Photovoltaic curtain wall system power generation efficiency prediction method and system

Abstract

The invention discloses a photovoltaic curtain wall system power generation efficiency prediction method and system, and the method comprises the steps: collecting environment data and power generation power data of a photovoltaic curtain wall system, carrying out the data integration through a time synchronization technology, and obtaining an original data set; performing data preprocessing operation on the original data set to obtain a target data set; constructing an initial power generation efficiency prediction model based on an LSTM neural network, and performing iterative optimization on the initial power generation efficiency prediction model by using an improved WOA algorithm through the target data set to obtain a target power generation efficiency prediction model based on an improved WOA-LSTM neural network model; and performing real-time prediction on the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model to obtain a prediction result. According to the invention, the accuracy and reliability of power generation efficiency prediction of the photovoltaic curtain wall system are effectively improved.

Description

Technical Field

[0001] The present invention belongs to the field of power generation prediction of building photovoltaic integrated curtain wall systems, and in particular relates to a method and system for predicting power generation efficiency of photovoltaic curtain wall systems. Background Art

[0002] In the global renewable energy utilization landscape, solar photovoltaic power generation occupies an important position. With the continuous growth of global demand for clean energy and the in-depth implementation of the concept of sustainable development, building integrated photovoltaics (BIPV), as a technical model that organically combines solar photovoltaic power generation systems with buildings, is gradually becoming an important development direction in the construction field and renewable energy utilization. Accurate prediction of photovoltaic power generation systems is crucial for the stable operation of power systems, energy scheduling, and maximizing the economic benefits of the system. Current power generation prediction methods mainly include physical methods, statistical methods, and machine learning methods. In practical applications, it is often necessary to use a combination of multiple methods to improve the accuracy and reliability of predictions. However, due to the randomness of meteorological conditions, the uncertainty of photovoltaic system performance, and data quality and integrity, there are still certain errors in current power generation predictions. When dealing with extreme weather or rapidly changing weather conditions, the improvement of prediction accuracy faces great challenges. This is also one of the key technical issues that need to be solved in the current widespread and efficient application of photovoltaic power generation systems in the field of buildings.

[0003] The "14th Five-Year Plan for Building Energy Conservation and Green Building Development" issued by the Ministry of Housing and Urban-Rural Development has clarified the goals and paths for the development of green buildings and promoted the widespread application of photovoltaic curtain wall systems for building photovoltaic integration. The application of these technologies has not only improved the energy self-sufficiency of buildings, but also provided new ideas for urban sustainable development. Existing research mainly focuses on the prediction of distributed rooftop photovoltaic power generation. In view of the complex and multi-factor influence of the building photovoltaic curtain wall generation point system, the drastic changes in environmental variables, and the difficulty in quantifying the internal influence mechanism, the existing photovoltaic curtain wall system power generation prediction methods have many shortcomings. At the same time, the traditional physical model prediction method requires precise physical parameters and is complex in calculation, which makes it difficult to adapt to the complex and changeable actual environment; the statistical method has poor adaptability to abnormal weather conditions; conventional machine learning methods such as ordinary neural networks are difficult to capture long-term and short-term dependencies when processing time series data, resulting in limited prediction accuracy and easy to fall into local optimal solutions, which cannot meet the needs of high-precision prediction of photovoltaic curtain wall power generation efficiency. Summary of the invention

[0004] In order to solve the above technical problems, the present invention provides a method and system for predicting the power generation efficiency of a photovoltaic curtain wall system. A method for predicting the power generation efficiency of a photovoltaic curtain wall system includes:

[0005] Collect the environmental data and power generation data of the photovoltaic curtain wall system, and integrate the data through time synchronization technology to obtain the original data set;

[0006] Perform data preprocessing operations on the original data set to obtain the target data set;

[0007] Construct an initial power generation efficiency prediction model based on the LSTM neural network, and use the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model;

[0008] Based on the target power generation efficiency prediction model, perform real-time prediction on the power generation efficiency of the photovoltaic curtain wall system to obtain the prediction result.

[0009] Preferably, the process of collecting the environmental data and power generation data of the photovoltaic curtain wall system includes:

[0010] Collect environmental data through an environmental data monitoring station built on the photovoltaic curtain wall system;

[0011] Build a photovoltaic curtain wall system by connecting photovoltaic curtain wall components in series and parallel, uniformly connect the photovoltaic curtain wall components to an inverter group, obtain real-time power generation data, and integrate the power generation data with environmental data through time synchronization technology to obtain the original data set;

[0012] Among them, the environmental data monitoring station includes a light intensity sensor, a component temperature sensor, an environmental temperature sensor, a humidity sensor, a wind speed sensor, and a pressure sensor;

[0013] The light intensity sensor is used to collect and obtain direct solar irradiance, diffuse irradiance, and reflected irradiance data;

[0014] The component temperature sensor is used to be set on the back of the photovoltaic curtain wall component to collect component temperature change data.

[0015] Preferably, the process of performing data preprocessing operations on the original data set includes:

[0016] Adopt a method based on the 3σ principle to remove outliers in the original data set. Data points outside the range of the mean ± 3 times the standard deviation are regarded as outliers and excluded;

[0017] For the missing values in the original data set, use linear interpolation to fill them, and then perform normalization processing on all the data to map it to the interval of [0, 1];

[0018] Among them, the formula expression of the normalization processing is:

[0019]

[0020] Among them, x norm is the normalized data, x is the original data, x max and x min are the maximum and minimum values in the original data respectively.

[0021] Preferably, the process of iteratively optimizing the initial power generation efficiency prediction model by using the improved WOA algorithm through the target data set includes:

[0022] Using the improved WOA algorithm to optimize the parameters of the LSTM neural network, encoding the weight matrix and bias vector of the LSTM neural network as the whale individual position vector in the WOA algorithm; among them, the weight matrix of the LSTM neural network includes the input weight, forgetting weight, output weight, and the weight from the hidden layer to the output layer;

[0023] Taking the prediction error as the fitness function of the WOA algorithm, through the iterative search of the improved WOA algorithm, using the mean square error as the evaluation index to obtain the parameter combination with the minimum fitness function, and determining the optimal LSTM neural network model structure and parameters;

[0024] During the optimization process, sort the whale individuals according to the fitness value, update the position and adjust the parameters of the individuals through the fitness value. After multiple iterations, obtain the optimal model parameters, and then construct the target power generation efficiency prediction model based on the improved WOA-LSTM neural network model.

[0025] Preferably, the improved WOA algorithm modifies the original update method in the position update formula of the algorithm by introducing an adaptive weight factor: X new = w·X rand - A·D;

[0026] Among them, w is the weight factor, X new is the updated position, X rand is the randomly selected individual position, A is the convergence factor, and D is the distance between the individual and the current optimal individual;

[0027] The weight factor w is adaptively adjusted according to the iteration times; the formula expression is:

[0028]

[0029] where w max and w min are the maximum and minimum values of the weight factor respectively, iter is the current iteration times, and iter max is the maximum iteration times.

[0030] Preferably, the improved WOA algorithm further includes a method of initializing the population with chaos, using the Logistic chaotic mapping to generate the initial population;

[0031] The process of using the Logistic chaotic mapping to generate the initial population in the method of initializing the population with chaos includes:

[0032] Randomly generate an initial value z0 within the parameter range of the chaotic mapping, and then generate a series of chaotic values through the iterative formula;

[0033] Map the chaotic values to the range of the search space as the initial population of the WOA algorithm;

[0034] Among them, the iterative formula is:

[0035] z n+1 = μ·z n ·(1 - z n )

[0036] Among them, μ is the chaotic parameter.

[0037] Preferably, the LSTM neural network structure includes an input layer, a hidden layer, and an output layer;

[0038] The input layer is used to receive the preprocessed environmental data and historical power generation data;

[0039] The hidden layer includes several LSTM units, which are used to learn the time series features and long-term dependence relationships in the data;

[0040] The output layer is used to output the predicted power generation value of the photovoltaic curtain wall system.

[0041] Preferably, the internal structure of the LSTM unit includes a forget gate, an input gate, an output gate, and a memory unit;

[0042] Through the combination of the forget gate, the input gate, the output gate, and the memory unit, the LSTM model network is allowed to selectively remember, forget, and output information when processing sequential data;

[0043] In the unit structure of the LSTM unit, the information that can be stored in the memory unit is determined by the input gate, and then the memory unit updates the memory unit at the current moment with a weight. The amount of information retained in the memory unit at the previous moment is determined by the forget gate, and finally the hidden state h at the current moment is determined by the output gate; t ;

[0044] The structural formula expression of the LSTM unit is:

[0045] f t = σ(Wc · [h t-1 σ(W f · [h t-1 , x t + b ρ )])

[0046] i t =σ(W c · [h t-1 , x t + b i )

[0047] C t =ft·C t -1 + i t · [tanh*W c · [h t-1 , x t + b c )]

[0048] O t =σ(W o · [h t-1 , x t + b c )

[0049] h t =O t · tanh(Ct)

[0050] Among them, f t is the output of the forget gate, i t is the output of the input gate, C t is the updated memory cell, O t is the output of the output gate, h t is the hidden state at the current moment.

[0051] Preferably, the process of real-time predicting the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model includes:

[0052] Input the real-time collected and preprocessed environmental data and historical power generation power data into the target power generation efficiency prediction model to predict the power generation power of the photovoltaic curtain wall system and obtain the actually predicted power generation power;

[0053] Calculate the predicted value of the power generation efficiency according to the rated power of the photovoltaic curtain wall system and the actually predicted power generation power;

[0054] Among them, the formula expression of the power generation efficiency is:

[0055]

[0056] Among them, η is the power generation efficiency, Ppred is the predicted power generation, P rated is the rated power of the photovoltaic curtain wall system.

[0057] The present invention also provides a power generation efficiency prediction system for a photovoltaic curtain wall system, including:

[0058] A data acquisition and preprocessing module, configured to collect the environmental data and power generation data of the photovoltaic curtain wall system, and perform data integration through time synchronization technology to obtain an original data set; and perform data preprocessing operations on the original data set to obtain a target data set;

[0059] A model construction and optimization module, connected to the data acquisition and preprocessing module, configured to construct an initial power generation efficiency prediction model based on the LSTM neural network, and use the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model;

[0060] A power generation efficiency prediction module, connected to the model construction and optimization module, configured to perform real-time prediction on the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model to obtain a prediction result.

[0061] Compared with the prior art, the present invention has the following advantages and technical effects:

[0062] Through steps such as data acquisition and preprocessing, optimization of the improved WOA algorithm, construction and optimization of the LSTM neural network model, and power generation efficiency prediction and model update, the present invention effectively improves the accuracy and reliability of the power generation efficiency prediction of the photovoltaic curtain wall system, has broad application prospects and practical value, and can provide important technical support for the development of the building photovoltaic integration industry. Description of the Drawings

[0063] The drawings constituting a part of this application are used to provide a further understanding of this application. The schematic embodiments of this application and their descriptions are used to explain this application and do not constitute an improper limitation to this application. In the drawings:

[0064] Figure 1 is a schematic flow chart of the WOA-LSTM model according to an embodiment of the present invention;

[0065] Figure 2 is a schematic diagram of the unit structure of the LSTM model according to an embodiment of the present invention;

[0066] Figure 3 is a graph of the output prediction result of the photovoltaic curtain wall system based on the WOA-LSTM model according to an embodiment of the present invention;

[0067] Figure 4This is the optimized output effect diagram of the photovoltaic curtain wall system based on the WOA-LSTM model in the embodiments of the present invention. Specific Embodiments

[0068] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will detail this application with reference to the accompanying drawings and in combination with the embodiments.

[0069] It should be noted that the steps shown in the flowchart of the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although the logical order is shown in the flowchart, in some cases, the steps shown or described can be executed in a different order than here.

[0070] Embodiment 1

[0071] As Figures 1-4 shown, this embodiment provides a method for predicting the power generation efficiency of a photovoltaic curtain wall system, including the following steps:

[0072] Collect the environmental data and power generation power data of the photovoltaic curtain wall system, and integrate the data through time synchronization technology to obtain the original data set;

[0073] Perform data preprocessing operations on the original data set to obtain the target data set;

[0074] Construct an initial power generation efficiency prediction model based on the LSTM neural network, and use the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model;

[0075] Based on the target power generation efficiency prediction model, perform real-time prediction on the power generation efficiency of the photovoltaic curtain wall system to obtain the prediction result.

[0076] Further, the process of collecting the environmental data and power generation power data of the photovoltaic curtain wall system includes:

[0077] Collect environmental data through the environmental data monitoring station built on the photovoltaic curtain wall system;

[0078] Build a photovoltaic curtain wall system by connecting photovoltaic curtain wall components in series and parallel, uniformly connect the photovoltaic curtain wall components to the inverter group, obtain real-time power generation power data, and integrate the power generation power data with the environmental data through time synchronization technology to obtain the original data set;

[0079] Among them, the environmental data monitoring station is generally built on the roof of the building photovoltaic curtain wall system, and is equipped with multiple high-precision sensors, including light intensity sensors, component temperature sensors, environmental temperature sensors, humidity sensors, wind speed sensors, and pressure sensors, etc., to ensure that environmental data can be comprehensively and accurately obtained.

[0080] The light intensity sensor is used to collect and obtain direct solar irradiance, diffuse irradiance, and reflected irradiance data;

[0081] At the same time, component temperature sensors are arranged on the back of the photovoltaic curtain wall components one by one to collect component temperature change data. The sampling frequency of the sensors is set to collect once every 15 minutes to ensure the continuity and timeliness of the data.

[0082] Furthermore, the process of performing data preprocessing operations on the original data set includes:

[0083] Adopt a method based on the 3σ principle to remove outliers in the original data set. Data points outside the range of the mean ± 3 times the standard deviation are regarded as outliers and removed;

[0084] For the missing values in the original data set, linear interpolation method is used to fill them to ensure the integrity of the data. Finally, all data is normalized and mapped to the interval of [0, 1];

[0085] Among them, the formula expression of the normalization process is:

[0086]

[0087] where, x norm is the normalized data, x is the original data, x max and x min are the maximum and minimum values in the original data respectively.

[0088] Through data preprocessing, the quality and stability of the data are improved, providing a good data foundation for subsequent model training.

[0089] Furthermore, the process of iteratively optimizing the initial power generation efficiency prediction model by using the improved WOA algorithm through the target data set includes:

[0090] Use the improved WOA algorithm to optimize the parameters of the LSTM neural network, and encode the weight matrix and bias vector of the LSTM neural network as the whale individual position vector in the WOA algorithm; among them, the weight matrix of the LSTM neural network includes input weight, forgetting weight, output weight, and the weight from the hidden layer to the output layer;

[0091] Taking the prediction error as the fitness function of the WOA algorithm, through the iterative search of the improved WOA algorithm, using the mean square error as the evaluation index to obtain the parameter combination with the minimum fitness function, and determining the optimal LSTM neural network model structure and parameters;

[0092] During the optimization process, the whale individuals are sorted according to the fitness value, and the position update and parameter adjustment of the individuals are carried out through the fitness value. After multiple iterations, the optimal model parameters are obtained, and then the target power generation efficiency prediction model based on the improved WOA-LSTM neural network model is constructed.

[0093] Furthermore, in order to improve the global search ability and convergence speed of the WOA (Whale Optimization Algorithm), the improved WOA algorithm modifies the original update method in the position update formula of the algorithm by introducing an adaptive weight factor: X new = w·X rand - A·D;

[0094] where w is the weight factor, X new is the updated position, X rand is the position of a randomly selected individual, A is the convergence factor, and D is the distance between the individual and the current optimal individual;

[0095] The weight factor w is adaptively adjusted according to the number of iterations; the formula expression is:

[0096]

[0097] where w max and w min are the maximum and minimum values of the weight factor respectively, iter is the current number of iterations, and iter max is the maximum number of iterations.

[0098] At the initial stage of the algorithm, w is larger, enabling the whale individuals to explore in a wider solution space. As the iteration progresses, w gradually decreases, enhancing the local search ability of the algorithm, thus better balancing global search and local search and avoiding premature convergence to local optimal solutions.

[0099] Furthermore, the improved WOA algorithm also includes the method of using chaotic initialization of the population, generating the initial population using the Logistic chaotic mapping;

[0100] The process of generating the initial population using the method of chaotic initialization of the population and using the Logistic chaotic mapping includes:

[0101] Randomly generate an initial value \(z_0\) within the parameter range of the chaotic map, and then generate a series of chaotic values through the iterative formula;

[0102] Map the chaotic values to the range of the search space as the initial population of the WOA algorithm; through chaotic initialization, the initial population is more evenly distributed in the search space, improving the global search ability and convergence speed of the algorithm and enhancing the possibility of jumping out of the local optimum.

[0103] Among them, the iterative formula is:

[0104] z n+1 = μ·z n ·(1 - z n )

[0105] Among them, μ is the chaotic parameter.

[0106] Furthermore, the LSTM neural network structure includes an input layer, a hidden layer, and an output layer;

[0107] The input layer is used to receive the pre - processed environmental data and historical power generation data;

[0108] The hidden layer includes several LSTM units, which are used to learn the time - series features and long - term dependence relationships in the data;

[0109] The output layer is used to output the predicted power generation value of the photovoltaic curtain wall system.

[0110] Furthermore, the internal structure of the LSTM unit includes a forget gate, an input gate, an output gate, and a memory unit;

[0111] In the internal structure of the LSTM unit, the flow and update of information are controlled through the input gate, forget gate, and output gate, effectively solving the problems of gradient disappearance and gradient explosion in traditional neural networks and being able to better handle long - term dependence relationships in time - series data.

[0112] The combination of the forget gate, input gate, output gate, and memory unit allows the LSTM model network to selectively remember, forget, and output information when processing sequence data;

[0113] In the unit structure of the LSTM unit, the input gate determines the information that can be stored in the memory unit, and then the memory unit updates the memory unit at the current moment through a weighted sum. The forget gate determines the amount of information retained in the memory unit at the previous moment, and finally the output gate determines the hidden state \(h\) at the current moment; t ;

[0114] The structural formula expression of the LSTM unit is:

[0115] f t = σ(Wc · [h t-1 σ(W f · [h t-1 , xt] + b ρ )])

[0116] i t =σ(W c · [h t-1 , x t + b i )

[0117] C t =ft·C t -1 + i t · [tanh*W c · [h t-1 , x t + b c )]

[0118] O t =σ(W o · [h t-1 , x t + b c )

[0119] h t =O t · tanh(Ct)

[0120] Among them, f t is the output of the forgetting gate, i t is the output of the input gate, C t is the updated memory cell, O t is the output of the output gate, h t is the hidden state at the current time.

[0121] Furthermore, the process of real-time predicting the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model includes:

[0122] Input the real-time collected and preprocessed environmental data and historical power generation power data into the target power generation efficiency prediction model to predict the power generation power of the photovoltaic curtain wall system, and obtain the actually predicted power generation power;

[0123] The prediction step length can be set to the power generation power value for the next 24 hours according to actual needs, and then the power generation efficiency prediction value can be calculated based on the rated power of the photovoltaic curtain wall system and the actually predicted power generation power;

[0124] Among them, the formula expression of the power generation efficiency is:

[0125]

[0126] Among them, η is the power generation efficiency, P pred is the predicted power generation power, and P rated is the rated power of the photovoltaic curtain wall system.

[0127] To ensure the prediction accuracy and timeliness of the model, the sliding window technique is used to update the model regularly. As new data is continuously generated, the newly collected data is added to the training dataset every 5 days, and the model is retrained and optimized. During the model update process, the improved WOA algorithm is used to optimize the parameters of the LSTM neural network again to adapt to the changes in the performance of the photovoltaic curtain wall system and the seasonal and long-term changes in environmental conditions. By continuously updating the model, the prediction model can always maintain a high prediction accuracy.

[0128] In this embodiment, the east elevation of a building in Zhuhai is selected, and the power generation efficiency of the building integrated photovoltaic curtain wall system is predicted according to the above method steps. First, light intensity sensors, temperature sensors, humidity sensors, wind speed sensors, and component temperature sensors are installed around the photovoltaic curtain wall, and the power generation power data is obtained from the system inverter, and data collection and preprocessing operations are performed. Then, the improved WOA algorithm is used to optimize the parameters of the LSTM neural network. After multiple iterations, the optimal model parameters are determined. The real-time data is input into the optimized model to predict the power generation power for the next 24 hours, and the power generation efficiency is calculated. During the actual operation process, the sliding window technique is used to add new data to the training set every 5 days to re-optimize the model. By comparing with traditional prediction methods, the method in this embodiment has significantly improved prediction accuracy. There are fluctuations in the actual power output and the predicted power output in some samples, but at most sample points, the actual power output and the predicted power output are relatively close. The system peak prediction effect is not good, and the system prediction value is on the low side; the loss of the optimized model is approximately 0.11 in the 0th round, which is significantly higher than the loss of the initial model. However, as the number of training rounds increases, the loss of the optimized model drops rapidly, and after about the 20th round, its loss value is always lower than the loss of the initial model, dropping below 0.04. It can more accurately predict the power generation efficiency of the photovoltaic curtain wall system, providing strong support for energy management and system optimization.

[0129] This embodiment effectively improves the accuracy and reliability of the power generation efficiency prediction of the photovoltaic curtain wall system through steps such as data collection and preprocessing, optimization of the improved WOA algorithm, construction and optimization of the LSTM neural network model, and power generation efficiency prediction and model update. It has broad application prospects and practical value, and can provide important technical support for the development of the building integrated photovoltaic industry.

[0130] Embodiment 2

[0131] Based on the same inventive concept, this embodiment also provides a power generation efficiency prediction system for a photovoltaic curtain wall system, including:

[0132] A data acquisition and preprocessing module, which is used to collect the environmental data and power generation power data of the photovoltaic curtain wall system, and integrate the data through time synchronization technology to obtain an original data set; and perform data preprocessing operations on the original data set to obtain a target data set;

[0133] A model construction and optimization module, connected to the data acquisition and preprocessing module, is used to construct an initial power generation efficiency prediction model based on the LSTM neural network, and use the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model;

[0134] A power generation efficiency prediction module, connected to the model construction and optimization module, is used to perform real-time prediction on the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model to obtain a prediction result.

[0135] The power generation efficiency prediction system for a photovoltaic curtain wall system provided in this embodiment has all the advantages of the power generation efficiency prediction method for a photovoltaic curtain wall system provided in Embodiment 1.

[0136] Embodiment 3

[0137] This embodiment also discloses a computer device, including a memory, a processor, and a computer program stored on the memory, and the processor executes the computer program to implement the steps of the method described in Embodiment 1.

[0138] Embodiment 4

[0139] This embodiment also discloses a computer-readable storage medium, on which a computer program is stored, and when the computer program is executed by a processor, it implements the steps of the method described in Embodiment 1.

[0140] Embodiment 5

[0141] This embodiment also discloses a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the steps of the method described in Embodiment 1.

[0142] The above is only a preferred specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed in the present application should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method for predicting the power generation efficiency of a photovoltaic curtain wall system, characterized in that, Including: Collecting the environmental data and power generation data of the photovoltaic curtain wall system, and integrating the data through time synchronization technology to obtain an original data set; Performing data preprocessing operations on the original data set to obtain a target data set; Constructing an initial power generation efficiency prediction model based on the LSTM neural network, and using the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model; Based on the target power generation efficiency prediction model, performing real-time prediction on the power generation efficiency of the photovoltaic curtain wall system to obtain a prediction result.

2. The method according to claim 1, wherein: The process of collecting the environmental data and power generation data of the photovoltaic curtain wall system includes: Collecting environmental data through an environmental data monitoring station built on the photovoltaic curtain wall system; Building a photovoltaic curtain wall system by connecting photovoltaic curtain wall components in series and parallel, uniformly connecting the photovoltaic curtain wall components to an inverter group, obtaining real-time power generation data, and integrating the power generation data with environmental data through time synchronization technology to obtain an original data set; Wherein, the environmental data monitoring station includes a light intensity sensor, a component temperature sensor, an environmental temperature sensor, a humidity sensor, a wind speed sensor, and a pressure sensor; The light intensity sensor is used to collect and obtain direct solar irradiance, diffuse irradiance, and reflected irradiance data; The component temperature sensor is used to be arranged on the back of the photovoltaic curtain wall component to collect component temperature change data.

3. The method according to claim 1, wherein: The process of performing data preprocessing operations on the original data set includes: Adopting a method based on the 3σ principle to remove outliers in the original data set, and regarding data points outside the range of the mean ± 3 times the standard deviation as outliers and eliminating them; For the missing values in the original data set, using linear interpolation to fill them, and then performing normalization processing on all the data to map it to the interval of [0, 1]; Wherein, the formula expression of the normalization processing is: where x norm is the normalized data, x is the original data, x max and x min are the maximum and minimum values in the original data, respectively.

4. The method according to claim 1, wherein: The process of using the improved WOA algorithm to iteratively optimize the initial power generation efficiency prediction model through the target data set includes: Using the improved WOA algorithm to optimize the parameters of the LSTM neural network, encoding the weight matrix and bias vector of the LSTM neural network as the whale individual position vector in the WOA algorithm; wherein, the weight matrix of the LSTM neural network includes input weights, forgetting weights, output weights, and weights from the hidden layer to the output layer; Taking the prediction error as the fitness function of the WOA algorithm, through the iterative search of the improved WOA algorithm, using the mean square error as the evaluation index to obtain the parameter combination with the minimum fitness function, and determining the optimal LSTM neural network model structure and parameters. During the optimization process, whale individuals are sorted according to the fitness value, and the positions of the individuals are updated and the parameters are adjusted through the fitness value. After multiple iterations, the optimal model parameters are obtained, and then a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model is constructed.

5. The method according to claim 4, wherein The improved WOA algorithm modifies the original update method in the position update formula of the algorithm to the following by introducing an adaptive weight factor: X new = w·X rand - A·D; where w is the weight factor, X new is the updated position, X rand is the randomly selected individual position, A is the convergence factor, and D is the distance between the individual and the current optimal individual; the weight factor w is adaptively adjusted according to the number of iterations; the formula expression is: where w max and w min are the maximum and minimum values of the weight factor respectively, iter is the current iteration number, and iter max is the maximum iteration number.

6. The method according to claim 4, wherein the improved WOA algorithm further includes a method of initializing the population with chaos, and uses the Logistic chaotic mapping to generate the initial population; The process of using the Logistic chaotic mapping to generate the initial population by the method of initializing the population with chaos includes: Randomly generate an initial value z0 within the parameter range of the chaotic mapping, and then generate a series of chaotic values through the iterative formula; Map the chaotic values to the range of the search space as the initial population of the WOA algorithm; wherein, the iterative formula is: z n+1 = μ·z n ·(1 - z n ) where μ is the chaotic parameter.

7. The method according to claim 1, wherein the LSTM neural network structure includes an input layer, a hidden layer and an output layer; the input layer is used to receive the preprocessed environmental data and historical power generation data; the hidden layer includes a number of LSTM units, which are used to learn the time series features and long-term dependence relationships in the data; the output layer is used to output the predicted power generation power value of the photovoltaic curtain wall system.

8. The method according to claim 7, wherein the internal structure of the LSTM unit includes a forgetting gate, an input gate, an output gate and a memory unit; Through the combination of the forgetting gate, the input gate, the output gate and the memory unit, the LSTM model network is allowed to selectively remember, forget and output information when processing sequence data; In the cell structure of the LSTM cell, the input gate determines the information that can be stored in the memory cell, and then the memory cell updates the memory cell at the current moment with a weight. The forget gate determines the amount of information retained in the memory cell at the previous moment, and finally the output gate determines the hidden state h at the current moment t ; The structural formula expression of the LSTM unit is: f t = σ(W c · h-1 σ(W f · [h t-1 , x t + b ρ )]) i t = σ(W c · [h t-1 , x t + b i ) C t = ft·C t -1 + i t ·[tanh(W c ·[h t-1 , x t + b c )] O t = σ(W o · [h t-1 , x t + b c ) h t = O t ·tanh(Ct) where, f t is the output of the forget gate, i t is the output of the input gate, C t is the updated memory cell, O t is the output of the output gate, h t is the hidden state at the current time.

9. The method according to claim 1, wherein The process of real-time predicting the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model includes: Input the real-time collected and preprocessed environmental data and historical power generation data into the target power generation efficiency prediction model to predict the power generation power of the photovoltaic curtain wall system, and obtain the actually predicted power generation power; According to the rated power of the photovoltaic curtain wall system and the actually predicted power generation power, calculate the power generation efficiency prediction value; wherein, the formula expression of the power generation efficiency is: Among them, η is the power generation efficiency, P pred is the predicted power generation power, P rated is the rated power of the photovoltaic curtain wall system.

10. A photovoltaic curtain wall system power generation efficiency prediction system, characterized in that, including: A data collection and preprocessing module, which is used to collect the environmental data and power generation power data of the photovoltaic curtain wall system, and integrate the data through time synchronization technology to obtain the original data set; and perform data preprocessing operations on the original data set to obtain the target data set; The model construction and optimization module, connected to the data collection and preprocessing module, is used to construct an initial power generation efficiency prediction model based on the LSTM neural network. Through the target data set, the initial power generation efficiency prediction model is iteratively optimized using the improved WOA algorithm to obtain a target power generation efficiency prediction model based on the improved WOA-LSTM neural network model; The power generation efficiency prediction module, connected to the model construction and optimization module, is used to perform real-time prediction of the power generation efficiency of the photovoltaic curtain wall system based on the target power generation efficiency prediction model to obtain a prediction result.

Images