Photovoltaic array equivalent simulation method and system based on data driving and storage medium
By performing IV characteristic analysis and Lamé curve approximation on the photovoltaic array, and training a hyperelliptic exponential prediction model using a BiLSTM model, the reliability and accuracy issues of the photovoltaic array control method were resolved, and efficient equivalent simulation was achieved.
Patent Information
- Application Number
- CN202511076539.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-08-01
- Publication Date
- 2025-11-14
AI Technical Summary
Existing photovoltaic array reference generators based on lookup tables or equation-based control methods suffer from insufficient reliability and accuracy, and equation-based photovoltaic array models are difficult to use directly for dynamic stability analysis.
A data-driven approach is adopted to construct an IV characteristic model by analyzing the IV characteristics of the photovoltaic array, and approximate it using the Lamé curve. The model is then combined with a BiLSTM model to train a hyperelliptic exponential prediction model, thereby achieving equivalent simulation of the photovoltaic array.
It improves the reliability and accuracy of equivalent simulation of photovoltaic arrays, simplifies iterative calculations, increases computational efficiency, and enhances robustness to load fluctuations.
Smart Images
Figure CN120951572A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of electrical automation, and specifically relates to a data-driven photovoltaic array equivalent simulation method, system, and storage medium. Background Technology
[0003] A photovoltaic (PV) array consists of a converter, a reference generator, and a controller. The reference generator embeds a PV equation or lookup table to ensure that the PV array's output corresponds to the IV characteristic curve. During application, the reference generator uses the embedded lookup table or approximate IV characteristic equation to generate a reference signal, thereby controlling the PV array.
[0004] Currently, photovoltaic (PV) array reference generators primarily rely on predefined values stored in lookup tables (corresponding to IV characteristic curves) to control the PV array. However, the reliability and accuracy of these predefined IV characteristic curves are not high. Equation-based PV arrays offer greater flexibility for precise control than lookup tables; however, they require iterative solutions to implicit IV relationships. Since traditional PV array modeling depends on implicit nonlinear equations (such as single / dual diode models), solving for IV curves requires complex iterations, making it difficult to directly apply to the dynamic stability analysis of PV systems. Summary of the Invention
[0005] One of the objectives of this invention is to provide a data-driven equivalent simulation method for photovoltaic arrays that is highly reliable, accurate, and efficient.
[0006] The second objective of this invention is to provide a system for implementing the data-driven photovoltaic array equivalent simulation method.
[0007] A third objective of this invention is to provide a storage medium on which a computer program is stored; when the computer program is executed by a processor, it implements the aforementioned data-driven photovoltaic array equivalent simulation method.
[0008] The data-driven photovoltaic array equivalent simulation method provided by this invention includes the following steps:
[0009] S1. Perform IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array;
[0010] S2. The numerical model of the target photovoltaic array constructed in step S1 is approximated based on the Lamé curve;
[0011] S3. Obtain historical data information of the target photovoltaic array;
[0012] S4. Using the data obtained in step S3, a superelliptic exponential prediction model for the target photovoltaic array is trained based on the BiLSTM model.
[0013] S5. Acquire data information of the target photovoltaic array in real time;
[0014] S6. Based on the superelliptic index prediction model of the target photovoltaic array obtained in step S4 and the data information obtained in step S5, the superelliptic index of the target photovoltaic array is predicted.
[0015] S7. Based on the hyperelliptic index of the target photovoltaic array obtained in step S6 and the data information obtained in step S2, complete the equivalent simulation of the photovoltaic array.
[0016] Step S1, which involves performing IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array, specifically includes the following steps:
[0017] The target photovoltaic array is equivalent to an equivalent circuit consisting of a current source, two resistors, and a PN junction diode.
[0018] The implicit equation of the target photovoltaic array is expressed as follows:
[0019]
[0020] In the formula i pv I is the output current of the target photovoltaic array. ph Let be the PV current of the target photovoltaic array, and G represents the solar radiation conditions for the photovoltaic module. n The set solar radiation conditions (value is 1000W / m) 2 ), I scn K represents the short-circuit current under standard test conditions. I For short-circuit current I sc The temperature coefficient, T is the temperature of the target photovoltaic array, T n Temperature under standard test conditions; I s It is the saturation current, and V ocn K is the open-circuit voltage of the photovoltaic array. V V is the temperature coefficient of the photovoltaic array, A is the set ideality factor, and V is the temperature coefficient of the photovoltaic array. t It is thermal voltage, and N is the number of cells in series, k is the Boltzmann constant, T is the absolute temperature of the target photovoltaic array, and q e The value of electron charge; v pv R is the output voltage of the target photovoltaic array; s R is the series resistance value. sh This represents the parallel resistance value.
[0021] Step S2 involves approximating the numerical model of the target photovoltaic array constructed in step S1 based on the Lamé curve, specifically including the following steps:
[0022] In Cartesian coordinates, the equation of the Lamé curve is expressed as:
[0023]
[0024] In the formula, a is the scaling factor for the horizontal axis; b is the scaling factor for the vertical axis; and n is the hyperellipse exponent.
[0025] Based on the equation of the Lamé curve, the mathematical expression of the characteristic equation of the target photovoltaic array is written as follows:
[0026]
[0027] In the formula, v is the sensed voltage, i is the sensed current, v≥0 and i≥0; V oc I is the open-circuit voltage. sc This is the short-circuit current;
[0028] The selected Lamé curve passes through the point (V) oc ,0)(0,I sc ) and (V mpp ,I mpp Therefore, the value of the voltage reference is calculated using the following formula:
[0029]
[0030] In the formula V mpp I is the voltage at the maximum power point; mpp This is the current at the maximum power point;
[0031] Let the first intermediate variable A and the second intermediate variable B be defined as follows:
[0032]
[0033] Through iterative calculation, the formula for calculating the hyperelliptic exponent n is obtained as follows:
[0034]
[0035] In the formula n i+1 The value of the hyperelliptic exponent n is obtained from the (i+1)th iteration; n i The value of the hyperelliptic exponent n is obtained from the i-th iteration;
[0036] For the Lamé curve, in the VS architecture (Visual Studio solution / project architecture), the current reference value i ref Represented as In a Client-Server (CS) architecture, the current voltage reference value v ref Represented as
[0037] Step S3, which involves obtaining historical data information of the target photovoltaic array, specifically includes the following steps:
[0038] Under different solar irradiance conditions, different temperature conditions, and conditions where the load change rate is greater than a set range value, historical data information of the target photovoltaic array is obtained;
[0039] The acquired data includes solar irradiance, temperature, load resistance, and the voltage, current, and power output of the target photovoltaic array.
[0040] Step S4 involves using the data obtained in step S3 to train a superelliptic exponential prediction model for the target photovoltaic array based on a BiLSTM model. This process includes the following steps:
[0041] The data obtained in step S3 is cleaned and normalized to construct a dataset;
[0042] Based on the BiLSTM model, a hyperelliptic exponential prediction model for the target photovoltaic array is constructed. The output of the constructed hyperelliptic exponential prediction model for the target photovoltaic array is expressed as follows:
[0043] h t =concat(h tf ,h tb )
[0044] In the formula h t This represents the hidden layer state of the model; `concat()` is the concatenation operation; h tf The hidden state of the forward LSTM; h tb This represents the hidden layer state of the backward LSTM;
[0045] The model employs a self-attention mechanism to emphasize important features. This mechanism uses a key-value-query pattern, assigning a value to the key based on the similarity between the key and the query. The self-attention mechanism ensures that the key only focuses on the top u most important queries. This is represented as...
[0046]
[0047] In the formula, A(Q,K,V) is the attention score; Softmax() is the softmax function; A sparse matrix containing only the first u queries; K is the keyword; d is the input dimension; V is the sentiment intensity feature for each part; k j Let K be the value of the j-th row;
[0048] The Informer uses the Distilling method to assign higher weights to the dominant features that contain the most information, and generates a focused self-attention feature map of the previous layer at layer j+1; represented as...
[0049]
[0050] In the formula Let `j+1` be the matrix of layer `j`; `MaxPooling()` is a max pooling operation with a stride of 2; `ELU()` is the activation function; `Conv1d()` is a one-dimensional convolution operation performed on the time series; att For attention operations;
[0051] After passing through the self-attention layer, the features are further processed and output through a fully connected layer to produce the final predicted hyperelliptic exponent; denoted as...
[0052] z = W o c s +b o
[0053]
[0054] In the formula, z is the input after linear transformation; W o b is the weight value; o For bias terms; c s The context vector generated by the self-attention layer; σ() is the softmax function; This is the predicted value of the hyperelliptic index;
[0055] During model training, the loss function is calculated using the following formula:
[0056]
[0057] In the formula, L is the loss function; m is the number of samples; Let n be the predicted value of the hyperelliptic index for the i-th sample; (i) The true value of the corresponding hyperelliptic exponent.
[0058] This invention also provides a system for implementing the data-driven photovoltaic array equivalent simulation method, comprising a model building module, an approximation processing module, a historical data acquisition module, a prediction model training module, a real-time data acquisition module, an exponential prediction module, and an equivalent simulation module; the model building module, approximation processing module, historical data acquisition module, prediction model training module, real-time data acquisition module, exponential prediction module, and equivalent simulation module are connected in series; the model building module is used to perform IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array, and upload the data information to the approximation processing module; the approximation processing module is used to perform approximation processing on the constructed numerical model of the target photovoltaic array based on the received data information, using the Lamé curve, and upload the data information to the historical data acquisition module; the historical data acquisition module is used to obtain... The system retrieves historical data information of the target photovoltaic array and uploads it to the prediction model training module. The prediction model training module, based on the received data and the acquired data, trains a hyperelliptic exponential prediction model for the target photovoltaic array using a BiLSTM model and uploads this data to the real-time data acquisition module. The real-time data acquisition module acquires real-time data information of the target photovoltaic array based on the received data and uploads this data to the exponential prediction module. The exponential prediction module, based on the received data, the acquired hyperelliptic exponential prediction model, and the data information, predicts the hyperelliptic exponent of the target photovoltaic array and uploads this data to the equivalent simulation module. The equivalent simulation module, based on the received data, the acquired hyperelliptic exponent, and the data information, performs an equivalent simulation of the photovoltaic array.
[0059] The present invention also provides a storage medium on which a computer program is stored; when the computer program is executed by a processor, it implements the data-driven photovoltaic array equivalent simulation method described above.
[0060] The data-driven photovoltaic array equivalent simulation method, system, and storage medium provided by this invention achieves modeling of the target photovoltaic array by performing IV characteristic analysis, IV characteristic modeling, and Lamé curve approximation. Then, a data-driven photovoltaic array equivalent simulation scheme is constructed using a BiLSTM model, which not only realizes the equivalent model of the photovoltaic array, but also has higher reliability, better accuracy, and higher efficiency. Attached Figure Description
[0061] Figure 1 This is a schematic diagram of the method flow of the present invention.
[0062] Figure 2 This is a schematic diagram of the simulation results of an embodiment of the method of the present invention.
[0063] Figure 3 This is a schematic diagram of the functional modules of the system of the present invention. Detailed Implementation
[0064] like Figure 1 The diagram shown is a flowchart of the method of the present invention: The data-driven photovoltaic array equivalent simulation method disclosed in this invention includes the following steps:
[0065] S1. Perform IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array; specifically including the following steps:
[0066] In the photovoltaic array description equation, the reference signal is generated from the sensed variables based on the PV characteristic equation;
[0067] The target photovoltaic array is equivalent to an equivalent circuit consisting of a current source, two resistors, and a PN junction diode. Because the mathematical equation describing the characteristic curve of the photovoltaic panel contains an exponential function, the equation becomes complex and implicit. Therefore, solving for the output voltage and current of the module often requires the use of complex and lengthy iterative expressions.
[0068] The implicit equation of the target photovoltaic array is expressed as follows:
[0069]
[0070] In the formula i pv I is the output current (A) of the target photovoltaic array. ph Let be the PV current (A) of the target photovoltaic array, and G represents the solar radiation conditions for the photovoltaic module. n The set solar radiation conditions (value is 1000W / m) 2 ), I scn K represents the short-circuit current under standard test conditions. I For short-circuit current I sc The temperature coefficient, T is the temperature of the target photovoltaic array, T n Temperature under standard test conditions; I s Let be the saturation current (A), and V ocn K is the open-circuit voltage of the photovoltaic array. V V is the temperature coefficient of the photovoltaic array, A is the set ideality factor, and V is the temperature coefficient of the photovoltaic array. t The thermal voltage (V) is given by the following value: N is the number of cells in series, k is the Boltzmann constant, T is the absolute temperature (Kelvin) of the target photovoltaic array, and q e The value of the electron charge (taken as 1.602 × 10⁻⁶). -19 C); v pvR is the output voltage (V) of the target photovoltaic array. s R is the series resistance value (Ω); sh The parallel resistance value (Ω);
[0071] S2. The numerical model of the target photovoltaic array constructed in step S1 is approximated based on the Lamé curve; specifically, the following steps are included:
[0072] Since the reference signal generated in the photovoltaic array depends on the sensed variable, to avoid tedious iterative calculations caused by the implicit nature of the signal, it is necessary to find a mathematical equation that can approximate the characteristic equation. This invention adopts the superelliptic approximation method. The superelliptic method provides a simple and fast approximation method for the IV characteristic equation. The shape of the superellipse (or Lamé curve) is similar to that of the IV curve, and at two special points (V... oc ,0) and (0,I sc At point (), the slopes of the tangent lines are approximately zero and infinity, respectively.
[0073] In Cartesian coordinates, the equation of the Lamé curve is expressed as:
[0074]
[0075] In the formula, a is the scaling factor for the horizontal axis; b is the scaling factor for the vertical axis; and n is the hyperellipse exponent.
[0076] Based on the equation of the Lamé curve, the mathematical expression of the characteristic equation of the target photovoltaic array is written as follows:
[0077]
[0078] In the formula, v is the sensed voltage, i is the sensed current, v≥0 and i≥0; V oc I is the open-circuit voltage. sc This is the short-circuit current;
[0079] The selected Lamé curve passes through the point (V) oc ,0)(0,I sc ) and (V mpp ,I mpp Therefore, the value of the voltage reference is calculated using the following formula:
[0080]
[0081] In the formula V mpp I is the voltage at the maximum power point; mpp This is the current at the maximum power point;
[0082] Let the first intermediate variable A and the second intermediate variable B be defined as follows:
[0083]
[0084] Through iterative calculation, the formula for calculating the hyperelliptic exponent n is obtained as follows:
[0085]
[0086] In the formula n i+1 The value of the hyperelliptic exponent n is obtained from the (i+1)th iteration; n i The value of the hyperelliptic exponent n is obtained from the i-th iteration;
[0087] For the Lamé curve, in the VS architecture (Visual Studio solution / project architecture), the current reference value i ref Represented as In a Client-Server (CS) architecture, the current voltage reference value v ref Represented as
[0088] S3. Obtain historical data information of the target photovoltaic array; specifically including the following steps:
[0089] Under different solar irradiance conditions, different temperature conditions, and conditions where the load change rate is greater than a set range value, historical data information of the target photovoltaic array is obtained;
[0090] The acquired data includes solar irradiance, temperature, and load resistance, as well as the voltage, current, and power output of the target photovoltaic array;
[0091] S4. Using the data obtained in step S3, a superelliptic exponential prediction model for the target photovoltaic array is trained based on the BiLSTM model; specifically, the following steps are included:
[0092] The data obtained in step S3 is cleaned and normalized to construct a dataset;
[0093] The model parameters also include short-circuit current, open-circuit voltage, maximum power point current, maximum power point voltage, temperature coefficient, operating temperature, and illuminance; the model output is the hyperelliptic exponential prediction value.
[0094] Based on the BiLSTM model, a hyperelliptic exponential prediction model for the target photovoltaic array is constructed. The output of the constructed hyperelliptic exponential prediction model for the target photovoltaic array is expressed as follows:
[0095] h t =concat(h tf ,h tb )
[0096] In the formula h tThis represents the hidden layer state of the model; `concat()` is the concatenation operation; h tf The hidden state of the forward LSTM; h tb This represents the hidden layer state of the backward LSTM;
[0097] The model employs a self-attention mechanism to emphasize important features. This mechanism uses a key-value-query pattern, assigning a value to the key based on the similarity between the key and the query. The self-attention mechanism ensures that the key only focuses on the top u most important queries. This is represented as...
[0098]
[0099] In the formula, A(Q,K,V) is the attention score; Softmax() is the softmax function; A sparse matrix containing only the first u queries; K is the keyword; d is the input dimension; V is the sentiment intensity feature for each part; k j Let K be the value of the j-th row;
[0100] The Informer uses the Distilling method to assign higher weights to the dominant features that contain the most information, and generates a focused self-attention feature map of the previous layer at layer j+1; represented as...
[0101]
[0102] In the formula Let `j+1` be the matrix of the (j+1)th layer; `MaxPooling()` is a max pooling operation with a stride of 2, which halves the computational cost of each layer, allowing the model to retain information from long input time series; `ELU()` is the activation function; `Conv1d()` performs a one-dimensional convolution operation on the time series; att For attention operations;
[0103] After passing through the self-attention layer, the features are further processed and output through a fully connected layer to produce the final predicted hyperelliptic exponent; denoted as...
[0104] z = W o c s +b o
[0105]
[0106] In the formula, z is the input after linear transformation; W o b is the weight value; o For bias terms; c s The context vector generated by the self-attention layer; σ() is the softmax function; This is the predicted value of the hyperelliptic index;
[0107] During model training, the loss function is calculated using the following formula:
[0108]
[0109] In the formula, L is the loss function; m is the number of samples; Let n be the predicted value of the hyperelliptic index for the i-th sample; (i) The true value of the corresponding hyperelliptic exponent;
[0110] By utilizing neural networks to establish a nonlinear mapping relationship between input parameters and the hyperelliptic exponent, this method can automatically learn complex nonlinear relationships and improve the fitting accuracy of the hyperelliptic curve to the photovoltaic cell IV curve.
[0111] S5. Acquire data information of the target photovoltaic array in real time;
[0112] S6. Based on the superelliptic index prediction model of the target photovoltaic array obtained in step S4 and the data information obtained in step S5, the superelliptic index of the target photovoltaic array is predicted. In specific implementation, the data information obtained in step S5 is cleaned and normalized, and then input into the superelliptic index prediction model of the target photovoltaic array to obtain the predicted superelliptic index of the target photovoltaic array.
[0113] S7. Based on the hyperelliptic index of the target photovoltaic array obtained in step S6 and the data information obtained in step S2, complete the equivalent simulation of the photovoltaic array; in specific implementation, based on the hyperelliptic index predicted in step S6, and following the modeling process in step S2, calculate the corresponding high-precision photovoltaic cell IV curve of the target photovoltaic array, thereby realizing the equivalent simulation of the photovoltaic array.
[0114] The method of the present invention will be further described below with reference to an embodiment:
[0115] Three different load resistors, 7Ω, 11Ω, and 20Ω, were selected to evaluate the system stability in the CSS (current source segment), MPP (maximum power segment), and VSS (voltage source segment), respectively. In each case, the inner loop gain T... v (s) Evaluation can be obtained, and simulation results are as follows: Figure 2 As shown.
[0116] pass Figure 2 As can be seen, the loop gain change is quite negligible for the three load resistors, which indicates that the method proposed in this invention is more robust and insensitive to load fluctuations.
[0117] like Figure 3The diagram shows the functional modules of the system of this invention: The system for implementing the data-driven photovoltaic array equivalent simulation method disclosed in this invention includes a model building module, an approximation processing module, a historical data acquisition module, a prediction model training module, a real-time data acquisition module, an exponential prediction module, and an equivalent simulation module; these modules are connected in series. The model building module performs IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array and uploads the data information to the approximation processing module. The approximation processing module performs approximation processing on the constructed numerical model of the target photovoltaic array based on the received data information, using the Lamé curve, and uploads the data information to the historical data acquisition module. The historical data acquisition module performs approximation processing on the received data information based on the received data information. The system acquires historical data of the target photovoltaic array and uploads it to the prediction model training module. The prediction model training module, based on the received and acquired data, trains a BiLSTM model to predict the hyperelliptic exponent of the target photovoltaic array and uploads this data to the real-time data acquisition module. The real-time data acquisition module acquires the target photovoltaic array's data in real-time and uploads this data to the exponent prediction module. The exponent prediction module, based on the received data, the acquired hyperelliptic exponent prediction model, and the data, predicts the hyperelliptic exponent of the target photovoltaic array and uploads this data to the equivalent simulation module. The equivalent simulation module, based on the received data, the acquired hyperelliptic exponent, and the data, performs an equivalent simulation of the photovoltaic array.
Claims
1. A data-driven method for equivalent simulation of photovoltaic arrays, comprising the following steps: S1. Perform IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array; S2. The numerical model of the target photovoltaic array constructed in step S1 is approximated based on the Lamé curve; S3. Obtain historical data information of the target photovoltaic array; S4. Using the data obtained in step S3, a superelliptic exponential prediction model for the target photovoltaic array is trained based on the BiLSTM model. S5. Acquire data information of the target photovoltaic array in real time; S6. Based on the superelliptic index prediction model of the target photovoltaic array obtained in step S4 and the data information obtained in step S5, the superelliptic index of the target photovoltaic array is predicted. S7. Based on the hyperelliptic index of the target photovoltaic array obtained in step S6 and the data information obtained in step S2, complete the equivalent simulation of the photovoltaic array.
2. The data-driven photovoltaic array equivalent simulation method according to claim 1, characterized in that... Step S1, which involves performing IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model of the target photovoltaic array, specifically includes the following steps: The target photovoltaic array is equivalent to an equivalent circuit consisting of a current source, two resistors, and a PN junction diode. The implicit equation of the target photovoltaic array is expressed as follows: In the formula i pv I is the output current of the target photovoltaic array. ph Let be the PV current of the target photovoltaic array, and G represents the solar radiation conditions for the photovoltaic module. n For the set solar conditions, I scn K represents the short-circuit current under standard test conditions. I For short-circuit current I sc The temperature coefficient, T is the temperature of the target photovoltaic array, T n Temperature under standard test conditions; I s It is the saturation current, and V ocn K is the open-circuit voltage of the photovoltaic array. V V is the temperature coefficient of the photovoltaic array, A is the set ideality factor, and V is the temperature coefficient of the photovoltaic array. t It is thermal voltage, and N is the number of cells in series, k is the Boltzmann constant, T is the absolute temperature of the target photovoltaic array, and q e The value of electron charge; v pv R is the output voltage of the target photovoltaic array; s R is the series resistance value. sh This represents the parallel resistance value.
3. The data-driven photovoltaic array equivalent simulation method according to claim 2, characterized in that... Step S2 involves approximating the numerical model of the target photovoltaic array constructed in step S1 based on the Lamé curve, specifically including the following steps: In Cartesian coordinates, the equation of the Lamé curve is expressed as: In the formula, 'a' is the scaling factor for the horizontal axis; b is the scaling factor for the vertical axis; n is the exponent of the hyperellipse; Based on the equation of the Lamé curve, the mathematical expression of the characteristic equation of the target photovoltaic array is written as follows: In the formula, v is the sensed voltage, i is the sensed current, v≥0 and i≥0; V oc I is the open-circuit voltage. sc This is the short-circuit current; The selected Lamé curve passes through the point (V) oc ,0)(0,I sc ) and (V mpp ,I mpp Therefore, the value of the voltage reference is calculated using the following formula: In the formula V mpp I is the voltage at the maximum power point; mpp This is the maximum power point current; Let the first intermediate variable A and the second intermediate variable B be defined as follows: Through iterative calculation, the formula for calculating the hyperelliptic exponent n is obtained as follows: In the formula n i+1 The value of the hyperelliptic exponent n is obtained from the (i+1)th iteration; n i The value of the hyperelliptic exponent n is obtained from the i-th iteration; For the Lamé curve, in the VS architecture, the current reference value i ref Represented as In the CS architecture, the current voltage reference value v ref Represented as 4. The data-driven photovoltaic array equivalent simulation method according to claim 3, characterized in that... Step S3, which involves obtaining historical data information of the target photovoltaic array, specifically includes the following steps: Under different solar irradiance conditions, different temperature conditions, and conditions where the load change rate is greater than a set range value, historical data information of the target photovoltaic array is obtained; The acquired data includes solar irradiance, temperature, load resistance, and the voltage, current, and power output of the target photovoltaic array.
5. The data-driven photovoltaic array equivalent simulation method according to claim 4, characterized in that... Step S4 involves using the data obtained in step S3 to train a superelliptic exponential prediction model for the target photovoltaic array based on a BiLSTM model. This process includes the following steps: The data obtained in step S3 is cleaned and normalized to construct a dataset; Based on the BiLSTM model, a hyperelliptic exponential prediction model for the target photovoltaic array is constructed. The output of the constructed hyperelliptic exponential prediction model for the target photovoltaic array is expressed as follows: h t =concat(h tf ,h tb ) In the formula h t This represents the hidden layer state of the model; `concat()` is the concatenation operation; h tf The hidden state of the forward LSTM; h tb This represents the hidden layer state of the backward LSTM; The model employs a self-attention mechanism to emphasize important features; Self-attention mechanisms employ a key-value-query model, assigning a value to a key by comparing its similarity to the query. This self-attention mechanism ensures that the key only focuses on the top u most important queries; represented as... In the formula, A(Q,K,V) is the attention score; Softmax() is the softmax function; A sparse matrix containing only the first u queries; K is the keyword; d is the input dimension; V is the sentiment intensity feature for each part; k j Let K be the value of the j-th row; The Informer uses the Distilling method to assign higher weights to the dominant features that contain the most information, and generates a focused self-attention feature map of the previous layer at layer j+1; represented as... In the formula Let `j+1` be the matrix of layer `j`; `MaxPooling()` is a max pooling operation with a stride of 2; `ELU()` is the activation function; `Conv1d()` is a one-dimensional convolution operation performed on the time series; att For attention operations; After passing through the self-attention layer, the features are further processed and output through a fully connected layer to produce the final predicted hyperelliptic exponent; denoted as... z=W o c s +b o In the formula, z is the input after linear transformation; W o b is the weight value; o For bias terms; c s The context vector generated by the self-attention layer; σ() is the softmax function; This is a predicted value for the hyperelliptic index; During model training, the loss function is calculated using the following formula: In the formula, L is the loss function; m is the number of samples; Let n be the predicted value of the hyperelliptic index for the i-th sample; (i) The true value of the corresponding hyperelliptic exponent.
6. A system for implementing the data-driven photovoltaic array equivalent simulation method according to any one of claims 1 to 5, characterized in that... The system includes a model building module, an approximation processing module, a historical data acquisition module, a prediction model training module, a real-time data acquisition module, an exponential prediction module, and an equivalent simulation module. These modules are connected in series. The model building module performs IV characteristic analysis on the target photovoltaic array to construct an IV characteristic model and uploads the data to the approximation processing module. The approximation processing module approximates the constructed numerical model of the target photovoltaic array based on the received data, using the Lamé curve, and uploads the data to the historical data acquisition module. The historical data acquisition module acquires historical data of the target photovoltaic array based on the received data and... The system comprises the following modules: a data information upload and prediction model training module; a prediction model training module, which uses the received data information and the acquired data information to train a BiLSTM model to predict the hyperelliptic exponent of the target photovoltaic array, and then uploads the data information to the real-time data acquisition module; a real-time data acquisition module, which uses the received data information to acquire the target photovoltaic array's data information in real time, and then uploads the data information to the exponent prediction module; an exponent prediction module, which uses the received data information, the acquired hyperelliptic exponent prediction model, and the data information to predict the hyperelliptic exponent of the target photovoltaic array, and then uploads the data information to the equivalent simulation module; and an equivalent simulation module, which uses the received data information, the acquired hyperelliptic exponent, and the data information to perform an equivalent simulation of the photovoltaic array.
7. A storage medium having a computer program stored thereon; said computer program, when executed by a processor, implements the data-driven photovoltaic array equivalent simulation method according to any one of claims 1 to 5.