is-better type, respectively:

[0027]

[0028] Wherein, s(x) is a deviation function, x represents the actual measurement value of a single state variable, x max and x min are the maximum critical value and the minimum critical value respectively specified by the relevant regulations.

[0029] Further, in step S3, the health state of the photovoltaic combiner box is divided into four state levels of normal, attention, abnormal and serious, and the membership function of each state level is represented as:

[0030]

[0031]

[0032] Wherein, s(x) is a deviation function, is a normal state level, is an attention state level, is an abnormal state level, is a serious state level.

[0033] By calculating the membership function of the state grade corresponding to the state variable, the membership function matrix P is obtained m :

[0034]

[0035] In the formula, indicates the membership of each index under different state grades.

[0036] Further, the state evaluation vector U is obtained: U = W * P m ;

[0037] W is the weight assignment calculated according to each corresponding state variable.

[0038] According to the calculated value of each element in the state evaluation vector and the corresponding expected value, the safety evaluation result Q0 is calculated:

[0039] In the formula, U c represents the calculated value of each element of the state evaluation vector, Ex c represents the expected value of the degradation degree corresponding to the cth element.

[0040] Further, in step S4, the fitness of the individual is represented as f d :

[0041]

[0042] In the formula, K represents a coefficient, y d is the actual output of the BP neural network, is the expected output of the BP neural network, and z is the number of population individuals.

[0043] Then, the selection operation is first performed and the roulette wheel selection method is selected, and the selection probability p d of each individual d is:

[0044]

[0045] In the formula, f d is the fitness of individual d, and z is the number of population individuals.

[0046] Then, the single-point real number crossover method is used, and the mth chromosome and the e th chromosome are crossed at the f th gene site: a mf = a mf (1-b) + a ef b, a ef = a ef (1-b) + a mf b;

[0047] where b is a random number between 0 and 1.

[0048] The fth gene of the e th individual is selected for mutation operation calculation, and the formula is as follows:

[0049]

[0050] where a max and a min are the upper and lower bounds of the gene a ef , g is the current iteration number, G max is the maximum evolution number, and r and r2 are random numbers between 0 and 1.

[0051] Another object of the present application is to provide an intelligent monitoring and early warning device for a photovoltaic combiner box, comprising a gas sensor, a temperature and humidity sensor, an intelligent monitoring module, an intelligent early warning module, a self-powered source, a controller and a buzzer alarm, the intelligent early warning module comprising the intelligent monitoring and early warning evaluation model for the photovoltaic combiner box as described in the above embodiments for realizing safety evaluation.

[0052] The self-powered source is connected with the intelligent monitoring module and the intelligent early warning module, the gas sensor and the temperature and humidity sensor are both connected with the intelligent monitoring module, the buzzer alarm is connected with the controller, the controller is connected with the intelligent early warning module, and the intelligent early warning module is connected with a monitoring center.

[0053] Further, the gas sensor and the temperature and humidity sensor are both arranged inside the box body of the photovoltaic combiner box, the gas sensor is used for monitoring the combustible gas concentration, the combustible gas proportion and the oxygen concentration, and the temperature and humidity sensor is used for monitoring the humidity of the photovoltaic combiner box and the temperature inside the photovoltaic combiner box.

[0054] Further, the intelligent monitoring module is used for collecting data and realizing the determination and calculation of the minimum ignition energy.

[0055] Compared with the prior art, the present application has the beneficial technical effects that:

[0056] (1) The GA-BP neural network of the present application can avoid falling into a local minimum in the training of the intelligent monitoring and early warning evaluation model and can improve the global optimization capability, thereby improving the prediction accuracy of the evaluation model.

[0057] (2) The present application establishes an intelligent monitoring and early warning index system, obtains the index weight of the state variable by using the analytic hierarchy process, calculates the safety evaluation result of the photovoltaic combiner box through historical data, judges whether the photovoltaic combiner box has a safety risk, improves the model accuracy through the GA-BP neural network, and realizes multi-index comprehensive intelligent monitoring and early warning.

[0058] (3)The present application aims at the hidden danger of flammable gas leakage affecting the safety of photovoltaic operation in petrochemical enterprises, and realizes real-time early warning evaluation of flammable gas leakage of photovoltaic combiner boxes in petrochemical enterprises by considering the influence of factors such as flammable gas concentration, oxygen concentration, minimum ignition energy and humidity. BRIEF DESCRIPTION OF DRAWINGS

[0059] The present application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0060] Figure 1 is a membership degree diagram of the health state grade of the photovoltaic combiner box in embodiment 1 of the present application.

[0061] Figure 2 is a flow chart of the intelligent monitoring and early warning evaluation model established based on the GA-BP neural network of the present application.

[0062] Figure 3 is a GA-BP neural network structure diagram of the present application.

[0063] Figure 4 is a connection structure diagram of the intelligent monitoring and early warning device of the photovoltaic combiner box of the present application. DETAILED DESCRIPTION

[0064] The following detailed description of the preferred implementation method of the present application and the included embodiments can more easily understand the content of the present application. Unless otherwise defined, all technical and scientific terms used herein have the same meaning as generally understood by those of ordinary skill in the art to which the present application belongs. When there is a conflict, the definition in the specification shall prevail.

[0065] The term "comprise", "include", "have", "contain", or any other variant thereof, used herein is intended to cover non-exclusive inclusion. For example, a composition, step, method, article or device comprising the listed elements does not necessarily limit to only those elements, but can include other elements not explicitly listed or inherent to such composition, step, method, article or device.

[0066] Embodiment 1

[0067] The flammable gas monitoring method currently used by petrochemical enterprises mainly monitors the gas concentration through sensors or detection equipment, and an alarm will be issued when the concentration exceeds a certain threshold. This method has certain limitations. In addition, the safety evaluation of photovoltaic combiner boxes will be affected by multiple factors such as flammable gas concentration, oxygen concentration, minimum ignition energy, temperature and humidity, and the current flammable gas monitoring often cannot cover the influence of these factors.

[0068] Based on the above situation, an intelligent monitoring and early warning evaluation model of photovoltaic combiner boxes in accordance with the needs of petrochemical enterprises is established.

[0069] A photovoltaic combiner box intelligent monitoring early warning evaluation model, comprising the following steps:

[0070] S1, selecting evaluation indexes, establishing an intelligent monitoring early warning index system.

[0071] The influence factors of combustible gas leakage on the safety of the photovoltaic combiner box are analyzed, and appropriate early warning indexes are selected, which should include the following state variables: combustible gas concentration, combustible gas proportion, oxygen concentration, minimum ignition energy, photovoltaic combiner box humidity and photovoltaic combiner box temperature, etc.

[0072] S2, the index weight of the state variable is calculated by using the analytic hierarchy process.

[0073] In order to accurately reflect the health status of the evaluation model, the weighting of the state variable must be relatively reasonable, the weight of the state variable is calculated by using the analytic hierarchy process, which has the characteristics of relative scientific and rigorous, and can greatly avoid the subjective bias produced by the assignment of experts.

[0074] The nine-level scale method is used to compare each state variable in the same level two by two, and the judgment matrix A relative to the same upper element is obtained: a = (A ij ) n×n .

[0075] In the formula, a ij is the scale of the importance of state variable i relative to state variable j, a ji is the scale of the importance of state variable j relative to state variable i, and a ii is the scale of the importance of state variable i relative to state variable i.

[0076] Wherein, a ij > 0, a ij = 1 / a ji , a ii = 1.

[0077] The judgment matrix A is a positive reciprocal matrix, which has a maximum eigenvalue and a unique eigenvalue. The corresponding eigenvector M is normalized, that is, the weight of each state variable is obtained:

[0078]

[0079] S3, the safety evaluation result of the photovoltaic combiner box is calculated.

[0080] Because the dimensions and orders of magnitude of the state variables are different, all the state variables participating in the evaluation need to be normalized. The deviation reflects the deviation degree of the state variable from the normal value.

[0081] When the actual measurement value exceeds the normal operation limit value, the deviation value is 1; when the actual measurement value is equal to the standard value, the deviation value is 0, at which time the state parameter is in the optimal state.

[0082] On this basis, for the state parameters of the more the better type and the less the better type, the deviation function is respectively established:

[0083]

[0084]

[0085] Wherein, s(x) is the deviation function, x represents the actual measurement value of a single state parameter, x max and x min are the maximum critical value and the minimum critical value respectively specified by the relevant regulations.

[0086] Combined with the running state of the electrical equipment of petrochemical enterprises, the health state of the photovoltaic combiner box is divided into four grades of normal, attention, abnormal and serious, and a state grade comment set is constructed.

[0087] In the safety evaluation of the photovoltaic combiner box, the semi-trapezoidal membership function is selected to express the membership degree of each health state, as shown in Figure 1 .

[0088] The health state of the photovoltaic combiner box is divided into four state grades of normal, attention, abnormal and serious, and the membership function of each state grade is expressed as:

[0089]

[0090]

[0091] Wherein, s(x) is the deviation function, is the normal state grade, is the attention state grade, is the abnormal state grade, is the serious state grade.

[0092] By calculating the membership function of the state grade corresponding to the state parameter, the membership function matrix P m is obtained:

[0093]

[0094] In the formula, indicates the membership degree of each index under different state grades.

[0095] Further, the state evaluation vector U is obtained: U=W*P m ; W is the weight assignment calculated by each corresponding state parameter.

[0096] According to the calculated value of each element in the state evaluation vector and the expected value corresponding thereto, a safety evaluation result Q0 is calculated:

[0097] wherein U c represents the calculated value of each element of the state evaluation vector, Ex c represents the expected value of the degradation degree corresponding to the cth element.

[0098] S4, the GA algorithm determines the initial weight and threshold of the BP neural network.

[0099] The state parameters of the photovoltaic combiner box and the safety evaluation result have been normalized in step three, so there is no need to pre-process the data. The GA algorithm takes the initial weight and threshold obtained from the safety evaluation result as the gene code of the genetic algorithm, and calculates the fitness of each individual according to the error of the BP neural network on the training set. The formula is as follows:

[0100] The fitness of the individual is represented as f d :

[0101] In the formula, K represents a coefficient, y d is the actual output of the BP neural network, is the expected output of the BP neural network, and z is the number of population individuals.

[0102] Subsequently, selection operation, crossover operation and mutation operation are performed in sequence, and when the number of iterations reaches a predetermined value, the genetic operation is stopped, and the best individual is taken as the initial weight and threshold of the BP neural network. Here, the number of iterations is set to 30 times.

[0103] First, selection operation is performed and roulette selection method is selected, and the selection probability p d of each individual d is:

[0104]

[0105] In the formula, f d is the fitness of individual d, and z is the number of population individuals.

[0106] Single-point real number crossover method is adopted, and the mth chromosome and the e th chromosome are crossed at the f th gene site: a mf = a mf (1-b) + a ef b, a ef = a ef (1-b) + a mf b.

[0107] In the formula, b is a random number between 0 and 1.

[0108] The fth gene of the e th individual is selected for mutation operation calculation, and the formula is as follows:

[0109]

[0110] In the formula, a max and a min are the upper limit and the lower limit of the gene a ef , g is the current iteration number, G max is the maximum evolution number, and r and r2 are random numbers between 0 and 1.

[0111] S5, establishing an intelligent monitoring and early warning evaluation model based on the GA-BP neural network.

[0112] The weight and the threshold value obtained by the GA algorithm optimization are used as the initial weight and the threshold value of the BP neural network, the output values of all neurons are obtained through the forward calculation of the BP neural network, the gradient values of the loss function for each weight and threshold value are calculated according to the error between the predicted value and the true value, all parameters in the BP neural network are updated using the gradient descent method, and finally the intelligent monitoring and early warning evaluation model based on the GA-BP neural network is constructed by continuously updating the weight and the threshold value.

[0113] The process of establishing the intelligent monitoring and early warning evaluation model based on the GA-BP neural network is as shown in Figure 2 .

[0114] (1) Determine the neural network structure. In this embodiment, the GA-BP neural network structure is as shown in Figure 3 . The GA-BP neural network includes an input layer, an output layer and a hidden layer, the input layer and the output layer are respectively responsible for the input and output of data, the input layer includes: combustible gas concentration, combustible gas proportion, oxygen concentration, minimum ignition energy, junction box humidity and junction box temperature, the output layer includes: safety evaluation result, and the hidden layer is responsible for the internal processing of data.

[0115] (2) Determine the initial weight and the threshold value of the GA-BP neural network.

[0116] Firstly, the data is preprocessed;

[0117] Then, it is used as the gene code of the genetic algorithm;

[0118] Next, the fitness of each individual is calculated, and the selection operation, the crossover operation and the mutation operation are sequentially performed to calculate the fitness;

[0119] Finally, it is judged whether the iteration number is completed, if not, the step of "carrying out selection operation, cross operation and mutation operation in turn, and calculating fitness" is repeated until the iteration number is completed, and the best individual is taken as the initial weight and threshold of the BP neural network.

[0120] (3) calculating the error of the BP neural network on the training set.

[0121] (4) weight and threshold update.

[0122] The weight and threshold obtained by the GA algorithm optimization are taken as the initial weight and threshold of the BP neural network, the output values of all neurons are obtained through the forward calculation of the BP neural network, the gradient values of the loss function for each weight and threshold are calculated according to the error between the predicted value and the true value, and all parameters in the BP neural network are updated using the gradient descent method.

[0123] (5) through continuously updating the weight and threshold, finally, the intelligent monitoring and early warning evaluation model based on the GA-BP neural network is constructed.

[0124] The GA-BP neural network can avoid falling into local minimum in the training of the intelligent monitoring and early warning evaluation model and improve the global optimization ability, and the prediction accuracy of the evaluation model is improved.

[0125] The present application establishes an intelligent monitoring and early warning index system, obtains the index weight of the state parameter by using the analytic hierarchy process, calculates the safety evaluation result of the photovoltaic combiner box through historical data, judges whether the photovoltaic combiner box has safety risks, and improves the model precision through the GA-BP neural network, and realizes multi-index comprehensive intelligent monitoring and early warning.

[0126] The present application aims at the hidden danger of flammable gas leakage affecting the safe operation of photovoltaic in petrochemical enterprises, realizes real-time early warning evaluation of flammable gas leakage of photovoltaic combiner box in petrochemical enterprises by considering the influence of factors such as flammable gas concentration, oxygen concentration, minimum ignition energy and humidity.

[0127] Example 2

[0128] A photovoltaic combiner box intelligent monitoring and early warning device, as shown in Figure 4 The device includes a gas sensor, a temperature and humidity sensor, an intelligent monitoring module, an intelligent early warning module, a self-powered power supply, a controller and a buzzer alarm, the intelligent early warning module includes the photovoltaic combiner box intelligent monitoring and early warning evaluation model of the embodiment 1 for realizing safety evaluation.

[0129] The self-powered source is connected with the intelligent monitoring module and the intelligent early warning module, the gas sensor and the temperature and humidity sensor are connected with the intelligent monitoring module, the buzzer alarm is connected with the controller, the controller is connected with the intelligent early warning module, and the intelligent early warning module is connected with the monitoring center.

[0130] The gas sensor and the temperature and humidity sensor are arranged in the box body of the photovoltaic combiner box, the gas sensor is used for monitoring the combustible gas concentration, the combustible gas proportion and the oxygen concentration, and the temperature and humidity sensor is used for monitoring the humidity of the photovoltaic combiner box and the temperature in the photovoltaic combiner box.

[0131] The controller connected with the intelligent early warning module and the buzzer alarm are used for sending the buzzer alarm information to the inspection personnel, guiding the shutdown operation of the combiner box and personnel evacuation in the dangerous situation.

[0132] The intelligent monitoring module realizes the determination and calculation of the minimum ignition energy according to the collected current and other electrical parameters in addition to the collection of the above data.

[0133] The intelligent early warning module contains an intelligent monitoring early warning evaluation model, which can realize safety evaluation, and when the model result exceeds the safety threshold, the information of the combustible gas early warning is sent to the monitoring center of the petrochemical enterprise.

[0134] The GA-BP neural network avoids falling into a local minimum in the training of the intelligent monitoring early warning evaluation model and can improve the global optimization capability, thereby improving the prediction accuracy of the evaluation model.

[0135] The present application establishes an intelligent monitoring early warning index system, obtains the index weight of the state parameter by using the analytic hierarchy process, calculates the safety evaluation result of the photovoltaic combiner box through historical data, judges whether the photovoltaic combiner box has a safety risk, improves the model precision through the GA-BP neural network, and realizes multi-index comprehensive intelligent monitoring and early warning.

[0136] The present application considers the influence of combustible gas concentration, oxygen concentration, minimum ignition energy and humidity, realizes real-time early warning evaluation of combustible gas leakage of the photovoltaic combiner box of the petrochemical enterprise.

[0137] Of course, the above description is not a limitation of the present application, and the present application is not limited to the above examples, and the changes, modifications, additions or replacements made by the person skilled in the art within the essential scope of the present application should also belong to the protection scope of the present application.

Claims

1. A smart monitoring, early warning, and evaluation model for photovoltaic combiner boxes, characterized in that, Includes the following steps: S1. Select evaluation indicators and establish an intelligent monitoring and early warning indicator system; S2. Calculate the index weights of the state parameters using the analytic hierarchy process. S3. Calculate the safety assessment results of the photovoltaic combiner box; S4, GA algorithm determines the initial weights and thresholds of BP neural network; S5. Establish an intelligent monitoring, early warning and evaluation model based on GA-BP neural network.

2. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 1, characterized in that, In step S1, the evaluation indicators include, but are not limited to, the following state parameters: combustible gas concentration, combustible gas ratio, oxygen concentration, minimum ignition energy, photovoltaic combiner box humidity, and photovoltaic combiner box internal temperature.

3. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 2, characterized in that, In step S3, the historical data is normalized to construct a set of status level comments, and the operational health status of the status parameters is represented by a semi-trapezoidal membership function. Based on the weights and membership calculation results of each evaluation index, the safety evaluation result of the photovoltaic combiner box is obtained.

4. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 3, characterized in that, In step S4, the GA algorithm uses the initial weights and thresholds obtained from the security assessment results as the gene encoding of the genetic algorithm. Based on the error of the BP neural network on the training set, it calculates the fitness of each individual and performs selection, crossover, and mutation operations in sequence. When the number of iterations reaches a predetermined value, the genetic operation stops, and the best individual is used as the initial weights and thresholds of the BP neural network.

5. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 4, characterized in that, In step S5, the weights and thresholds obtained through the GA algorithm optimization are used as the initial weights and thresholds of the BP neural network. The output values ​​of all neurons are obtained through forward computation of the BP neural network. Based on the error between the predicted value and the true value, the gradient value of the loss function with respect to each weight and threshold is calculated. The gradient descent method is used to update all parameters in the BP neural network. By continuously updating the weights and thresholds, an intelligent monitoring, early warning and evaluation model based on the GA-BP neural network is finally constructed.

6. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 2, characterized in that, In step S2, the nine-level scaling method is used to compare each state parameter at the same level pairwise, resulting in a judgment matrix A relative to the same upper-level element: A = (a ij ) n×n ; In the formula, a ij a is a scale for the importance of state parameter i relative to state parameter j. ji a is a scale for the importance of state parameter j relative to state parameter i. ii Let be a scale indicating the importance of state parameter i relative to state parameter i, where a ij >0, a ij =1 / a ji a ii =1; The judgment matrix A is a positive reciprocal matrix, which has a maximum eigenvalue and a unique eigenvalue. Normalizing the corresponding eigenvector M yields the weights of each state parameter.

7. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 5 or 6, characterized in that, In step S3, since the dimensions and orders of magnitude of the state parameters are different, all state parameters involved in the evaluation are normalized and the deviation is used to characterize the safety of the photovoltaic combiner box. When the actual measured value exceeds the normal operation limit, the deviation value is 1. When the actual measured value is equal to the standard value, the deviation is 0, and the state parameter is in the optimal state.

8. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 7, characterized in that, In step S3, deviation functions are established for the state parameters of the "larger the value, the better" and "smaller the value, the better" types, respectively: Where s(x) is the deviation function, and x represents the actual measured value of a single state parameter. max and x min These are the maximum and minimum critical values ​​specified in the relevant regulations.

9. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 8, characterized in that, In step S3, the health status of the photovoltaic combiner box is divided into four levels: normal, attention, abnormal, and severe. The membership function of each level is expressed as follows: Where s(x) is the deviation function, This is the normal status level. To pay attention to the status level, This is an abnormal state level. The severity level is [not specified]. By calculating the membership function of the state level corresponding to the state parameter, the membership function matrix P is obtained. m : In the formula, This indicates the membership degree of each indicator under different state levels; Further, we obtain the state evaluation vector U: U = W * P m W represents the weights calculated for each corresponding state parameter. Based on the calculated values ​​of each element in the state evaluation vector and their corresponding expected values, the security assessment result Q0 is calculated: Among them, U c Ex represents the calculated value of each element in the state evaluation vector. c This represents the expected degradation value corresponding to the c-th element.

10. The intelligent monitoring, early warning, and evaluation model for photovoltaic combiner boxes according to claim 9, characterized in that, In step S4, the fitness of an individual is represented as f. d : In the formula, K represents the coefficient, and y d This is the actual output of the BP neural network. Let z be the expected output of the BP neural network, and z be the number of individuals in the population. Next, a selection operation is performed, and the roulette wheel selection method is chosen. The selection probability p for each individual d is... d for: In the formula, f d Let d be the fitness of individual d, and z be the number of individuals in the population; Then, using the single-point real number crossover method, the crossover operation is performed between the m-th chromosome and the e-th chromosome at the f-th gene locus: a mf =a mf (1-b)+a ef b, a ef =a ef (1-b)+a mf b; In the formula, b is a random number between 0 and 1; The mutation operation is performed on the f-th gene of the e-th individual, and the formula is as follows: In the formula, a max and a min Gene a ef The upper and lower bounds of G, where g is the current iteration number, and G max The maximum number of evolutions is given by r, and r2 is a random number between 0 and 1.

11. A photovoltaic combiner box intelligent monitoring and early warning device, characterized in that, It includes a gas sensor, a temperature and humidity sensor, an intelligent monitoring module, an intelligent early warning module, a self-powered power supply, a controller, and a buzzer alarm. The intelligent early warning module includes the photovoltaic combiner box intelligent monitoring and early warning evaluation model according to any one of claims 1-10 for realizing safety evaluation. The self-powered power supply is connected to the intelligent monitoring module and the intelligent early warning module. The gas sensor and the temperature and humidity sensor are both connected to the intelligent monitoring module. The buzzer alarm is connected to the controller. The controller is connected to the intelligent early warning module. The intelligent early warning module is connected to the monitoring center.

12. The intelligent monitoring and early warning device for photovoltaic combiner boxes according to claim 11, characterized in that, Both the gas sensor and the temperature and humidity sensor are located inside the photovoltaic combiner box. The gas sensor is used to monitor the concentration of combustible gas, the proportion of combustible gas, and the oxygen concentration. The temperature and humidity sensor is used to monitor the humidity and the temperature inside the photovoltaic combiner box.

13. The intelligent monitoring and early warning device for photovoltaic combiner boxes according to claim 12, characterized in that, The intelligent monitoring module is used to collect data and calculate the minimum ignition energy.

Images