Photovoltaic module life cycle assessment method and system based on big data

Abstract

The present invention discloses a photovoltaic module life cycle assessment method and system based on big data, including collecting working data and monitoring data of photovoltaic modules, preprocessing the working data and the monitoring data; obtaining impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data characterizes the internal decay of photovoltaic modules; the second impact data characterizes the external loss coefficient of photovoltaic modules; life prediction is performed according to the working data to obtain predicted life, a photovoltaic module life cycle assessment model is constructed according to the impact data and the predicted life, the photovoltaic module life cycle assessment model is optimized, the data to be assessed is input into the photovoltaic module life cycle assessment model, and the assessment result is output. The method improves the accuracy of photovoltaic module life cycle assessment based on big data, has good interpretability, and can be directly applied to photovoltaic module life cycle assessment system.

Description

Technical Field

[0001] The present invention relates to the field of assessment, and in particular to a photovoltaic component life cycle assessment method and system based on big data. Background Art

[0002] As the global demand for renewable energy continues to grow, photovoltaic power generation, as a clean and renewable energy form, has become increasingly important. As the core component of the photovoltaic power generation system, the performance stability and service life of photovoltaic modules are directly related to the power generation efficiency and economic benefits of the entire system. However, during the long-term outdoor operation of photovoltaic modules, they will be affected by many factors, such as natural environmental conditions (such as temperature, humidity, light intensity, wind and sand, etc.), material aging, mechanical stress, etc. These factors will cause the performance of photovoltaic modules to gradually decline or even damage.

[0003] At present, the performance evaluation of photovoltaic modules mostly relies on traditional regular inspection and maintenance methods. This method is not only time-consuming and labor-intensive, but also difficult to accurately reflect the actual operating status and remaining life of photovoltaic modules in real time. In addition, with the rapid development of big data technology, all walks of life are actively exploring how to use big data to optimize decision-making and improve efficiency. In the photovoltaic field, how to effectively use the large amount of data generated during the operation of photovoltaic modules to build a scientific and accurate life cycle assessment model has become one of the hot topics in industry research.

[0004] Therefore, it is necessary to develop a photovoltaic module life cycle assessment method based on big data. This method can collect the working data and monitoring data of photovoltaic modules in real time and accurately, and through advanced preprocessing technology and data analysis algorithms, deeply explore the laws and information behind the data, and build a scientific, accurate, and dynamically adjustable photovoltaic module life cycle assessment model to achieve comprehensive, accurate, and real-time photovoltaic module life cycle assessment. Summary of the invention

[0005] The purpose of the present invention is to provide a photovoltaic module life cycle assessment method based on big data.

[0006] To achieve the above object, the present invention is implemented according to the following technical solutions:

[0007] The present invention comprises the following steps:

[0008] Collecting working data and monitoring data of photovoltaic components, and preprocessing the working data and the monitoring data;

[0009] Obtaining impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module;

[0010] Performing a life prediction based on the working data to obtain a predicted life, and constructing a photovoltaic component life cycle assessment model based on the impact data and the predicted life;

[0011] The photovoltaic component life cycle assessment model is optimized, the data to be assessed is input into the photovoltaic component life cycle assessment model, and an assessment result is output.

[0012] Furthermore, the method for obtaining the internal attenuation includes:

[0013] The internal impact data of the photovoltaic modules are input into the deep learning model to obtain the internal predicted lifespan, and the decision matrix is ​​constructed using the normalized internal impact data;

[0014] The correlation analysis between internal impact data and internal predicted lifespan is performed to obtain the correlation degree, which is expressed as:

[0015]

[0016] The zth internal influence data at the sth moment is The zth internal predicted life span at the sth time is H s,z , internal impact data With the internal predicted life H s,z The correlation degree is Internal impact data The probability of occurrence is Internal impact data With the internal predicted life H s,z The probability of simultaneous occurrence is The resolution coefficient is ζ, the upper limit of the monitoring time is n, and the importance of the zth internal impact data at the sth time is The zth interference constant at the sth moment is u s,z ;

[0017] Calculate the importance of internal impact data:

[0018]

[0019] The zth adjustment factor at the sth moment is The amount of internal impact data is The contribution of the zth internal impact data to the life reduction at the sth moment is

[0020] Calculate the decision value of internal impact data:

[0021]

[0022] The normalized internal impact data The zth decision value at the sth moment is

[0023] Calculate the inner decay based on the decision value:

[0024]

[0025] The first weight is The second weight is The zth internal predicted life span at time s-1 is H s+1,z , the internal decay of the zth internal impact data is θ z , the error coefficient is θ z .

[0026] Furthermore, the method for obtaining the external loss coefficient includes:

[0027] Calculate the information entropy of external impact data:

[0028]

[0029] The number of external influence data is α, and the i-th external influence data is g i , the i-th external influence data at the s-th time is g i,s , external impact data g i The information entropy of h(g i ), the upper limit of time is

[0030] Calculate the information gain rate:

[0031]

[0032] The information gain rate of external influence data is φ(g i ), the jth external influence data at the sth moment is g j,s , the jth external influence data at the s+1th moment is g j,s+1 , external impact data g j,s and external influence data g j,s+1 The offset is The i-th external influence data at the s+1th time is g i,s+1 , external impact data g i,s and external influence data g i,s+1 The offset is

[0033] Calculate the Gini index:

[0034]

[0035] The external impact data g iThe Gini index is ui(g i ), using a long short-term memory network to obtain the external predicted lifespan of the component;

[0036] The external loss coefficient is calculated according to the Gini index, and the expression is:

[0037]

[0038] The proportional control coefficient is ξ, and the i-th external predicted life The probability of At the same time, the i-th external predicted life span appears and external influence data g i The probability of External predicted lifespan and external influence data g i The correlation degree is The L2 norm function is ||·||2, and the external loss coefficient of the i-th test is

[0039] Furthermore, a method for performing life prediction according to the working data to obtain a predicted life includes:

[0040] Arrange the working data in chronological order and use it as input data. Calculate the attention weight of the input data in the encoder:

[0041]

[0042]

[0043] The attention weight of the w-th input data at the s-th time is The query vector is q, the number of input data is m, and the attention scoring function is The hyperbolic tangent function is tanh(·), transposed to T, and the weight matrix is ​​V z , the hidden layer state at the s-1th moment is f s-1 , the cell state at time s-1 is y s-1 , the weight vector is λ z , the weight matrix of the linear transformation is C z , the wth input data is b w , the wth input data at the sth time is The κth input data at the sth time is

[0044] Update the input data according to the attention weight, the expression is:

[0045]

[0046] The input data updated at time s is The number of input data is n, and the expression for updating the hidden layer state is:

[0047]

[0048] The nonlinear activation function is The updated hidden layer state at the sth moment is f s , calculate the decoder input:

[0049]

[0050] The hidden layer state weight at the sth moment is The wth hidden layer state is f w , the hidden layer state at time s-1 is d s-1 , the cell state at time s-1 is x s-1 , the cell state at time s is x s , the encoder weight vector is λ d , the decoder weight matrix is ​​λ d ;

[0051] The predicted life of the PV module is obtained based on the input of the decoder, and the expression is:

[0052]

[0053] The predicted life span at time s is The hidden state of the decoder at time s is d1 s , the predicted life span at the s-1th moment is The decoder bias is σ, the decoder weight is τ, and the weight vector bias is γ λ , the weight matrix bias is γ w , the actual life span at the s-1th moment is ρ s-1 .

[0054] Furthermore, a method for constructing a photovoltaic component life cycle assessment model based on the impact data and the predicted lifespan includes:

[0055] Construct the objective function:

[0056]

[0057] The objective function of the cth photovoltaic module is The spatial parameters of the cth photovoltaic module are The external loss coefficient of the cth photovoltaic module is The internal decay of the cth PV module is θ c , the predicted lifetime of the cth PV module is The actual lifetime of the cth PV module is ρ c,i , the shadow anomaly is The predicted lifetime of the cth PV module is The actual lifetime of the cth PV module is ρ c , predict life expectancy and actual life span ρ c The loss function is

[0058] The PV module life cycle assessment model includes a deviation analysis algorithm, an autoencoder, a deep learning matching algorithm, and a life cycle assessment method;

[0059] The deviation analysis algorithm evaluates the deviation between the predicted value and the actual value of the input data to obtain the deviation amount, and selects the input data according to the importance of the deviation amount to obtain important data;

[0060] The autoencoder compresses and reconstructs important data through unsupervised learning, extracting the influential features of key data;

[0061] The deep learning matching algorithm performs matching analysis based on the similarity between the influencing features and the influencing data to obtain matching data;

[0062] The life cycle assessment method is based on quantifying the working conditions, consumption and environmental load of photovoltaic components during their life cycle to conduct a life cycle assessment of matching data and important data, and taking the life cycle with the minimum objective function as the assessment result.

[0063] Furthermore, the method for optimizing the photovoltaic module life cycle assessment model includes:

[0064] The prey captured by the searcher is regarded as the optimal solution, and the position of the searcher is regarded as the solution. The position of the searcher is updated by the modulus of the random number and the coefficient vector. The expression is:

[0065]

[0066] β=2r2

[0067] The search subposition of the t+1th iteration is The coefficient vector is β, the random numbers between 0 and 1 are r, r1, r2, the random number between -1 and 1 is v1, and the prey position of the tth iteration is The spiral coefficient is μ, the current number of iterations is t, and the maximum number of iterations is t max , search sub-position for

[0068] Update the spiral coefficient, the expression is:

[0069]

[0070] When r<0.5, the search sub-position is updated according to the random position to obtain the adjusted search sub-position, and the expression is:

[0071]

[0072]

[0073] The random position of the searcher in the tth iteration is The adjusted search subposition for the t+1th iteration is

[0074] The step coefficient is introduced, and the search sub-position is updated and adjusted according to the step coefficient to obtain the optimized search sub-position. The expression is:

[0075]

[0076]

[0077] The optimized search subposition for the t+1th iteration is The adjusted search subposition for the tth iteration is The search subspeed of the tth iteration is The step size coefficient for the tth iteration is The upper limit of the search range is The lower limit of the search interval is The maximum step size factor is The minimum step size coefficient is

[0078] Continue to iterate until the maximum number of iterations is reached, otherwise update the spiral coefficient and re-select the optimal solution.

[0079] The second aspect is the photovoltaic module life cycle assessment system based on big data, including:

[0080] Data acquisition module: used to collect working data and monitoring data of photovoltaic components, and pre-process the working data and monitoring data;

[0081] Data processing module: used to obtain impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module;

[0082] Prediction construction module: used for performing life prediction according to the working data to obtain predicted life, and constructing a photovoltaic component life cycle assessment model according to the impact data and the predicted life;

[0083] Constructing an optimization module: for optimizing the photovoltaic component life cycle assessment model, inputting the data to be assessed into the photovoltaic component life cycle assessment model, and outputting the assessment result.

[0084] The beneficial effects of the present invention are:

[0085] The present invention is a photovoltaic module life cycle assessment method and system based on big data. Compared with the prior art, the present invention has the following technical effects:

[0086] The present invention can improve the accuracy of photovoltaic module life cycle assessment through preprocessing, obtaining impact data, life prediction, model construction and model optimization steps, thereby improving the precision of photovoltaic module life cycle assessment, optimizing photovoltaic module life cycle assessment, greatly saving resources, and improving work efficiency, and can realize intelligent assessment of photovoltaic module life cycle, and perform multi-data input for photovoltaic module life cycle assessment in real time, which is of great significance to photovoltaic module life cycle assessment, can adapt to photovoltaic module life cycle assessment of different standards and different photovoltaic module life cycle assessment needs, and has a certain universality. BRIEF DESCRIPTION OF THE DRAWINGS

[0087] Figure 1 This is a flowchart of the steps of the photovoltaic module life cycle assessment method based on big data of the present invention. DETAILED DESCRIPTION

[0088] The present invention is further described below by means of specific embodiments. The illustrative embodiments and descriptions of the present invention are used to explain the present invention but are not intended to limit the present invention.

[0089] The photovoltaic module life cycle assessment method and system based on big data of the present invention include the following steps:

[0090] like Figure 1 As shown, in this embodiment, the following steps are included:

[0091] Collecting working data and monitoring data of photovoltaic components, and preprocessing the working data and the monitoring data;

[0092] In the actual evaluation, the working data are energy consumption, greenhouse gas emissions, production process, equipment efficiency, power generation efficiency, and power output; the monitoring data are irradiance, temperature, wind speed, humidity, quality control, installation location, installation method, installation time, fault repair, cleaning and maintenance, service time, retirement reason, remaining life, wind and sand, snow load, light time, light intensity, and weather conditions;

[0093] The photovoltaic modules installed in XXX factory in June 2018 were taken as the research objects;

[0094] Obtaining impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module;

[0095] In the actual evaluation, the internal decay and external loss coefficients of the PV modules in September 2021 were 0.259 and 0.326 respectively;

[0096] Performing a life prediction based on the working data to obtain a predicted life, and constructing a photovoltaic component life cycle assessment model based on the impact data and the predicted life;

[0097] In actual assessments, the predicted lifetime of PV panels in September 2021 was 15 years and 7 months;

[0098] The data to be evaluated is input into the photovoltaic module life cycle evaluation model, and the evaluation result is output.

[0099] In this embodiment, the method for obtaining the internal attenuation includes:

[0100] The internal impact data of the photovoltaic modules are input into the deep learning model to obtain the internal predicted lifespan, and the decision matrix is ​​constructed using the normalized internal impact data;

[0101] The correlation analysis between internal impact data and internal predicted lifespan is performed to obtain the correlation degree, which is expressed as:

[0102]

[0103] The zth internal influence data at the sth moment is The zth internal predicted life span at the sth time is H s,z , internal impact data With the internal predicted life H s,z The correlation degree is Internal impact data The probability of occurrence is Internal impact data With the internal predicted life H s,z The probability of simultaneous occurrence is The resolution coefficient is ζ, and the upper limit of monitoring time is The importance of the zth internal influence data at the sth time is The zth interference constant at the sth moment is u s,z ;

[0104] Calculate the importance of internal impact data:

[0105]

[0106] The zth adjustment factor at the sth moment is The amount of internal impact data is The contribution of the zth internal impact data to the life reduction at the sth moment is

[0107] Calculate the decision value of internal impact data:

[0108]

[0109] The normalized internal impact data The zth decision value at the sth moment is

[0110] Calculate the inner decay based on the decision value:

[0111]

[0112] The first weight is The second weight is The zth internal predicted life span at time s-1 is H s+1,z , the internal decay of the zth internal impact data is θ z , the error coefficient is θ z .

[0113] In this embodiment, the method for obtaining the external loss coefficient includes:

[0114] Calculate the information entropy of external impact data:

[0115]

[0116] The number of external influence data is α, and the i-th external influence data is g i , the i-th external influence data at the s-th time is g i,s , external impact data g i The information entropy of i ), the upper limit of time is

[0117] Calculate the information gain rate:

[0118]

[0119] The information gain rate of external influence data is φ(g i ), the jth external influence data at the sth moment is g j,s , the jth external influence data at the s+1th moment is g j,s+1 , external impact data g j,s and external influence data g j,s+1 The offset is The i-th external influence data at the s+1th time is gi,s+1 , external impact data g i,s and external influence data g i,s+1 The offset is

[0120] Calculate the Gini index:

[0121]

[0122] The external impact data g i The Gini index is ui(g i ), using a long short-term memory network to obtain the external predicted lifespan of the component;

[0123] The external loss coefficient is calculated according to the Gini index, and the expression is:

[0124]

[0125] The proportional control coefficient is ξ, and the i-th external predicted life The probability of At the same time, the i-th external predicted life span appears and external influence data g i The probability of External predicted lifespan and external influence data g i The correlation degree is The L2 norm function is ||·||2, and the external loss coefficient of the i-th test is

[0126] In this embodiment, the method for performing life prediction according to the working data to obtain the predicted life includes:

[0127] Arrange the working data in chronological order and use it as input data. Calculate the attention weight of the input data in the encoder:

[0128]

[0129]

[0130] The attention weight of the w-th input data at the s-th time is The query vector is q, the number of input data is m, and the attention scoring function is The hyperbolic tangent function is tanh(·), transposed to T, and the weight matrix is ​​V z , the hidden layer state at the s-1th moment is f s-1 , the cell state at time s-1 is y s-1 , the weight vector is λ z , the weight matrix of the linear transformation is C z, the wth input data is b w , the wth input data at the sth time is The κth input data at the sth time is

[0131] Update the input data according to the attention weight, the expression is:

[0132]

[0133] The input data updated at time s is The number of input data is n, and the expression for updating the hidden layer state is:

[0134]

[0135] The nonlinear activation function is The updated hidden layer state at the sth moment is f s , calculate the decoder input:

[0136]

[0137] The hidden layer state weight at the sth moment is The wth hidden layer state is f w , the hidden layer state at the s-1th moment is d s-1 , the cell state at time s-1 is x s-1 , the cell state at time s is x s , the encoder weight vector is λ d , the decoder weight matrix is ​​λ d ;

[0138] The predicted life of the PV module is obtained based on the input of the decoder, and the expression is:

[0139]

[0140] The predicted life span at time s is The hidden state of the decoder at time s is d1 s , the predicted life span at the s-1th moment is The decoder bias is σ, the decoder weight is τ, and the weight vector bias is γ λ , the weight matrix bias is γ w , the actual life span at the s-1th moment is ρ s-1 .

[0141] In this embodiment, the method for constructing a photovoltaic component life cycle assessment model based on the impact data and the predicted lifespan includes:

[0142] Construct the objective function:

[0143]

[0144] The objective function of the cth photovoltaic module is The spatial parameters of the cth photovoltaic module are The external loss coefficient of the cth photovoltaic module is The internal decay of the cth PV module is θ c , the predicted lifetime of the cth PV module is The actual lifetime of the cth PV module is ρ c,i , the shadow anomaly is The predicted lifetime of the cth PV module is The actual lifetime of the cth PV module is ρ c , predict life expectancy and actual life span ρ c The loss function is

[0145] The PV module life cycle assessment model includes a deviation analysis algorithm, an autoencoder, a deep learning matching algorithm, and a life cycle assessment method;

[0146] The deviation analysis algorithm evaluates the deviation between the predicted value and the actual value of the input data to obtain the deviation amount, and selects the input data according to the importance of the deviation amount to obtain important data;

[0147] The autoencoder compresses and reconstructs important data through unsupervised learning, extracting the influential features of key data;

[0148] The deep learning matching algorithm performs matching analysis based on the similarity between the influencing features and the influencing data to obtain matching data;

[0149] The life cycle assessment method is based on quantifying the working conditions, consumption and environmental load of photovoltaic components during their life cycle to conduct a life cycle assessment of matching data and important data, and taking the life cycle with the minimum objective function as the assessment result.

[0150] In this embodiment, the method for optimizing the photovoltaic module life cycle assessment model includes:

[0151] The prey captured by the searcher is regarded as the optimal solution, and the position of the searcher is regarded as the solution. The position of the searcher is updated by the modulus of the random number and the coefficient vector. The expression is:

[0152]

[0153] β=2r2

[0154] The search subposition of the t+1th iteration is The coefficient vector is β, the random numbers between 0 and 1 are r, r1, r2, the random number between -1 and 1 is v1, and the prey position of the tth iteration is The spiral coefficient is μ, the current number of iterations is t, and the maximum number of iterations is t max , search sub-position for

[0155] Update the spiral coefficient, the expression is:

[0156]

[0157] When r<0.5, the search sub-position is updated according to the random position to obtain the adjusted search sub-position, and the expression is:

[0158]

[0159]

[0160] The random position of the searcher in the tth iteration is The adjusted search subposition for the t+1th iteration is

[0161] The step coefficient is introduced, and the search sub-position is updated and adjusted according to the step coefficient to obtain the optimized search sub-position. The expression is:

[0162]

[0163]

[0164] The optimized search subposition for the t+1th iteration is The adjusted search subposition for the tth iteration is The search subspeed of the tth iteration is The step size coefficient for the tth iteration is The upper limit of the search range is The lower limit of the search interval is The maximum step size factor is The minimum step size coefficient is

[0165] Continue to iterate until the maximum number of iterations is reached, otherwise update the spiral coefficient and re-select the optimal solution.

[0166] The second aspect is the photovoltaic module life cycle assessment system based on big data, including:

[0167] Data acquisition module: used to collect working data and monitoring data of photovoltaic components, and pre-process the working data and monitoring data;

[0168] Data processing module: used to obtain impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module;

[0169] Prediction construction module: used for performing life prediction according to the working data to obtain predicted life, and constructing a photovoltaic component life cycle assessment model according to the impact data and the predicted life;

[0170] Constructing an optimization module: for optimizing the photovoltaic component life cycle assessment model, inputting the data to be assessed into the photovoltaic component life cycle assessment model, and outputting the assessment result.

[0171] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A photovoltaic module life cycle assessment method based on big data, characterized in that: The following steps are involved: Collecting working data and monitoring data of photovoltaic components, and preprocessing the working data and the monitoring data; Obtaining impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module; Performing a life prediction according to the working data to obtain a predicted life, and constructing a photovoltaic component life cycle assessment model according to the impact data and the predicted life; including: Construct the objective function: The objective function of the cth photovoltaic module is , the spatial parameters of the cth photovoltaic module are , the external loss coefficient of the cth photovoltaic module is , the internal decay of the cth photovoltaic module is , the predicted lifetime of the cth PV module is , the actual life of the i-th PV module of the c-th PV module is , the shadow anomaly is , the predicted lifetime of the cth PV module is , the actual life of the cth photovoltaic module is , predict life expectancy and actual lifespan The loss function is ; The PV module life cycle assessment model includes a deviation analysis algorithm, an autoencoder, a deep learning matching algorithm, and a life cycle assessment method; The deviation analysis algorithm evaluates the variance between the predicted value and the actual value of the input data to obtain the deviation amount, and selects the input data according to the importance of the deviation amount to obtain important data; The autoencoder compresses and reconstructs important data through unsupervised learning, extracting the influential features of key data; The deep learning matching algorithm performs matching analysis based on the similarity between the influencing features and the influencing data to obtain matching data; The life cycle assessment method is based on quantifying the working conditions, consumption and environmental load of photovoltaic modules during their life cycle to evaluate the matching data and important data, and taking the life cycle with the minimum objective function as the evaluation result; The photovoltaic component life cycle assessment model is optimized, the data to be assessed is input into the photovoltaic component life cycle assessment model, and an assessment result is output.

2. The photovoltaic module life cycle assessment method based on big data according to claim 1 is characterized in that: The method for obtaining the internal attenuation comprises: The internal impact data of the photovoltaic modules are input into the deep learning model to obtain the internal predicted lifespan, and the decision matrix is ​​constructed using the normalized internal impact data; The correlation analysis between internal impact data and internal predicted lifespan is performed to obtain the correlation degree, which is expressed as: Among them, the The internal impact data is , the sth time The internal predicted lifespan is , internal impact data Internal predicted lifespan The correlation degree is , internal impact data The probability of occurrence is , the resolution coefficient is The upper limit of monitoring time is , at the sth moment The importance of internal impact data is , the sth time The interference constant is ; Calculate the importance of internal impact data: Among them, the The adjustment factor is , the number of internal impact data is , No. The contribution of the life reduction of the internal impact data at the sth moment is ; Calculate the decision value of internal impact data: The normalized internal impact data is , the sth time The decision value is ; Calculate the inner decay based on the decision value: The first weight is , the second weight is , the first The internal predicted lifespan is , No. The internal attenuation of the internal impact data is , the error coefficient is .

3. The photovoltaic module life cycle assessment method based on big data according to claim 1 is characterized in that: The method for obtaining the external loss coefficient includes: Calculate the information entropy of external impact data: The number of external impact data is , the i-th external impact data is , the i-th external influence data at the s-th time is , external impact data The information entropy is , the time limit is ; Calculate the information gain rate: The information gain rate of external influence data is , the jth external influence data at the sth moment is , the jth external influence data at the s+1th moment is , external impact data and external influence data The offset is , the i-th external influence data at the s+1th time is , external impact data and external influence data The offset is ; Calculate the Gini index: The i-th external influence data The Gini index is , using long short-term memory networks to obtain the external predicted lifetime of components; The external loss coefficient is calculated according to the Gini index, and the expression is: The proportional control coefficient is , the i-th external predicted life span appears The probability of , and the i-th external predicted life span appears and external influence data The probability of , external predicted lifespan and external influence data The correlation degree is , the L2 norm function is , the external loss coefficient of the i-th test is .

4. The photovoltaic module life cycle assessment method based on big data according to claim 1 is characterized in that: The method for performing life prediction according to the working data to obtain the predicted life includes: Arrange the working data in chronological order and use it as input data. Calculate the attention weight of the input data in the encoder: The attention weight of the w-th input data at the s-th time is , the query vector is q, the number of input data is m, and the attention scoring function at the sth moment is , the hyperbolic tangent function is , transposed to T, the weight matrix is , the hidden layer state at the s-1th moment is , the cell state at time s-1 is , the weight vector is , the weight matrix of the linear transformation is , the wth input data is , the wth input data at the sth time is , at the sth moment The input data is ; Update the input data according to the attention weight, the expression is: The input data updated at time s is , the number of input data is n, and the hidden layer state is updated. The expression is: The nonlinear activation function is , the updated hidden layer state at the sth moment is , calculate the decoder input: The wth hidden layer state is , the hidden layer state at the s-1th moment is , the cell state at time s-1 is , the cell state at time s is , the encoder weight vector is , the decoder weight matrix is ; The predicted life of the PV module is obtained based on the input of the decoder, and the expression is: The predicted life span at time s is , the hidden state of the decoder at time s is , the predicted life span at the s-1th moment is , the decoder bias is , the decoder weight is , the weight vector bias is , the weight matrix bias is , the actual life span at the s-1th moment is .

5. The photovoltaic module life cycle assessment method based on big data according to claim 1 is characterized in that: The method for optimizing the photovoltaic module life cycle assessment model comprises: The prey captured by the searcher is regarded as the optimal solution, and the position of the searcher is regarded as the solution. The position of the searcher is updated by the modulus of the random number and the coefficient vector. The expression is: The search subposition of the t+1th iteration is , the coefficient vector is , the random numbers between 0 and 1 are r, , , a random number between -1 and 1 is , the prey position at the tth iteration is , the spiral coefficient is , the current number of iterations is t, and the maximum number of iterations is , search sub-position for ; Update the spiral coefficient, the expression is: when When , the search sub-position is updated according to the random position to obtain the adjusted search sub-position, the expression is: The random position of the searcher in the tth iteration is , the adjusted search subposition for the t+1th iteration is ; The step coefficient is introduced, and the search sub-position is updated and adjusted according to the step coefficient to obtain the optimized search sub-position. The expression is: The optimized search subposition for the t+1th iteration is , the adjusted search subposition for the tth iteration is , the search subspeed of the tth iteration is , the step size coefficient of the tth iteration is , the upper limit of the search interval is , the lower limit of the search interval is , the maximum step size coefficient is , the minimum step size coefficient is ; Continue to iterate until the maximum number of iterations is reached, otherwise update the spiral coefficient and re-select the optimal solution.

6. A photovoltaic module life cycle assessment system based on big data, used to execute the method described in any one of claims 1 to 5, characterized in that: include: Data acquisition module: used to collect working data and monitoring data of photovoltaic components, and pre-process the working data and monitoring data; Data processing module: used to obtain impact data through the monitoring data; the impact data includes first impact data and second impact data; the first impact data represents the internal decay of the photovoltaic module; the second impact data represents the external loss coefficient of the photovoltaic module; Prediction construction module: used for performing life prediction according to the working data to obtain predicted life, and constructing a photovoltaic component life cycle assessment model according to the impact data and the predicted life; Constructing an optimization module: for optimizing the photovoltaic component life cycle assessment model, inputting the data to be assessed into the photovoltaic component life cycle assessment model, and outputting the assessment result.

Images