A method and system for optimizing the operating efficiency of a photovoltaic inverter

Abstract

The present invention provides a method and system for optimizing the operating efficiency of a photovoltaic inverter, which belongs to the technical field of photovoltaic inverters. The method includes: obtaining the output current and output voltage of a photovoltaic cell, the operating temperature and operating efficiency of a photovoltaic inverter, performing surface fitting operations respectively, obtaining a first reference domain and a second reference domain, and obtaining corresponding reference data based on the first reference domain and the second reference domain, and performing neural network training based on the reference data to obtain a neural network model, and optimizing the neural network model using a surface fitting method to obtain an optimal neural network model, and finally optimizing the operating efficiency of the photovoltaic inverter based on the optimal neural network model. The present invention uses surface fitting and neural networks to optimize the operating efficiency of a photovoltaic inverter from multiple dimensions. Through continuous optimization, the operating efficiency of the photovoltaic inverter can be kept in a relatively good or even optimal state for a long time, thereby improving the power generation and economic benefits of the entire system.

Description

Technical Field

[0001] The present invention relates to the technical field of photovoltaic inverters, and in particular to a method and system for optimizing the operating efficiency of a photovoltaic inverter. Background Art

[0002] As the core component of the solar power generation system, the operating efficiency of the photovoltaic inverter is directly related to the power generation and economic benefits of the entire system. In order to improve the operating efficiency of the photovoltaic inverter, it can be optimized from the following aspects:

[0003] 1. Maximum Power Point Tracking (MPPT)

[0004] Maximum power point tracking is an important means of optimizing the operation of photovoltaic inverters. By real-time detection of the operating voltage and current of photovoltaic panels, the MPPT algorithm can quickly find the maximum power point of photovoltaic cells, allowing the photovoltaic system to maintain the best working state under different environmental conditions.

[0005] 2. Maximum Efficiency Point Tracking (MEPT)

[0006] In addition to maximum power point tracking, maximum efficiency point tracking is also an effective optimization method. MEPT not only considers the output power of the photovoltaic panel, but also comprehensively considers the overall efficiency of the system, including the conversion efficiency of the inverter, heat dissipation loss, etc., so as to find the operating point with the highest overall system efficiency.

[0007] 3. Device loss analysis

[0008] Conduct loss analysis on key components in photovoltaic inverters, including power devices, filter components, etc., and reduce unnecessary energy losses and improve inverter efficiency by improving device selection and optimizing heat dissipation design.

[0009] 4. System Design Optimization

[0010] By optimizing the circuit design and electromagnetic design of the inverter, the energy loss inside the system can be reduced and the conversion efficiency can be improved. At the same time, the control strategy of the inverter is reasonably designed to adapt to different lighting conditions and load requirements.

[0011] 5. Component Matching

[0012] The matching of photovoltaic inverter and photovoltaic panel is also the key to improve efficiency. The appropriate inverter should be selected according to the characteristics of photovoltaic panels to ensure the matching between the two and reduce the loss of energy during the conversion process.

[0013] 6. Voltage Space Vector Modulation (SVPWM)

[0014] The use of voltage space vector modulation technology can improve the output voltage waveform quality of the inverter, reduce harmonic components, and thus improve the overall efficiency of the system.

[0015] 7. Use of high-efficiency inverter

[0016] Choosing high-efficiency photovoltaic inverter products is a direct means to improve system efficiency. High-efficiency inverters usually have higher conversion efficiency, lower self-loss and better heat dissipation performance.

[0017] 8. Application of Intelligent Control System

[0018] By introducing an intelligent control system, real-time monitoring and intelligent control of photovoltaic inverters can be achieved. The intelligent control system can automatically adjust the working state of the inverter according to factors such as lighting conditions and load demand, so that it is always in the optimal operating range, thereby improving system efficiency.

[0019] However, the above methods are basically considered and optimized from a single dimension. Therefore, the present invention attempts to provide a multi-dimensional optimization solution. Summary of the invention

[0020] The present invention provides a method and system for optimizing the operating efficiency of a photovoltaic inverter, which optimizes the operating efficiency of the photovoltaic inverter from multiple dimensions by using surface fitting and neural network methods.

[0021] The embodiment of this specification discloses a method for optimizing the operating efficiency of a photovoltaic inverter, including:

[0022] Obtain the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter;

[0023] The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain;

[0024] The output voltage is used as a row vector, and the operating temperature is used as a column vector to form a second matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a second reference domain;

[0025] Based on the first reference domain and the second reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature are obtained;

[0026] Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone;

[0027] Preprocessing the first training set and the first test set;

[0028] Based on the preprocessed first training set, model training is performed to obtain a first neural network model;

[0029] Based on the preprocessed first test set, obtaining a first accuracy, a first precision, and a first recall of the first neural network model;

[0030] The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain;

[0031] The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operation efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain;

[0032] Based on the third reference domain and the fourth reference domain, with the goal of optimizing the reference operation efficiency, using the first test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0033] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal first neural network model;

[0034] The operating efficiency of photovoltaic inverter is optimized based on the optimal first neural network model.

[0035] In some embodiments of this specification, a ratio of an actual output power of a photovoltaic inverter to its rated output power, that is, a load rate, is obtained;

[0036] The load rate is used as a row vector, the operating temperature is used as a column vector, a fifth matrix is ​​formed, and then the operating efficiency is introduced, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fifth reference domain;

[0037] Based on the first reference domain, the second reference domain and the fifth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference load rate are obtained;

[0038] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference load rate into a second training set and a second test set;

[0039] Preprocessing the second training set and the second test set;

[0040] Based on the preprocessed second training set, model training is performed to obtain a second neural network model;

[0041] Based on the preprocessed second test set, obtaining a second accuracy, a second precision, and a second recall of the second neural network model;

[0042] The second recall rate is used as a row vector, and the second accuracy rate is used as a column vector to form a sixth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a sixth reference domain;

[0043] The second precision is used as a row vector, and the second accuracy rate is used as a column vector to form a seventh matrix, and then the reference operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a seventh reference domain;

[0044] Based on the sixth reference domain and the seventh reference domain, with the goal of optimizing the reference operation efficiency, using the second test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0045] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal second neural network model;

[0046] The operating efficiency of photovoltaic inverter is optimized based on the optimal second neural network model.

[0047] In some embodiments of this specification, a harmonic current of a photovoltaic power generation system is obtained;

[0048] The harmonic current is used as a row vector, and the operating temperature is used as a column vector to form an eighth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as an eighth reference domain;

[0049] Based on the first reference domain, the second reference domain and the eighth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic current are obtained;

[0050] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic current into a third training set and a third test set;

[0051] Preprocessing the third training set and the third test set;

[0052] Based on the preprocessed third training set, model training is performed to obtain a third neural network model;

[0053] Based on the preprocessed third test set, obtaining a third accuracy, a third precision, and a third recall rate of the third neural network model;

[0054] The third recall rate is used as a row vector, and the third accuracy rate is used as a column vector to form a ninth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a ninth reference domain;

[0055] The third precision is used as a row vector, and the third accuracy rate is used as a column vector to form a tenth matrix, and then the reference operation efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a tenth reference domain;

[0056] Based on the ninth reference domain and the tenth reference domain, with the goal of optimizing the reference operation efficiency, using the third test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0057] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal third neural network model;

[0058] The photovoltaic inverter operating efficiency is optimized based on the optimal third neural network model.

[0059] In some embodiments of this specification, obtaining a harmonic voltage of a photovoltaic power generation system;

[0060] The harmonic voltage is used as a row vector, and the operating temperature is used as a column vector to form an eleventh matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic voltage surface fitting equation, and the three-dimensional surface is recorded as an eleventh reference domain;

[0061] Based on the first reference domain, the second reference domain and the eleventh reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic voltage are obtained;

[0062] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic voltage into a fourth training set and a fourth test set;

[0063] Preprocessing the fourth training set and the fourth test set;

[0064] Based on the preprocessed fourth training set, model training is performed to obtain a fourth neural network model;

[0065] Based on the preprocessed fourth test set, obtaining a fourth accuracy, a fourth precision, and a fourth recall rate of the fourth neural network model;

[0066] The fourth recall rate is used as a row vector, and the fourth accuracy rate is used as a column vector to form a twelfth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twelfth reference domain;

[0067] The fourth precision is used as a row vector, the fourth accuracy rate is used as a column vector, a thirteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a thirteenth reference domain;

[0068] Based on the twelfth reference domain and the thirteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fourth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0069] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fourth neural network model;

[0070] The photovoltaic inverter operating efficiency is optimized based on the optimal fourth neural network model.

[0071] In some embodiments of this specification, the harmonic impedance of the photovoltaic power generation system is obtained;

[0072] The harmonic impedance is used as a row vector, and the operating temperature is used as a column vector to form a fourteenth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic impedance surface fitting equation, and the three-dimensional surface is recorded as a fourteenth reference domain;

[0073] Based on the first reference domain, the second reference domain and the fourteenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic impedance are obtained;

[0074] Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference harmonic impedance into a fifth training set and a fifth test set;

[0075] Preprocessing the fifth training set and the fifth test set;

[0076] Based on the preprocessed fifth training set, model training is performed to obtain a fifth neural network model;

[0077] Based on the preprocessed fifth test set, obtaining a fifth accuracy, a fifth precision, and a fifth recall rate of the fifth neural network model;

[0078] The fifth recall rate is used as a row vector, and the fifth accuracy rate is used as a column vector to form a fifteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a fifteenth reference domain;

[0079] The fifth precision is used as a row vector, the fifth accuracy rate is used as a column vector, a sixteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a sixteenth reference domain;

[0080] Based on the fifteenth reference domain and the sixteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fifth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0081] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fifth neural network model;

[0082] The photovoltaic inverter operating efficiency is optimized based on the optimal fifth neural network model.

[0083] In some embodiments of this specification, obtaining a tilt angle of a photovoltaic cell;

[0084] The tilt angle is used as a row vector, and the operating temperature is used as a column vector to form a seventeenth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a tilt angle surface fitting equation, and the three-dimensional surface is recorded as the seventeenth reference domain;

[0085] Based on the first reference domain, the second reference domain and the seventeenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference tilt angle are obtained;

[0086] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference tilt angle into a sixth training set and a sixth test set;

[0087] Preprocessing the sixth training set and the sixth test set;

[0088] Based on the preprocessed sixth training set, model training is performed to obtain a sixth neural network model;

[0089] Based on the preprocessed sixth test set, obtaining a sixth accuracy, a sixth precision, and a sixth recall rate of the sixth neural network model;

[0090] The sixth recall rate is used as a row vector, and the sixth accuracy rate is used as a column vector to form an eighteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as the eighteenth reference domain;

[0091] The sixth precision is used as a row vector, the sixth accuracy rate is used as a column vector, a nineteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a nineteenth reference domain;

[0092] Based on the eighteenth reference domain and the nineteenth reference domain, with the goal of optimizing the reference operation efficiency, using the sixth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0093] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal sixth neural network model;

[0094] The photovoltaic inverter operating efficiency is optimized based on the optimal sixth neural network model.

[0095] In some embodiments of this specification, the ambient noise decibel of the photovoltaic inverter is obtained;

[0096] The ambient noise decibel is used as a row vector and the operating temperature is used as a column vector to form a twentieth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and an ambient noise decibel surface fitting equation, and the three-dimensional surface is recorded as a twentieth reference domain;

[0097] Based on the first reference domain, the second reference domain and the twentieth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference ambient noise decibel are obtained;

[0098] Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference ambient noise decibel into a seventh training set and a seventh test set;

[0099] Preprocessing the seventh training set and the seventh test set;

[0100] Based on the preprocessed seventh training set, model training is performed to obtain a seventh neural network model;

[0101] Based on the preprocessed seventh test set, obtaining a seventh accuracy, a seventh precision, and a seventh recall rate of the seventh neural network model;

[0102] The seventh recall rate is used as a row vector, and the seventh accuracy rate is used as a column vector to form a twenty-first matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twenty-first reference domain;

[0103] The seventh precision is used as a row vector, the seventh accuracy rate is used as a column vector, a twenty-second matrix is ​​formed, and then the reference operating efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a twenty-second reference domain;

[0104] Based on the twenty-first reference domain and the twenty-second reference domain, with the goal of optimizing the reference operation efficiency, using the seventh test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0105] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal seventh neural network model;

[0106] The photovoltaic inverter operating efficiency is optimized based on the optimal seventh neural network model.

[0107] The embodiments of this specification also disclose a photovoltaic inverter operation efficiency optimization system, which is used to implement any of the photovoltaic inverter operation efficiency optimization methods described above. The photovoltaic inverter operation efficiency optimization system includes:

[0108] Get module for Get Obtain the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter;

[0109] The first fitting module is used to:

[0110] The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain;

[0111] The output voltage is used as a row vector, the operating temperature is used as a column vector, a second matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a second reference domain;

[0112] A reference module, for obtaining a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature based on a first reference domain and a second reference domain with the goal of optimizing the operating efficiency;

[0113] Neural network modules for:

[0114] Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone;

[0115] Preprocessing the first training set and the first test set;

[0116] Based on the preprocessed first training set, model training is performed to obtain a first neural network model;

[0117] Based on the preprocessed first test set, obtaining a first accuracy, a first precision, and a first recall of the first neural network model;

[0118] The second fitting module is used to:

[0119] The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain;

[0120] The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operating efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain;

[0121] A first optimization module is used to select the optimal accuracy, optimal precision and optimal recall rate of the model corresponding to the optimal reference operation efficiency based on the third reference domain and the fourth reference domain and taking the optimal reference operation efficiency as the goal, using the first test set;

[0122] A second optimization module is used to optimize the model parameters by adjusting the hyperparameters, modifying the model structure and adopting regularization based on the optimal accuracy, the optimal precision and the optimal recall rate, and select the model parameters corresponding to the optimal reference running efficiency based on the validation set to obtain the optimal first neural network model;

[0123] The efficiency optimization module is used to optimize the operating efficiency of the photovoltaic inverter based on the optimal first neural network model.

[0124] The embodiments of this specification can at least achieve the following beneficial effects:

[0125] By obtaining the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter, the surface fitting operation is performed respectively to obtain the first reference domain and the second reference domain, and based on the first reference domain and the second reference domain, the corresponding reference data is obtained, and the neural network training is performed based on the reference data to obtain the neural network model, and the surface fitting method is used to optimize the neural network model to obtain the optimal neural network model, and finally the photovoltaic inverter operating efficiency is optimized based on the optimal neural network model. The surface fitting and neural network are used to optimize the operating efficiency of the photovoltaic inverter from multiple dimensions. Through continuous optimization, the operating efficiency of the photovoltaic inverter can be kept in a relatively good or even optimal state for a long time, thereby improving the power generation and economic benefits of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0126] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0127] Figure 1 It is a schematic diagram of the method for optimizing the operating efficiency of a photovoltaic inverter involved in the present invention. DETAILED DESCRIPTION

[0128] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the embodiments of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0129] The disclosure below provides many different embodiments or examples to implement different structures of the embodiments of the present invention. In order to simplify the disclosure of the embodiments of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the embodiments of the present invention. In addition, the embodiments of the present invention can repeat reference numbers and / or reference letters in different examples, and this repetition is for the purpose of simplification and clarity, which itself does not indicate the relationship between the various embodiments and / or settings discussed.

[0130] The embodiments of the present invention are described in detail below with reference to the accompanying drawings.

[0131] like Figure 1As shown, the embodiment of this specification discloses a method for optimizing the operating efficiency of a photovoltaic inverter, including:

[0132] Obtain the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter;

[0133] The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain;

[0134] The output voltage is used as a row vector, and the operating temperature is used as a column vector to form a second matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a second reference domain;

[0135] Based on the first reference domain and the second reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature are obtained;

[0136] Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone;

[0137] Preprocessing the first training set and the first test set;

[0138] Based on the preprocessed first training set, model training is performed to obtain a first neural network model;

[0139] Based on the preprocessed first test set, obtaining a first accuracy, a first precision, and a first recall of the first neural network model;

[0140] The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain;

[0141] The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operation efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain;

[0142] Based on the third reference domain and the fourth reference domain, with the goal of optimizing the reference operation efficiency, using the first test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0143] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal first neural network model;

[0144] The operating efficiency of photovoltaic inverter is optimized based on the optimal first neural network model.

[0145] In some embodiments of this specification, a ratio of an actual output power of a photovoltaic inverter to its rated output power, that is, a load rate, is obtained;

[0146] The load rate is used as a row vector, the operating temperature is used as a column vector, a fifth matrix is ​​formed, and then the operating efficiency is introduced, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fifth reference domain;

[0147] Based on the first reference domain, the second reference domain and the fifth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference load rate are obtained;

[0148] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference load rate into a second training set and a second test set;

[0149] Preprocessing the second training set and the second test set;

[0150] Based on the preprocessed second training set, model training is performed to obtain a second neural network model;

[0151] Based on the preprocessed second test set, obtaining a second accuracy, a second precision, and a second recall of the second neural network model;

[0152] The second recall rate is used as a row vector, and the second accuracy rate is used as a column vector to form a sixth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a sixth reference domain;

[0153] The second precision is used as a row vector, and the second accuracy rate is used as a column vector to form a seventh matrix, and then the reference operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a seventh reference domain;

[0154] Based on the sixth reference domain and the seventh reference domain, with the goal of optimizing the reference operation efficiency, using the second test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0155] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal second neural network model;

[0156] The operating efficiency of photovoltaic inverter is optimized based on the optimal second neural network model.

[0157] In some embodiments of this specification, a harmonic current of a photovoltaic power generation system is obtained;

[0158] The harmonic current is used as a row vector, and the operating temperature is used as a column vector to form an eighth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as an eighth reference domain;

[0159] Based on the first reference domain, the second reference domain and the eighth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic current are obtained;

[0160] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic current into a third training set and a third test set;

[0161] Preprocessing the third training set and the third test set;

[0162] Based on the preprocessed third training set, model training is performed to obtain a third neural network model;

[0163] Based on the preprocessed third test set, obtaining a third accuracy, a third precision, and a third recall rate of the third neural network model;

[0164] The third recall rate is used as a row vector, and the third accuracy rate is used as a column vector to form a ninth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a ninth reference domain;

[0165] The third precision is used as a row vector, and the third accuracy rate is used as a column vector to form a tenth matrix, and then the reference operation efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a tenth reference domain;

[0166] Based on the ninth reference domain and the tenth reference domain, with the goal of optimizing the reference operation efficiency, using the third test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0167] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal third neural network model;

[0168] The photovoltaic inverter operating efficiency is optimized based on the optimal third neural network model.

[0169] In some embodiments of this specification, obtaining a harmonic voltage of a photovoltaic power generation system;

[0170] The harmonic voltage is used as a row vector, and the operating temperature is used as a column vector to form an eleventh matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic voltage surface fitting equation, and the three-dimensional surface is recorded as an eleventh reference domain;

[0171] Based on the first reference domain, the second reference domain and the eleventh reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic voltage are obtained;

[0172] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic voltage into a fourth training set and a fourth test set;

[0173] Preprocessing the fourth training set and the fourth test set;

[0174] Based on the preprocessed fourth training set, model training is performed to obtain a fourth neural network model;

[0175] Based on the preprocessed fourth test set, obtaining a fourth accuracy, a fourth precision, and a fourth recall rate of the fourth neural network model;

[0176] The fourth recall rate is used as a row vector, and the fourth accuracy rate is used as a column vector to form a twelfth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twelfth reference domain;

[0177] The fourth precision is used as a row vector, the fourth accuracy rate is used as a column vector, a thirteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a thirteenth reference domain;

[0178] Based on the twelfth reference domain and the thirteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fourth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0179] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fourth neural network model;

[0180] The photovoltaic inverter operating efficiency is optimized based on the optimal fourth neural network model.

[0181] In some embodiments of this specification, the harmonic impedance of the photovoltaic power generation system is obtained;

[0182] The harmonic impedance is used as a row vector, and the operating temperature is used as a column vector to form a fourteenth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic impedance surface fitting equation, and the three-dimensional surface is recorded as a fourteenth reference domain;

[0183] Based on the first reference domain, the second reference domain and the fourteenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic impedance are obtained;

[0184] Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference harmonic impedance into a fifth training set and a fifth test set;

[0185] Preprocessing the fifth training set and the fifth test set;

[0186] Based on the preprocessed fifth training set, model training is performed to obtain a fifth neural network model;

[0187] Based on the preprocessed fifth test set, obtaining a fifth accuracy, a fifth precision, and a fifth recall rate of the fifth neural network model;

[0188] The fifth recall rate is used as a row vector, and the fifth accuracy rate is used as a column vector to form a fifteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a fifteenth reference domain;

[0189] The fifth precision is used as a row vector, the fifth accuracy rate is used as a column vector, a sixteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a sixteenth reference domain;

[0190] Based on the fifteenth reference domain and the sixteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fifth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0191] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fifth neural network model;

[0192] The photovoltaic inverter operating efficiency is optimized based on the optimal fifth neural network model.

[0193] In some embodiments of this specification, obtaining a tilt angle of a photovoltaic cell;

[0194] The tilt angle is used as a row vector, and the operating temperature is used as a column vector to form a seventeenth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a tilt angle surface fitting equation, and the three-dimensional surface is recorded as the seventeenth reference domain;

[0195] Based on the first reference domain, the second reference domain and the seventeenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference tilt angle are obtained;

[0196] Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference tilt angle into a sixth training set and a sixth test set;

[0197] Preprocessing the sixth training set and the sixth test set;

[0198] Based on the preprocessed sixth training set, model training is performed to obtain a sixth neural network model;

[0199] Based on the preprocessed sixth test set, obtaining a sixth accuracy, a sixth precision, and a sixth recall rate of the sixth neural network model;

[0200] The sixth recall rate is used as a row vector, and the sixth accuracy rate is used as a column vector to form an eighteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as the eighteenth reference domain;

[0201] The sixth precision is used as a row vector, the sixth accuracy rate is used as a column vector, a nineteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a nineteenth reference domain;

[0202] Based on the eighteenth reference domain and the nineteenth reference domain, with the goal of optimizing the reference operation efficiency, using the sixth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0203] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal sixth neural network model;

[0204] The photovoltaic inverter operating efficiency is optimized based on the optimal sixth neural network model.

[0205] In some embodiments of this specification, the ambient noise decibel of the photovoltaic inverter is obtained;

[0206] The ambient noise decibel is used as a row vector and the operating temperature is used as a column vector to form a twentieth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and an ambient noise decibel surface fitting equation, and the three-dimensional surface is recorded as a twentieth reference domain;

[0207] Based on the first reference domain, the second reference domain and the twentieth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference ambient noise decibel are obtained;

[0208] Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference ambient noise decibel into a seventh training set and a seventh test set;

[0209] Preprocessing the seventh training set and the seventh test set;

[0210] Based on the preprocessed seventh training set, model training is performed to obtain a seventh neural network model;

[0211] Based on the preprocessed seventh test set, obtaining a seventh accuracy, a seventh precision, and a seventh recall rate of the seventh neural network model;

[0212] The seventh recall rate is used as a row vector, and the seventh accuracy rate is used as a column vector to form a twenty-first matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twenty-first reference domain;

[0213] The seventh precision is used as a row vector, the seventh accuracy rate is used as a column vector, a twenty-second matrix is ​​formed, and then the reference operating efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a twenty-second reference domain;

[0214] Based on the twenty-first reference domain and the twenty-second reference domain, with the goal of optimizing the reference operation efficiency, using the seventh test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency;

[0215] Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal seventh neural network model;

[0216] The photovoltaic inverter operating efficiency is optimized based on the optimal seventh neural network model.

[0217] The embodiments of this specification also disclose a photovoltaic inverter operation efficiency optimization system, which is used to implement any of the photovoltaic inverter operation efficiency optimization methods described above. The photovoltaic inverter operation efficiency optimization system includes:

[0218] Get module for Get Obtain the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter;

[0219] The first fitting module is used to:

[0220] The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain;

[0221] The output voltage is used as a row vector, the operating temperature is used as a column vector, a second matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a second reference domain;

[0222] A reference module, for obtaining a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature based on a first reference domain and a second reference domain with the goal of optimizing the operating efficiency;

[0223] Neural network modules for:

[0224] Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone;

[0225] Preprocessing the first training set and the first test set;

[0226] Based on the preprocessed first training set, model training is performed to obtain a first neural network model;

[0227] Based on the preprocessed first test set, obtaining a first accuracy, a first precision, and a first recall of the first neural network model;

[0228] The second fitting module is used to:

[0229] The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain;

[0230] The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operation efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain;

[0231] A first optimization module is used to select the optimal accuracy, optimal precision and optimal recall rate of the model corresponding to the optimal reference operation efficiency based on the third reference domain and the fourth reference domain and taking the optimal reference operation efficiency as the goal, using the first test set;

[0232] A second optimization module is used to optimize the model parameters by adjusting the hyperparameters, modifying the model structure and adopting regularization based on the optimal accuracy, the optimal precision and the optimal recall rate, and select the model parameters corresponding to the optimal reference running efficiency based on the validation set to obtain the optimal first neural network model;

[0233] The efficiency optimization module is used to optimize the operating efficiency of the photovoltaic inverter based on the optimal first neural network model.

[0234] It is clear that the above-mentioned embodiments can be used in combination, such as using the optimal first neural network model to the optimal seventh neural network model (based on the mean of the output) to jointly optimize the operating efficiency of the photovoltaic inverter.

[0235] For example: first use operating efficiency, output current, output voltage, operating temperature, load rate, harmonic current, harmonic voltage, harmonic impedance, tilt angle and ambient noise decibels to perform surface fitting, and obtain corresponding reference domains respectively, and then take the optimal operating efficiency as the goal to obtain corresponding reference data, and perform neural network training based on these reference data to obtain corresponding neural network models, and perform surface fitting based on the accuracy, precision and recall of these neural network models to obtain corresponding reference domains respectively, and then take the optimal operating efficiency as the goal to obtain the optimal accuracy, optimal precision and optimal recall of the model corresponding to the optimal reference operating efficiency, based on the optimal accuracy, optimal precision and optimal recall, optimize the model parameters by adjusting hyperparameters, modifying the model structure and using regularization, and select the model parameters corresponding to the optimal reference operating efficiency based on the validation set to obtain the optimal neural network model, and finally optimize the operating efficiency of the photovoltaic inverter based on the optimal neural network model.

[0236] The present invention adopts a surface fitting method, which can well reflect the influence of output current, output voltage, operating temperature, load rate, harmonic current, harmonic voltage, harmonic impedance, tilt angle and environmental noise decibel on the operating efficiency respectively, as well as the influence of output current, output voltage, operating temperature, load rate, harmonic current, harmonic voltage, harmonic impedance, tilt angle and environmental noise decibel on the operating efficiency in different combinations.

[0237] It is understandable that each time optimization is performed, these data and their corresponding operating efficiency can be obtained and then used to update the reference domain. Through long-term updates, the optimization effect and accuracy can be improved.

[0238] It is understandable that corresponding functional modules may be added when implementing the methods of the above embodiments.

[0239] In practice, before surface fitting, the matrix can be preprocessed to remove NaN values ​​or 0 values, and then imported into the running efficiency to perform the surface fitting operation.

[0240] It should be understood that the above-mentioned surface fitting operation can use existing software tools, such as the Curve Fitting tool in Matlab. The above-mentioned modules can be data reading modules of existing data detection (collection) / measurement devices / sensors or photovoltaic inverter operation efficiency optimization systems, and can directly obtain corresponding data.

[0241] It is clear that by obtaining the corresponding data in the operation efficiency optimization system of the photovoltaic inverter in advance through existing equipment, a large amount of test data can be obtained. These test data can be simulation experiment data in the laboratory or actual measured or operation data of the operation efficiency optimization system of the photovoltaic inverter in practice.

[0242] The above-described embodiments are used to illustrate the present invention, not to limit the present invention, so changes in the exemplified values ​​or replacement of equivalent elements should still fall within the scope of the present invention.

[0243] From the above detailed description, it can be understood by those skilled in the art that the present invention can indeed achieve the aforementioned objectives and is in compliance with the provisions of the Patent Law.

[0244] Although the preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic creative concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. It should be pointed out that any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

[0245] It should be noted that the above description of the relevant process is only for example and explanation, and does not limit the scope of application of this specification. For those skilled in the art, various modifications and changes can be made to the process under the guidance of this specification. However, these modifications and changes are still within the scope of this specification.

[0246] The basic concepts have been described above. Obviously, for those of ordinary skill in the art who have read this application, the above invention disclosure is only for example and does not constitute a limitation of this application. Although not explicitly stated here, those of ordinary skill in the art may make various modifications, improvements and amendments to this application. Such modifications, improvements and amendments are suggested in this application, so such modifications, improvements and amendments still belong to the spirit and scope of the exemplary embodiments of this application.

[0247] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment" or "an embodiment" or "an alternative embodiment" mentioned twice or more in different positions in this specification does not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0248] In addition, it will be appreciated by those skilled in the art that various aspects of the present application may be illustrated and described by a number of patentable categories or situations, including any new and useful combination of processes, machines, products or substances, or any new and useful improvements thereto. Therefore, various aspects of the present application may be implemented entirely by hardware, entirely by software (including firmware, resident software, microcode, etc.), or by a combination of hardware and software. The above hardware or software may all be referred to as "units", "modules" or "systems". In addition, various aspects of the present application may take the form of a computer program product embodied in one or more computer-readable media, wherein computer-readable program code is contained therein.

[0249] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the implementation of the various components described above can be embodied in a hardware device, it can also be implemented as a pure software solution, for example, installation on an existing server or mobile device.

[0250] Similarly, it should be noted that in order to simplify the description disclosed in this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of the application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this method of the application should not be interpreted as reflecting the intention that the claimed object requires more features than those explicitly stated in each claim. On the contrary, the subject of the invention should have fewer features than the above single embodiment.

Claims

1. A method for optimizing the operating efficiency of a photovoltaic inverter, characterized in that: include: Obtain the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter; The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain; The output voltage is used as a row vector, and the operating temperature is used as a column vector to form a second matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a second reference domain; Based on the first reference domain and the second reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature are obtained; Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone; Preprocessing the first training set and the first test set; Based on the preprocessed first training set, model training is performed to obtain a first neural network model; Based on the preprocessed first test set, obtaining a first accuracy, a first precision, and a first recall of the first neural network model; The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain; The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operation efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain; Based on the third reference domain and the fourth reference domain, with the goal of optimizing the reference operation efficiency, using the first test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal first neural network model; The operating efficiency of photovoltaic inverter is optimized based on the optimal first neural network model.

2. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Obtain the ratio of the actual output power of the photovoltaic inverter to its rated output power, that is, the load rate; The load rate is used as a row vector, the operating temperature is used as a column vector, a fifth matrix is ​​formed, and then the operating efficiency is introduced, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fifth reference domain; Based on the first reference domain, the second reference domain and the fifth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference load rate are obtained; Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference load rate into a second training set and a second test set; Preprocessing the second training set and the second test set; Based on the preprocessed second training set, model training is performed to obtain a second neural network model; Based on the preprocessed second test set, obtaining a second accuracy, a second precision, and a second recall of the second neural network model; The second recall rate is used as a row vector, and the second accuracy rate is used as a column vector to form a sixth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a sixth reference domain; The second precision is used as a row vector, and the second accuracy rate is used as a column vector to form a seventh matrix, and then the reference operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a seventh reference domain; Based on the sixth reference domain and the seventh reference domain, with the goal of optimizing the reference operation efficiency, using the second test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal second neural network model; The operating efficiency of photovoltaic inverter is optimized based on the optimal second neural network model.

3. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Obtain harmonic current of photovoltaic power generation system; The harmonic current is used as a row vector, and the operating temperature is used as a column vector to form an eighth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as an eighth reference domain; Based on the first reference domain, the second reference domain and the eighth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic current are obtained; Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic current into a third training set and a third test set; Preprocessing the third training set and the third test set; Based on the preprocessed third training set, model training is performed to obtain a third neural network model; Based on the preprocessed third test set, obtaining a third accuracy, a third precision, and a third recall rate of the third neural network model; The third recall rate is used as a row vector, and the third accuracy rate is used as a column vector to form a ninth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a ninth reference domain; The third precision is used as a row vector, and the third accuracy rate is used as a column vector to form a tenth matrix, and then the reference operation efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a tenth reference domain; Based on the ninth reference domain and the tenth reference domain, with the goal of optimizing the reference operation efficiency, using the third test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal third neural network model; The photovoltaic inverter operating efficiency is optimized based on the optimal third neural network model.

4. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Obtain the harmonic voltage of the photovoltaic power generation system; The harmonic voltage is used as a row vector, and the operating temperature is used as a column vector to form an eleventh matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic voltage surface fitting equation, and the three-dimensional surface is recorded as an eleventh reference domain; Based on the first reference domain, the second reference domain and the eleventh reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic voltage are obtained; Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference harmonic voltage into a fourth training set and a fourth test set; Preprocessing the fourth training set and the fourth test set; Based on the preprocessed fourth training set, model training is performed to obtain a fourth neural network model; Based on the preprocessed fourth test set, obtaining a fourth accuracy, a fourth precision, and a fourth recall rate of the fourth neural network model; The fourth recall rate is used as a row vector, and the fourth accuracy rate is used as a column vector to form a twelfth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twelfth reference domain; The fourth precision is used as a row vector, the fourth accuracy rate is used as a column vector, a thirteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a thirteenth reference domain; Based on the twelfth reference domain and the thirteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fourth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fourth neural network model; The photovoltaic inverter operating efficiency is optimized based on the optimal fourth neural network model.

5. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Obtain the harmonic impedance of the photovoltaic power generation system; The harmonic impedance is used as a row vector, and the operating temperature is used as a column vector to form a fourteenth matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface and a harmonic impedance surface fitting equation, and the three-dimensional surface is recorded as a fourteenth reference domain; Based on the first reference domain, the second reference domain and the fourteenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference harmonic impedance are obtained; Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference harmonic impedance into a fifth training set and a fifth test set; Preprocessing the fifth training set and the fifth test set; Based on the preprocessed fifth training set, model training is performed to obtain a fifth neural network model; Based on the preprocessed fifth test set, obtaining a fifth accuracy, a fifth precision, and a fifth recall rate of the fifth neural network model; The fifth recall rate is used as a row vector, and the fifth accuracy rate is used as a column vector to form a fifteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a fifteenth reference domain; The fifth precision is used as a row vector, the fifth accuracy rate is used as a column vector, a sixteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a sixteenth reference domain; Based on the fifteenth reference domain and the sixteenth reference domain, with the goal of optimizing the reference operation efficiency, using the fifth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal fifth neural network model; The photovoltaic inverter operating efficiency is optimized based on the optimal fifth neural network model.

6. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Get the tilt angle of the photovoltaic cell; The tilt angle is used as a row vector, and the operating temperature is used as a column vector to form a seventeenth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and a tilt angle surface fitting equation, and the three-dimensional surface is recorded as the seventeenth reference domain; Based on the first reference domain, the second reference domain and the seventeenth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference tilt angle are obtained; Arrange the reference operating efficiency, the reference output current, the reference output voltage, the reference operating temperature and the reference tilt angle into a sixth training set and a sixth test set; Preprocessing the sixth training set and the sixth test set; Based on the preprocessed sixth training set, model training is performed to obtain a sixth neural network model; Based on the preprocessed sixth test set, obtaining a sixth accuracy, a sixth precision, and a sixth recall rate of the sixth neural network model; The sixth recall rate is used as a row vector, and the sixth accuracy rate is used as a column vector to form an eighteenth matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as the eighteenth reference domain; The sixth precision is used as a row vector, the sixth accuracy rate is used as a column vector, a nineteenth matrix is ​​formed, and then the reference operation efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a nineteenth reference domain; Based on the eighteenth reference domain and the nineteenth reference domain, with the goal of optimizing the reference operation efficiency, using the sixth test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal sixth neural network model; The photovoltaic inverter operating efficiency is optimized based on the optimal sixth neural network model.

7. The method for optimizing the operating efficiency of a photovoltaic inverter according to claim 1, characterized in that: Get the ambient noise decibel of the photovoltaic inverter; The ambient noise decibel is used as a row vector and the operating temperature is used as a column vector to form a twentieth matrix, and then the operating efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface and an ambient noise decibel surface fitting equation, and the three-dimensional surface is recorded as a twentieth reference domain; Based on the first reference domain, the second reference domain and the twentieth reference domain, with the goal of optimizing the operating efficiency, a reference operating efficiency, a reference output current, a reference output voltage, a reference operating temperature and a reference ambient noise decibel are obtained; Arrange the reference operating efficiency, reference output current, reference output voltage, reference operating temperature and reference ambient noise decibel into a seventh training set and a seventh test set; Preprocessing the seventh training set and the seventh test set; Based on the preprocessed seventh training set, model training is performed to obtain a seventh neural network model; Based on the preprocessed seventh test set, obtaining a seventh accuracy, a seventh precision, and a seventh recall rate of the seventh neural network model; The seventh recall rate is used as a row vector, and the seventh accuracy rate is used as a column vector to form a twenty-first matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a twenty-first reference domain; The seventh precision is used as a row vector, the seventh accuracy rate is used as a column vector, a twenty-second matrix is ​​formed, and then the reference operating efficiency is imported, a surface fitting operation is performed, and a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a twenty-second reference domain; Based on the twenty-first reference domain and the twenty-second reference domain, with the goal of optimizing the reference operation efficiency, using the seventh test set, selecting the optimal accuracy, optimal precision, and optimal recall rate of the model corresponding to the optimal reference operation efficiency; Based on the optimal accuracy, optimal precision and optimal recall rate, the model parameters are optimized by adjusting the hyperparameters, modifying the model structure and adopting regularization, and based on the validation set, the model parameters corresponding to the optimal reference running efficiency are selected to obtain the optimal seventh neural network model; The photovoltaic inverter operating efficiency is optimized based on the optimal seventh neural network model.

8. A photovoltaic inverter operation efficiency optimization system, characterized in that: A method for optimizing the operating efficiency of a photovoltaic inverter according to any one of claims 1 to 7, wherein the operating efficiency optimization system of the photovoltaic inverter comprises: An acquisition module, used to acquire the output current and output voltage of the photovoltaic cell, the operating temperature and operating efficiency of the photovoltaic inverter; The first fitting module is used to: The output current is used as a row vector, the operating temperature is used as a column vector, a first matrix is ​​formed, the operating efficiency is then introduced, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a first reference domain; The output voltage is used as a row vector, and the operating temperature is used as a column vector to form a second matrix, and then the operating efficiency is introduced, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a second reference domain; A reference module, for obtaining a reference operating efficiency, a reference output current, a reference output voltage and a reference operating temperature based on a first reference domain and a second reference domain with the goal of optimizing the operating efficiency; Neural network modules for: Arrange the reference operating efficiency, the reference output current, the reference output voltage and the reference operating temperature into a first training set and a first test set, and use the reference operating efficiency as a validation set alone; Preprocessing the first training set and the first test set; Based on the preprocessed first training set, model training is performed to obtain a first neural network model; based on the preprocessed first test set, a first accuracy, a first precision and a first recall rate of the first neural network model are obtained; The second fitting module is used to: The first recall rate is used as a row vector, and the first accuracy rate is used as a column vector to form a third matrix, and then the reference running efficiency is imported, and a surface fitting operation is performed to obtain a three-dimensional surface, and the three-dimensional surface is recorded as a third reference domain; The first precision is used as a row vector, the first accuracy rate is used as a column vector, a fourth matrix is ​​formed, the reference operation efficiency is then imported, a surface fitting operation is performed, a three-dimensional surface is obtained, and the three-dimensional surface is recorded as a fourth reference domain; A first optimization module is used to select the optimal accuracy, optimal precision and optimal recall rate of the model corresponding to the optimal reference operation efficiency based on the third reference domain and the fourth reference domain and taking the optimal reference operation efficiency as the goal, using the first test set; A second optimization module is used to optimize the model parameters by adjusting the hyperparameters, modifying the model structure and adopting regularization based on the optimal accuracy, the optimal precision and the optimal recall rate, and select the model parameters corresponding to the optimal reference running efficiency based on the validation set to obtain the optimal first neural network model; The efficiency optimization module is used to optimize the operating efficiency of the photovoltaic inverter based on the optimal first neural network model.

Images