Photovoltaic power generation grid-connected limited power control method and device and computer equipment

Abstract

The invention provides a photovoltaic power generation grid-connected limited power control method and device and computer equipment, and belongs to the technical field of new energy grid-connected power control. The method comprises the following steps: acquiring a power prediction value required by a load at a next moment and a maximum output power prediction value of a photovoltaic power station; judging whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated during photovoltaic power generation grid connection and the predicted value of the power required by the load at the next moment; if yes, the constructed MPC module is used for generating the optimal duty ratio, so that the power output by the DC / DC converter in photovoltaic power generation under the control of the pulse modulation signal with the optimal duty ratio tracks the target power value, and otherwise, MPPT maximum power point tracking control is adopted on the DC / DC converter side. By means of the method, flexible power control during photovoltaic power generation grid connection is achieved, and the method has the advantages of being high in response speed, high in real-time performance, high in adaptability to variable loads and the like.

Description

Technical Field

[0001] The present invention relates to the technical field of new energy grid-connected power control, and in particular to a photovoltaic power generation grid-connected power limiting control method, a photovoltaic power generation grid-connected power limiting control device, a computer device and a machine-readable storage medium. Background Art

[0002] Photovoltaic power generation has the characteristics of randomness, intermittency, periodicity and volatility. Therefore, large and medium-sized photovoltaic power stations should be equipped with photovoltaic power generation grid-connected power control systems. Photovoltaic power generation system grid-connected power control refers to controlling the power delivered by photovoltaic power stations to loads in a certain way to meet the power requirements of loads and track and adjust them. The purpose is to match the output power of photovoltaic power stations with the load requirements, achieve a stable supply of electricity, and maximize the use of photovoltaic power generation.

[0003] At present, the main power limit control methods for photovoltaic power generation grid connection are DC / DC converter PI control and inverter active power regulation control. DC / DC converter PI control ensures stable operation within a limited power range by precisely controlling the output voltage and current of the converter. Its control process is mainly as follows: real-time monitoring of the output voltage and current of the DC / DC converter, and calculation of the current output power, and comparison of the real-time monitored output power with the set power limit value. If the current output power exceeds the power limit value, power limit control is required. By optimizing the control parameters of the PID algorithm, the accuracy and response speed of power limit control can be further improved. Inverter active power regulation control is usually achieved by changing the phase and amplitude of the output current and voltage of the inverter, and by controlling the conduction time of the IGBT in the inverter circuit to change the output voltage, thereby controlling the amplitude of the output active power. It can be achieved by adjusting the control parameters of the inverter or using a specific control algorithm. Some advanced inverters are also equipped with a decoupling control function of active power and reactive power to more accurately adjust the output active power. The main problems of PI control of DC / DC converter are as follows: 1) It is mainly applicable to linear systems, and the processing of constraints is relatively weak; 2) It has poor adaptability to load changes. When the load changes, the control effect of the PI controller may be affected, resulting in a decrease in system performance; 3) Under the action of integral regulation, if the system error exists for a long time, the integral term may continue to grow, resulting in integral saturation, which may cause the system to over-respond or even cause system instability; 4) The dynamic response speed is slow, and the control variable needs to be gradually adjusted through the integral action. The main problems of inverter active power regulation control are as follows: 1) The power regulation of photovoltaic inverters usually involves the adjustment of multiple parameters such as voltage, frequency and switching state, and the operation is relatively complex and the response speed is slow; 2) The photovoltaic inverter needs to have good grid adaptability to ensure stable connection with the grid and power quality, which requires the photovoltaic inverter to have complex protection functions and control strategies, which increases the complexity and cost of the system; 3) When photovoltaic power generation is connected to the plant system of a thermal power plant, there are still many problems in the output power and regulation control of the inverter, which need to be improved and optimized.

[0004] In summary, there is an urgent need to improve the power limiting control scheme for photovoltaic power generation grid-connected. Summary of the invention

[0005] In view of this, the purpose of the embodiments of the present invention is to provide a photovoltaic power generation grid-connected power limiting control method, a photovoltaic power generation grid-connected power limiting control device, a computer device and a machine-readable storage medium, so as to overcome one or more defects in the photovoltaic power generation grid-connected power limiting control scheme based on DC / DC converter PI control and inverter active power regulation control in the prior art.

[0006] In order to achieve the above-mentioned object, a first aspect of an embodiment of the present invention provides a photovoltaic power generation grid-connected power limiting control method, the method comprising:

[0007] Obtain the predicted value of the load power required at the next moment and the predicted value of the maximum output power of the photovoltaic power station;

[0008] Determine whether the predicted maximum output power value of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted power required by the load at the next moment; if so, use the constructed MPC model to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value, otherwise, MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted maximum output power value of the photovoltaic power station at the next moment;

[0009] The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

[0010] In a specific embodiment of the present invention, before determining whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, the method further includes:

[0011] The first time series is formed by the characteristic variables collected at consecutive moments;

[0012] The first time series is input into the constructed transmission and transformation loss prediction model to predict the power loss value generated when the photovoltaic power generation is connected to the grid at the next moment;

[0013] Among them, the maximum collection time of the characteristic variable is the current time, the characteristic variable includes the power transmission and transformation state variable that changes the power loss of photovoltaic power generation grid-connected, and the power transmission and transformation loss prediction model is obtained by training and optimizing the parameters of the sequence prediction network of the initial state.

[0014] In a specific embodiment of the present invention, the characteristic variables also include the voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics of the transformer in the photovoltaic power generation grid-connected line. The voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics at continuous moments are obtained by extracting features from the second time series using the constructed feature extraction network. The second time series is composed of the transformer voltage amplitude and the transformer current amplitude collected at continuous moments.

[0015] In a specific embodiment of the present invention, the feature extraction network includes a 2D convolutional layer, a pooling layer and a Flatten layer connected in sequence, wherein:

[0016] A 2D convolution layer, used for receiving a two-dimensional matrix converted from the second time series and performing feature extraction on the two-dimensional matrix;

[0017] The pooling layer is used to perform pooling operations on the feature maps output after feature extraction;

[0018] The Flatten layer is used to convert the feature map output after the pooling operation into a one-dimensional vector.

[0019] In a specific embodiment of the present invention, training and parameter optimization of the sequence prediction network in the initial state include:

[0020] Use historical data of feature variables to construct training samples;

[0021] Use training samples to pre-train and pre-optimize hyperparameters of sequence prediction networks;

[0022] Use the characteristic variables collected in real time to construct real samples;

[0023] Use real samples to perform incremental training and hyperparameter optimization on the pre-trained and hyperparameter-optimized sequence prediction network;

[0024] Among them, the particle swarm optimization algorithm is used to optimize the hyperparameters.

[0025] In a specific embodiment of the present invention, the DC / DC converter uses a boost circuit as a DC / DC boost circuit, and the space state equation is expressed as:

[0026]

[0027] Wherein, x represents the state variable, A represents the state matrix, u represents the input vector, the input vector is the target voltage value output by the DC / DC converter determined according to the scheduling instruction, B represents the input matrix, represents the derivative of the state variable, y represents the output vector, the output vector is the difference between 1 and the predicted value of the duty cycle of the pulse modulation signal controlling the DC / DC converter, C represents the output matrix, and D represents the direct transfer matrix. In a specific embodiment of the present invention, the constraints of the MPC model include: the duty cycle is greater than or equal to 0 and less than or equal to 1; the target power value is within a preset fluctuation range.

[0028] In a specific embodiment of the present invention, when the constructed MPC model is used to generate the optimal duty cycle, an optimization algorithm is used to optimize the predicted time domain length of the MPC model.

[0029] In a specific embodiment of the present invention, an optimization algorithm is used to optimize the prediction time domain length of the MPC model, including:

[0030] Initialize particle swarm;

[0031] Calculate the fitness value of each particle according to the fitness function;

[0032] Execute particle swarm update step;

[0033] If the number of iterations is reached or the particle fitness value of the particle swarm reaches the expected value, the optimal individual in the current particle swarm is output as the optimal solution for the predicted time domain length, otherwise the particle swarm update step is executed again;

[0034] Wherein, the particle swarm updating step comprises:

[0035] Determine the current individual extremum of each particle and the current global optimal solution of the particle swarm;

[0036] Update the speed and position of each particle;

[0037] The fitness function is an evaluation function that characterizes the degree of error between the output voltage of the DC / DC converter calculated by the MPC model and the target voltage value required to be output by the DC / DC converter determined according to the scheduling instruction.

[0038] In a specific embodiment of the present invention, the power transmission and transformation state variables include ambient temperature, load and transformer usage time.

[0039] A second aspect of an embodiment of the present invention provides a photovoltaic power generation grid-connected power limiting control device, the device comprising:

[0040] An acquisition module is used to obtain the predicted value of the load power required at the next moment and the predicted value of the maximum output power of the photovoltaic power station;

[0041] The power limit control module is used to determine whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment; if so, the constructed MPC model is used to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value; otherwise, the MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted value of the maximum output power of the photovoltaic power station at the next moment;

[0042] The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

[0043] A third aspect of an embodiment of the present invention provides a computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the photovoltaic power generation grid-connected power limiting control method described in the first aspect of an embodiment of the present invention is implemented.

[0044] A fourth aspect of the embodiments of the present invention provides a machine-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the photovoltaic power generation grid-connected power limiting control method described in the first aspect of the embodiments of the present invention is implemented.

[0045] In the above technical solution, the real-time output power of photovoltaic power generation when connected to the grid is dynamically adjusted according to the prediction results of the load required power and the prediction results of the maximum output power of the photovoltaic power station, including two adjustment strategies. The first is to adopt MPPT maximum power point tracking control on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted value of the maximum output power of the photovoltaic power station at the next moment, and the second is to use the constructed MPC model to generate the optimal duty cycle so that the output power on the DC / DC converter side tracks the required power. Based on this, the flexible adjustment of the photovoltaic power generation grid-connected power is realized. Compared with the traditional power limit control, it ensures the maximum utilization of photovoltaic power generation energy, and at the same time, ensures the stable operation of the power grid and the reliability of power supply. In addition, the MPC control based on the DC / DC converter side abandons the DC / DC converter PI control and inverter active power regulation control in the traditional power limiting control, effectively overcoming the defects of the DC / DC converter PI control in the traditional power limiting control, such as weak constraint processing capability, poor adaptability to load changes and slow dynamic response speed, as well as the defects of the inverter active power regulation control in the traditional power limiting control, such as complex control process, slow response speed, complex control components and high cost.

[0046] Other features and advantages of the embodiments of the present invention will be described in detail in the subsequent detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The accompanying drawings are used to provide a further understanding of the embodiments of the present invention and constitute a part of the specification. Together with the following specific implementations, they are used to explain the embodiments of the present invention, but do not constitute a limitation on the embodiments of the present invention. In the accompanying drawings:

[0048] Figure 1 A first flow chart of a photovoltaic power generation grid-connected power limiting control method according to an embodiment of the present invention is schematically shown;

[0049] Figure 2 A second flow chart of the photovoltaic power generation grid-connected power limiting control method according to an embodiment of the present invention is schematically shown;

[0050] Figure 3 The first construction process diagram of the power transmission and transformation loss prediction model in a specific application example is schematically shown;

[0051] Figure 4 The second construction process diagram of the power transmission and transformation loss prediction model in a specific application example is schematically shown;

[0052] Figure 5 The structure diagram of the power supply circuit from the photovoltaic power station where the DC / DC converter is located to the load in a specific application example is schematically shown;

[0053] Figure 6 The schematic diagram of the structure of the system for connecting photovoltaic power generation to a 6kV thermal power plant in a specific application example is shown;

[0054] Figure 7 A schematic diagram shows a structural block diagram of a photovoltaic power generation grid-connected power limiting control device according to an embodiment of the present invention;

[0055] Figure 8 The following schematically shows a block diagram of a computer device according to an embodiment of the present invention. DETAILED DESCRIPTION

[0056] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings in the embodiments of the present invention. It should be understood that the specific implementation methods described herein are only used to illustrate and explain the embodiments of the present invention, and are not used to limit the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0057] It should be noted that if the embodiments of the present invention involve directional indications (such as up, down, left, right, front, back, etc.), the directional indications are only used to explain the relative position relationship, movement status, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication will also change accordingly.

[0058] In addition, if there are descriptions involving "first", "second", etc. in the embodiments of the present invention, the descriptions of "first", "second", etc. are only used for descriptive purposes and cannot be understood as indicating or suggesting their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In addition, the technical solutions between the various embodiments can be combined with each other, but they must be based on the ability of ordinary technicians in the field to implement them. When the combination of technical solutions is contradictory or cannot be implemented, it should be deemed that such a combination of technical solutions does not exist and is not within the scope of protection required by the present invention.

[0059] See also Figure 1 A photovoltaic power generation grid-connected power limiting control method provided by an embodiment of the present invention may include the following steps.

[0060] Step S100, obtaining the load power prediction value at the next moment and the photovoltaic power station maximum output power prediction value at the next moment. The load power prediction value at the next moment can be derived from the load power prediction data, and the photovoltaic power station maximum output power prediction value at the next moment can be derived from the photovoltaic power station maximum output power prediction data.

[0061] Step S102, determine whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment; if so, use the constructed MPC model to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value, otherwise the MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted value of the maximum output power of the photovoltaic power station at the next moment. The target power value refers to the power value required to be output by the DC / DC converter determined according to the scheduling instruction, that is, the required power value. The spatial state equation used in the MPC (Model Predictive Control) model is constructed based on the circuit of the photovoltaic power station where the DC / DC converter is located to supply power to the load.

[0062] It is known that the MPC model can adopt a dynamic matrix control (DMC) model, a model algorithm control (MAC) model, a generalized predictive control (GPC) model, and a known and publicly improved MPC model. The spatial state equation used by the MPC model needs to be set in combination with the specific application scenario, and this embodiment does not describe this part in detail.

[0063] As in the above embodiment, in order to control the power size of photovoltaic power generation grid connection at the next moment, firstly, the predicted value of the load power required at the next moment is extracted from the predicted data of the load power required, and the predicted value of the maximum output power of the photovoltaic power station at the next moment is extracted from the predicted data of the maximum output power of the photovoltaic power station, and then the photovoltaic power generation grid connection output power is dynamically adjusted in combination with the power loss value generated when the photovoltaic power generation grid connection is combined. At the same time, in order to ensure the response speed, control cost, adaptability to variable loads, etc. during dynamic adjustment, when the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation grid connection is combined and the predicted value of the load power required at the next moment, the pre-built MPC model is used to control the output power of the DC / DC converter side, thereby realizing the flexible adjustment of the photovoltaic power generation grid connection power, ensuring the stable operation of the power grid and the reliability of the power supply, and taking into account real-time, low cost, adaptability to variable loads, flexibility and scalability. Among them, flexibility is reflected in: the dynamic adjustment of the photovoltaic power generation grid connection output power includes two adjustment strategies. The first adjustment strategy is: when the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, the MPC prediction on the DC / DC converter side is used to track the required power value. The second adjustment strategy is: when the predicted value of the maximum output power of the photovoltaic power station at the next moment is less than or equal to the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, the MPPT maximum power point tracking control is adopted on the DC / DC converter side. Scalability is reflected in: the MPC model has strong scalability, and the control strategy of the MPC model can be adjusted according to the specific photovoltaic power generation grid-connected scenario, such as adjusting the spatial state equation, the control domain length, the prediction domain length, etc.

[0064] The power loss when photovoltaic power generation is connected to the grid is mainly caused by the transmission and transformation lines and devices. Therefore, the power loss value generated when photovoltaic power generation is connected to the grid is also called the transmission and transformation loss value generated when photovoltaic power generation is connected to the grid.

[0065] Combination Figure 2 As shown, in a specific embodiment of the present invention, the transmission and transformation loss value is obtained by dynamic prediction. Accordingly, before judging whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, the following steps may also be included:

[0066] Step S101, using the constructed power transmission and transformation prediction model, predict the power loss value generated when photovoltaic power generation is connected to the grid at the next moment.

[0067] Specifically, step S101 may include:

[0068] The first time series is formed by the characteristic variables collected at consecutive moments;

[0069] The first time series is input into the constructed transmission and transformation loss prediction model to predict the power loss value generated when the photovoltaic power generation is connected to the grid at the next moment;

[0070] Among them, the maximum collection time of the characteristic variables is the current time, the characteristic variables include the power transmission and transformation state variables that change the power transmission and transformation loss, and the power transmission and transformation loss prediction model is obtained through the sequence prediction network training and parameter optimization of the initial state.

[0071] Specifically, photovoltaic power generation grid-connected lines involve complex step-up transformers, step-down transformers, transmission lines, etc. The transmission line loss needs to consider the influence of line material, ambient temperature, and the heat generated when current passes through the conductor on the resistance, etc. The step-up transformer and step-down transformer and other transformer losses need to consider copper loss, iron loss, etc. It can be seen that there are many factors that affect the power loss of transmission and transformation, covering a wide range of dimensions. The power transmission and transformation state variables that change the power transmission and transformation losses described in the above embodiments should be understood as the principal components of these influencing factors extracted by dimensionality reduction analysis or combined with expert experience. Dimensionality reduction analysis can adopt methods such as PCA principal component analysis.

[0072] In a comparative embodiment, the power transmission and transformation loss value is a preset value, and the process of determining the preset value may include the following implementation steps:

[0073] The specific photovoltaic power generation grid-connected line is simplified by using, for example, the equivalent impedance method, and a simplified circuit model of the photovoltaic power generation grid-connected line is constructed after simplification. In the simplified circuit model, complex step-up transformers, step-down transformers, transmission lines, etc. are simplified and calculated into equivalent loss impedances using equivalent parameters;

[0074] Calculate the transmission and transformation loss value based on the simplified circuit model.

[0075] It can be seen that the equivalent impedance method regards the line resistance as a constant value when simplifying the circuit, but in fact the wire resistance will change with factors such as temperature, because in most application scenarios, the ambient temperature is difficult to keep constant, which leads to errors in the calculation results. Secondly, the equivalent impedance method mainly calculates the resistive loss in the circuit, ignoring the influence of the inductive loss on the total line loss. In addition, the transformer power loss is affected by multiple factors such as voltage level, manufacturing process and material, load, ambient temperature, cooling method and loss type (iron loss and copper loss). The equivalent loss impedance that is equivalent to a constant will lead to errors in the calculation results. It can be seen that in the above comparative embodiment, the accuracy of the transmission and transformation loss based on the preset value is limited, and the dynamic transmission and transformation loss prediction based on the transmission and transformation loss prediction model takes into account the real-time dynamic changes of the photovoltaic power generation grid-connected line, so that the transmission and transformation loss value added to the photovoltaic power generation grid-connected power limit control is more accurate, thereby improving the accuracy of the photovoltaic power generation grid-connected power limit control. In addition, the dynamic transmission and transformation loss prediction based on the transmission and transformation loss prediction model enhances the flexibility and scalability of the photovoltaic power generation grid-connected power limit control method implemented in the embodiment of the present invention. Flexibility and scalability are reflected in the following aspects: the power transmission and transformation loss prediction model can be flexibly modified or expanded according to different application scenarios and requirements, such as introducing more characteristic variables to improve the prediction accuracy of the power transmission and transformation loss prediction model and adapt to specific application scenarios and requirements, such as modifying the dimensions of characteristic variables to adapt to specific application scenarios and requirements.

[0076] For example, in a specific embodiment, in order to ensure the real-time and reliability of the transmission and transformation loss prediction, the sequence prediction network may adopt an LSTM network or the like.

[0077] As an improvement of the above embodiments, the following embodiment proposes an improved transmission and transformation loss prediction model, which introduces the voltage amplitude fluctuation characteristics and current amplitude fluctuation characteristics of the transformer in the photovoltaic power generation grid-connected line into the characteristic variables used as model input. The voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics need to be obtained through a feature extraction network. Specifically, the process of extracting the voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics by the feature extraction network may include:

[0078] The transformer voltage amplitude and transformer current amplitude collected at consecutive moments form a second time series;

[0079] The second time series is input into the constructed feature extraction network to extract the voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics at the continuous moment.

[0080] For example, the feature extraction network may adopt a CNN network. The CNN network includes a 2D convolution layer, a pooling layer, and a Flatten layer connected in sequence, where:

[0081] A 2D convolution layer, used for receiving a two-dimensional matrix converted from the second time series and performing feature extraction on the two-dimensional matrix;

[0082] The pooling layer is used to perform pooling operations on the feature maps output after feature extraction, such as maximum pooling;

[0083] The Flatten layer is used to convert the feature map output after the pooling operation into a one-dimensional vector.

[0084] Because the voltage amplitude fluctuation and current amplitude fluctuation of the transformer will cause changes in iron loss, copper loss, and the heat generation of the iron core and winding of the transformer winding, thereby affecting the power loss of the transformer, based on this, in the above-mentioned improved embodiment, the voltage amplitude fluctuation characteristics and current amplitude fluctuation characteristics of the transformer are introduced into the characteristic variables, thereby improving the prediction accuracy of the transmission and transformation loss.

[0085] Taking the above CNN network as an example, the extraction process of voltage amplitude fluctuation characteristics and current amplitude fluctuation characteristics is explained as follows:

[0086] 1) The transformer voltage amplitude and current amplitude at continuous moments are collected by sensors, etc., and the transformer voltage amplitude and current amplitude at continuous moments are used to form a second time series, which is converted into a two-dimensional matrix X∈R T×V Input CNN network, where dimension T represents the time step and dimension V represents the transformer voltage amplitude and transformer current amplitude.

[0087] 2) The spatial features of the two-dimensional matrix X are extracted through the 2D convolution layer. The convolution kernel used in the 2D convolution layer is represented as K∈R k×f , where k represents the time window size, f represents the feature window size, and the convolution operation performed at each position (t, v) in the two-dimensional matrix X can be expressed as:

[0088]

[0089] In the above formula, Y i [t,v] represents the output feature map obtained by performing a convolution operation on the position (t,v) using the i-th convolution kernel, K i [m,n] represents the element in the mth row and nth column of the i-th convolution kernel. The dot product symbol in the above formula represents the element multiplication including the bias term and the ReLU activation function.

[0090] 3) In order to reduce the spatial dimension of the data, the output feature map of the 2D convolutional layer is subjected to the maximum pooling operation to obtain the output feature map Z. In the pooling window, for each position (t′, v′):

[0091] Z[t′,v′]=max{(t,v)∈pool window(t′,v′)}Y[t,v];

[0092] 4) Use the Flatten layer to convert the output feature map Z into a one-dimensional vector so that it can be used as an input parameter of the transmission and transformation loss prediction model.

[0093] For example, in a specific embodiment, an incremental learning method is used when training the sequence prediction network in the initial state. The specific process may include:

[0094] Use historical data of feature variables to construct training samples;

[0095] Use training samples to pre-train and pre-optimize hyperparameters of sequence prediction networks;

[0096] Use the characteristic variables collected in real time to construct real samples;

[0097] Use real samples to perform incremental training and hyperparameter optimization on the pre-trained and hyperparameter pre-optimized sequence prediction network.

[0098] As in the above embodiment, the accuracy of the constructed power transmission and transformation loss prediction model is improved through incremental learning.

[0099] For example, in a specific embodiment, when training and optimizing parameters of the sequence prediction network in the initial state, a particle swarm optimization algorithm (PSO algorithm) is also used to optimize hyperparameters. The specific process may include:

[0100] Use historical data of feature variables to construct training samples;

[0101] Use training samples to pre-train and pre-optimize hyperparameters of sequence prediction networks;

[0102] Use the characteristic variables collected in real time to construct real samples;

[0103] Use real samples to perform incremental training and hyperparameter optimization on the pre-trained and hyperparameter-optimized sequence prediction network;

[0104] Among them, the particle swarm optimization algorithm is used to optimize the hyperparameters.

[0105] For example, the process of hyperparameter optimization using particle swarm optimization algorithm is as follows:

[0106] 1) Using the prediction accuracy of the sequence prediction network, the following evaluation function is designed:

[0107]

[0108] In the above formula, SCORE represents the evaluation function value, TP represents true positive examples, TN represents true negative examples, FP represents false positive examples, and FN represents false negative examples.

[0109] 2) Use each particle to represent a set of hyperparameters, use the above evaluation function to evaluate the position of the particle swarm on the validation set, and update the speed and position of the particle swarm.

[0110] Among them, the particle velocity update formula is: The particle position update formula is: In the above two formulas, ω represents the inertia weight, c1, c2 represent the learning factor, r1, r2 represent the random number, and v i represents the velocity of the ith particle, p i is the position of the ith particle, b i is the best position found by the ith particle so far, g is the global best position found by the group so far, and the size limit of the hyperparameter, that is, the boundary limit of the particle position, H bounds =[H min ,H max ], clip represents the clip function.

[0111] 3) Repeat the previous step to obtain the global optimal solution, which is the optimal hyperparameter of the sequence prediction network, and apply it to the pre-training and incremental training of the sequence prediction network.

[0112] As an improvement to the above embodiments, when the constructed MPC model is used to generate the optimal duty cycle, an optimization algorithm is used to optimize the predicted time domain length of the MPC model. The optimization algorithm may be a genetic optimization algorithm, a particle swarm optimization algorithm, or the like.

[0113] As in the above embodiment, the prediction time domain length N in the MPC model is adjusted in real time by using an optimization algorithm to fully capture the dynamic characteristics of photovoltaic power generation when it is connected to the grid, ensuring that the computational complexity and prediction error are not increased. Preferably, the iterative process of optimizing the prediction time domain length N is synchronized with the iterative process in the MPC model to better meet the real-time update of the prediction time domain length of the MPC model.

[0114] In a specific embodiment, the prediction time domain length of the MPC model is optimized using a particle swarm optimization algorithm, which specifically includes the following implementation steps:

[0115] Initialize the particle swarm, set the size, initial position and initial velocity of the particle swarm;

[0116] Calculate the fitness value of each particle according to the fitness function;

[0117] Execute particle swarm update step;

[0118] If the number of iterations is reached or the particle fitness value of the particle swarm reaches the expected value, the optimal individual in the current particle swarm is output as the optimal solution for the predicted time domain length, otherwise the particle swarm update step is executed again;

[0119] Among them, the particle swarm update step includes:

[0120] Determine the current individual optimal solution of each particle and the current global optimal solution of the particle swarm;

[0121] Update the speed and position of each particle;

[0122] The fitness function is an evaluation function that characterizes the degree of error between the output voltage of the DC / DC converter calculated by the MPC model and the target voltage value required to be output by the DC / DC converter determined according to the scheduling instruction.

[0123] In a specific embodiment, a genetic optimization algorithm is used to optimize the prediction time domain length of the MPC model, and the fitness function constructed is the same as the above-mentioned embodiment based on the particle swarm optimization algorithm.

[0124] Taking the optimization of the prediction time domain length of the MPC model by using the particle swarm optimization algorithm as an example, the process of real-time updating of the prediction time domain length can be specifically described as follows:

[0125] 1) Based on the mean square error (MSE) between the output voltage U0′ of the DC / DC converter calculated by the MPC model and the target voltage value U0 required to be output by the DC / DC converter determined according to the scheduling instruction, an evaluation function, i.e., a fitness function, is designed. The evaluation function is expressed as:

[0126]

[0127] In the above formula, score represents the evaluation function value, and n' represents the number of samples.

[0128] 2) Each particle represents a predicted time domain length. The evaluation function designed in the previous step is used to evaluate the position of each particle in the particle swarm, and the speed and position of each particle in the particle swarm are updated.

[0129] The formula for updating particle velocity is as follows: The formula for updating the particle position is as follows: In the above two formulas, ω' is the inertia weight, c1', c2' are learning factors, r1', r2' are random numbers, and v i ' is the velocity of the ith particle, p i ' is the position of the i-th particle, b i' is the best position found by the ith particle so far, g' is the global best position found by the group so far, and the size limit of the prediction time domain length, that is, the boundary limit of the particle position N bounds =[N min ,N max ].

[0130] 3) Repeat the previous step to find the position of the particle with the highest fitness, that is, the optimal prediction time domain length. The iteration process of the particle swarm is synchronized with the iteration process of the MPC model.

[0131] In a specific application of the above embodiment, the characteristic variables only include the power transmission and transformation state variables that change the power loss when photovoltaic power generation is connected to the grid, and the power transmission and transformation state variables include ambient temperature, load, and transformer usage time, etc. The power transmission and transformation loss prediction model is constructed using the LSTM neural network, and the particle swarm optimization algorithm is introduced to perform hyperparameter optimization when constructing the power transmission and transformation prediction model. Figure 3 The diagram shows the training and prediction process of the power transmission and transformation loss prediction model at this time.

[0132] In another specific application of the above embodiment, the characteristic variables also include transformer voltage amplitude fluctuation characteristics and transformer current amplitude fluctuation characteristics, a CNN network is used to extract the transformer voltage amplitude fluctuation characteristics and transformer current amplitude fluctuation characteristics, an LSTM neural network is used to construct a power transmission and transformation loss prediction model, and a particle swarm optimization algorithm is introduced to perform hyperparameter optimization when constructing the power transmission and transformation prediction model. Figure 4 The diagram shows the training and prediction process of the power transmission and transformation loss prediction model at this time.

[0133] In another specific application of the above embodiment, the photovoltaic power station adopts a Boost circuit as a DC / DC boost circuit, and the photovoltaic power station is connected to a 6kV thermal power plant system. Figure 5 The structure diagram of the power supply circuit from the photovoltaic power station where the DC / DC converter is located to the load is shown. Figure 6 The schematic diagram of the structure of the photovoltaic power generation system connected to a 6kV thermal power plant is shown. Figure 5 and Figure 6 In, P W represents the predicted transmission and transformation loss at the next moment, P C Indicates the predicted value of the load power required at the next moment, P PV It represents the predicted value of the maximum output power of the photovoltaic power station at the next moment. The CNN-PSO-LSTM model represents that the above embodiment combines the transformer voltage amplitude fluctuation characteristics extracted based on the CNN network, the transformer current amplitude fluctuation characteristics extracted based on the CNN network, and the transmission and transformation loss prediction model based on the hyperparameter optimization of the particle swarm optimization algorithm. The transmission and transformation loss prediction model is constructed based on the LSTM network.

[0134] Combination Figure 5 The structure diagram of the power supply circuit of the photovoltaic power station where the DC / DC converter is located to the load is shown in the figure. The construction process of the spatial state equation of the MPC model is as follows:

[0135] 1) Establish the spatial state equation of the Boost circuit:

[0136]

[0137] Wherein, x represents the state variable, A represents the state matrix, u represents the input vector, the input vector is the target voltage value output by the DC / DC converter determined according to the scheduling instruction, B represents the input matrix, represents the derivative of the state variable, y represents the output vector, the output vector is the difference between 1 and the predicted value of the duty cycle of the pulse modulation signal controlling the DC / DC converter, C represents the output matrix, and D represents the direct transfer matrix.

[0138] 2) Establish the spatial state equation of the MPC model:

[0139]

[0140] Among them, i(t) and u(t) are state variables, the target voltage value U0 of the DC / DC converter output determined by the scheduling instruction is the input, D represents the duty cycle, and D′=1-D is the output. Figure 5 In the circuit structure diagram shown: L It represents the resistance connected to the positive output terminal of photovoltaic power generation in the circuit structure. The output voltage of photovoltaic power generation is represented as Upv. i(t) represents the current flowing through the resistor r L The current, resistance r L Connected with inductor L, inductor L is far away from resistor r L One end is connected to the collector of transistor S and the positive electrode of diode, and the negative electrode of diode is connected to resistor r C The first end and the load R L The first end of the resistor r C The second end of the capacitor C is connected to the first end of the capacitor C, and the second end of the capacitor C is connected to the load R L The second end of each is connected to the emitter of transistor S, the emitter of transistor S is connected to the negative output terminal of photovoltaic power generation, the voltage across capacitor C is u(t), and the load R L The voltage across the two ends is the output voltage of the DC / DC converter.

[0141] In this specific application, the constraints of the MPC model may include: min ≤U0≤U max ; 0≤D′≤1, where U min Indicates the fluctuation lower limit of the target voltage value U0, Umax Indicates the upper limit of fluctuation of the target voltage value U0.

[0142] Based on the above constraints and the establishment of the spatial state equation, the initial MPC model is constructed, and the final MPC model is obtained after setting the sampling frequency, adjusting the parameters, and selecting the appropriate weights.

[0143] Based on the pre-construction of the CNN-PSO-LSTM model and the MPC model, the specific process of the photovoltaic power generation grid-connected power limit control in this specific application includes:

[0144] Step A1: Obtain the predicted value P of the load power required at the next moment C and the predicted maximum output power of the photovoltaic power station at the next moment P PV ;

[0145] Step A2: Use the CNN-PSO-LSTM model to predict the transmission and transformation loss value P at the next moment. W ;

[0146] Step A3: Compare the predicted value P of the load power required at the next moment C , the predicted value of the maximum output power of the photovoltaic power station at the next moment P PV and the transmission and transformation loss value P at the next moment W , if P W +P C ≥P PV , then the MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted maximum output power of the photovoltaic power station at the next moment. If P W +P C <P PV , then MPC control based on real-time optimization of the predicted time domain length is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the target power value. The target power value refers to the power value required to be output by the DC / DC converter determined according to the scheduling instruction.

[0147] Figure 7 The structure block diagram of the photovoltaic power generation grid-connected power limiting control device 400 according to an embodiment of the present invention is schematically shown. Specifically, the photovoltaic power generation grid-connected power limiting control device 400 provided by the embodiment of the present invention includes an acquisition module 410 and a power limiting control module 420, wherein:

[0148] An acquisition module 410 is used to acquire a predicted value of load power required at the next moment and a predicted value of maximum output power of the photovoltaic power station;

[0149] The power limiting control module 420 is used to determine whether the predicted maximum output power value of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted power required by the load at the next moment; if so, the constructed MPC model is used to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value; otherwise, the MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted maximum output power value of the photovoltaic power station at the next moment;

[0150] The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

[0151] In a specific embodiment, the photovoltaic power generation grid-connected power limiting control device 400 also includes a transmission and transformation loss prediction module, which is used to form a first time series from the characteristic variables collected at consecutive moments before judging whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, and input the first time series into the constructed transmission and transformation loss prediction model to predict the power loss value generated when the photovoltaic power generation is connected to the grid at the next moment, wherein the maximum collection moment of the characteristic variable is the current moment, and the characteristic variable includes the transmission and transformation state variable that changes the photovoltaic power generation grid-connected power loss, and the transmission and transformation loss prediction model is obtained by training the sequence prediction network in the initial state and optimizing the parameters.

[0152] In a specific embodiment, the characteristic variables also include the voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics of the transformer in the photovoltaic power generation grid-connected line. The voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics at continuous moments are obtained by extracting features from the second time series using a constructed feature extraction network. The second time series is composed of the transformer voltage amplitude and the transformer current amplitude collected at continuous moments.

[0153] In a specific embodiment, the feature extraction network includes a 2D convolutional layer, a pooling layer, and a Flatten layer connected in sequence, wherein:

[0154] A 2D convolution layer, used for receiving a two-dimensional matrix converted from the second time series and performing feature extraction on the two-dimensional matrix;

[0155] The pooling layer is used to perform pooling operations on the feature maps output after feature extraction;

[0156] The Flatten layer is used to convert the feature map output after the pooling operation into a one-dimensional vector.

[0157] In a specific embodiment, training and parameter optimization of the sequence prediction network in the initial state include:

[0158] Use historical data of feature variables to construct training samples;

[0159] Use training samples to pre-train and pre-optimize hyperparameters of sequence prediction networks;

[0160] Use the characteristic variables collected in real time to construct real samples;

[0161] Use real samples to perform incremental training and hyperparameter optimization on the pre-trained and hyperparameter-optimized sequence prediction network;

[0162] Among them, the particle swarm optimization algorithm is used to optimize the hyperparameters.

[0163] In a specific embodiment, the DC / DC converter uses a boost circuit as a DC / DC boost circuit, and the space state equation is expressed as:

[0164]

[0165] Wherein, x represents the state variable, A represents the state matrix, u represents the input vector, the input vector is the target voltage value output by the DC / DC converter determined according to the scheduling instruction, B represents the input matrix, represents the derivative of the state variable, y represents the output vector, the output vector is the difference between 1 and the predicted value of the duty cycle of the pulse modulation signal controlling the DC / DC converter, C represents the output matrix, and D represents the direct transfer matrix.

[0166] In a specific embodiment, the constraints of the MPC model include: the duty cycle is greater than or equal to 0 and less than or equal to 1; and the target power value is within a preset fluctuation range.

[0167] In a specific embodiment, when the constructed MPC model is used to generate the optimal duty cycle, an optimization algorithm is used to optimize the predicted time domain length of the MPC model.

[0168] In a specific embodiment, an optimization algorithm is used to optimize the prediction time domain length of the MPC model, including:

[0169] Initialize particle swarm;

[0170] Calculate the fitness value of each particle according to the fitness function;

[0171] Execute particle swarm update step;

[0172] If the number of iterations is reached or the particle fitness value of the particle swarm reaches the expected value, the optimal individual in the current particle swarm is output as the optimal solution for the predicted time domain length, otherwise the particle swarm update step is executed again;

[0173] Wherein, the particle swarm updating step comprises:

[0174] Determine the current individual extremum of each particle and the current global optimal solution of the particle swarm;

[0175] Update the speed and position of each particle;

[0176] The fitness function is an evaluation function that characterizes the degree of error between the output voltage of the DC / DC converter calculated by the MPC model and the target voltage value required to be output by the DC / DC converter determined according to the scheduling instruction.

[0177] In a specific embodiment, the power transmission and transformation state variables include ambient temperature, load, and transformer usage time.

[0178] In a specific embodiment, the photovoltaic power generation grid-connected power limiting control device 400 provided by the present invention can be implemented in the form of a computer program. The computer program can be used in the following manner: Figure 8 The computer device shown in the figure is run. The memory of the computer device can store various program modules constituting the photovoltaic power generation grid-connected power limiting control device 400, for example, Figure 7 The acquisition module 410 and the power limit control module 420 are shown. The computer program composed of various program modules enables the processor to execute the steps of the photovoltaic power generation grid-connected power limit control method of various embodiments of the present invention described in this specification.

[0179] Figure 8The block diagram of the composition of a computer device according to an embodiment of the present invention is schematically shown. Specifically, the computer device provided by the embodiment of the present invention may be a terminal. The computer device includes a processor A01, a network interface A02, a display screen A04, an input device A05 and a memory (not shown in the figure) connected via a system bus. Among them, the processor A01 of the computer device is used to provide computing and control capabilities. The memory of the computer device includes an internal memory A03 and a non-volatile storage medium A06. The non-volatile storage medium A06 stores an operating system B01 and a computer program B02. The internal memory A03 provides an environment for the operation of the operating system B01 and the computer program B02 in the non-volatile storage medium A06. The network interface A02 of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor A01, a photovoltaic power generation grid-connected power limiting control method is implemented. The display screen A04 of the computer device can be a liquid crystal display screen or an electronic ink display screen, and the input device A05 of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0180] The embodiment of the present invention further provides a machine-readable storage medium on which a computer program is stored. When the computer program is executed by a processor, the above-mentioned photovoltaic power generation grid-connected power limiting control method is implemented.

[0181] The embodiment of the present invention also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing the following method steps:

[0182] Obtain the predicted value of the load power required at the next moment and the predicted value of the maximum output power of the photovoltaic power station;

[0183] Determine whether the predicted maximum output power value of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted power required by the load at the next moment; if so, use the constructed MPC model to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value, otherwise, MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted maximum output power value of the photovoltaic power station at the next moment;

[0184] The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

[0185] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. Ordinary technicians in this field can understand and implement it without paying creative labor.

[0186] Those skilled in the art will appreciate that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0187] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0188] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A photovoltaic power generation grid-connected power limiting control method, characterized in that: The method comprises: Obtain the predicted value of the load power required at the next moment and the predicted value of the maximum output power of the photovoltaic power station; Determine whether the predicted maximum output power value of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted power required by the load at the next moment; if so, use the constructed MPC model to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value, otherwise, MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted maximum output power value of the photovoltaic power station at the next moment; The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

2. The photovoltaic power generation grid-connected power limiting control method according to claim 1, characterized in that: Before determining whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment, the method further includes: The first time series is formed by the characteristic variables collected at consecutive moments; The first time series is input into the constructed transmission and transformation loss prediction model to predict the power loss value generated when the photovoltaic power generation is connected to the grid at the next moment; Among them, the maximum collection time of the characteristic variable is the current time, the characteristic variable includes the power transmission and transformation state variable that changes the power loss of photovoltaic power generation grid-connected, and the power transmission and transformation loss prediction model is obtained by training and optimizing the parameters of the sequence prediction network of the initial state.

3. The photovoltaic power generation grid-connected power limiting control method according to claim 2, characterized in that: The characteristic variables also include the voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics of the transformer in the photovoltaic power generation grid-connected line. The voltage amplitude fluctuation characteristics and the current amplitude fluctuation characteristics at continuous moments are obtained by extracting features from the second time series using the constructed feature extraction network. The second time series is composed of the transformer voltage amplitude and the transformer current amplitude collected at continuous moments.

4. The photovoltaic power generation grid-connected power limiting control method according to claim 3 is characterized in that: The feature extraction network includes a 2D convolutional layer, a pooling layer, and a Flatten layer connected in sequence, wherein: A 2D convolution layer, used for receiving a two-dimensional matrix converted from the second time series and performing feature extraction on the two-dimensional matrix; The pooling layer is used to perform pooling operations on the feature maps output after feature extraction; The Flatten layer is used to convert the feature map output after the pooling operation into a one-dimensional vector.

5. The photovoltaic power generation grid-connected power limiting control method according to claim 3, characterized in that: Train and optimize the parameters of the initial sequence prediction network, including: Use historical data of feature variables to construct training samples; Use training samples to pre-train and pre-optimize hyperparameters of sequence prediction networks; Use the characteristic variables collected in real time to construct real samples; Use real samples to perform incremental training and hyperparameter optimization on the pre-trained and hyperparameter-optimized sequence prediction network; Among them, the particle swarm optimization algorithm is used to optimize the hyperparameters.

6. The photovoltaic power generation grid-connected power limiting control method according to claim 1, characterized in that: The DC / DC converter adopts a boost circuit as a DC / DC boost circuit, and the space state equation is expressed as: Wherein, x represents the state variable, A represents the state matrix, u represents the input vector, the input vector is the target voltage value output by the DC / DC converter determined according to the scheduling instruction, B represents the input matrix, represents the derivative of the state variable, y represents the output vector, the output vector is the difference between 1 and the predicted value of the duty cycle of the pulse modulation signal controlling the DC / DC converter, C represents the output matrix, and D represents the direct transfer matrix.

7. The photovoltaic power generation grid-connected power limiting control method according to claim 1, characterized in that: The constraints of the MPC model include: the duty cycle is greater than or equal to 0 and less than or equal to 1; the target power value is within a preset fluctuation range.

8. The photovoltaic power generation grid-connected power limiting control method according to claim 1, characterized in that: When the constructed MPC model is used to generate the optimal duty cycle, an optimization algorithm is used to optimize the prediction time domain length of the MPC model.

9. The photovoltaic power generation grid-connected power limiting control method according to claim 8, characterized in that: The optimization algorithm is used to optimize the prediction time domain length of the MPC model, including: Initialize particle swarm; Calculate the fitness value of each particle according to the fitness function; Execute particle swarm update step; If the number of iterations is reached or the particle fitness value of the particle swarm reaches the expected value, the optimal individual in the current particle swarm is output as the optimal solution for the predicted time domain length, otherwise the particle swarm update step is executed again; Wherein, the particle swarm updating step comprises: Determine the current individual extremum of each particle and the current global optimal solution of the particle swarm; Update the speed and position of each particle; The fitness function is an evaluation function that characterizes the degree of error between the output voltage of the DC / DC converter calculated by the MPC model and the target voltage value required to be output by the DC / DC converter determined according to the scheduling instruction.

10. The photovoltaic power generation grid-connected power limiting control method according to claim 2, characterized in that: The power transmission and transformation state variables include ambient temperature, load and transformer usage time.

11. A photovoltaic power generation grid-connected power limiting control device, characterized in that: The device comprises: An acquisition module is used to obtain the predicted value of the load power required at the next moment and the predicted value of the maximum output power of the photovoltaic power station; The power limit control module is used to determine whether the predicted value of the maximum output power of the photovoltaic power station at the next moment is greater than the sum of the power loss value generated when the photovoltaic power generation is connected to the grid and the predicted value of the power required by the load at the next moment; if so, the constructed MPC model is used to generate the optimal duty cycle, so that the power output by the DC / DC converter in the photovoltaic power station under the control of the pulse modulation signal with a duty cycle of the optimal duty cycle tracks the target power value; otherwise, the MPPT maximum power point tracking control is adopted on the DC / DC converter side, so that the output power of the DC / DC converter tracks the predicted value of the maximum output power of the photovoltaic power station at the next moment; The target power value is the power value required to be output by the DC / DC converter determined according to the dispatch instruction, and the spatial state equation of the MPC model is established based on the circuit in which the photovoltaic power station where the DC / DC converter is located supplies power to the load.

12. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the program, the photovoltaic power generation grid-connected power limiting control method described in any one of claims 1 to 10 is implemented.

13. A machine-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the photovoltaic power generation grid-connected power limiting control method according to any one of claims 1 to 10 is implemented.

Images