Transformer capacity optimization method and device for photovoltaic access power grid, equipment and medium

Abstract

The invention discloses a transformer capacity optimization method and device for a photovoltaic access power grid and electronic equipment. The method comprises the steps that photovoltaic historical output related information and power grid historical load related information are acquired; inputting photovoltaic historical output related information into a trained first prediction model to predict photovoltaic future output data; inputting the historical load related information of the power grid into a trained second prediction model to predict future load data of the power grid; and inputting the predicted photovoltaic future output data and the power grid future load data into a preset optimization model so as to calculate and obtain the optimized transformer capacity of the distributed photovoltaic access power grid. According to the method, the photovoltaic processing and the power grid load are reasonably predicted and optimized to obtain the optimized photovoltaic access mode and the transformer capacity, so that the waste of resources is avoided, and the stability and the operation efficiency of the whole power system are further improved.

Description

Technical Field

[0001] The present application belongs to the technical field of power systems, and in particular, relates to a method, device and electronic equipment for optimizing the capacity of a transformer for photovoltaic access to a power grid. Background Art

[0002] Photovoltaic power generation is a clean energy technology that relies on solar energy to generate electricity. In today's rapidly developing energy field, photovoltaic power generation systems (PV) have become an important form of energy supply. Among them, the basic equipment of distributed photovoltaic power generation systems includes photovoltaic cell modules, photovoltaic array brackets, DC combiner boxes, DC distribution cabinets, grid-connected inverters, transformers, AC distribution cabinets and other equipment. Its operating mode is to convert the output of solar energy into DC distribution cabinets through DC combiner boxes under the condition of solar radiation, and then convert it into AC power by grid-connected inverters to supply grid loads. Excess or insufficient power is adjusted by connecting to the grid. However, photovoltaic power generation is affected by many factors such as weather and light, and its output is uncertain and volatile, which is significantly different from traditional deterministic power sources.

[0003] The impact of high-proportion distributed photovoltaic access to the grid has gradually become prominent, especially in rural mountainous areas where the grid structure is relatively weak. High-proportion photovoltaic access requires a significant increase in grid capacity and dispatching management capabilities, which has created huge investment pressure on the power system. The traditional distribution network structure and operation mode are difficult to cope with the random characteristics of photovoltaic power generation. If we follow the traditional model, only considering deterministic power sources and blindly increasing grid capacity to adapt to the volatility of photovoltaic power generation will lead to over-investment and waste of resources.

[0004] In addition, when photovoltaic power is connected to the grid, it is necessary to ensure that the photovoltaic grid-connected capacity matches the grid capacity, while considering the limitation of transformer capacity to ensure the stable operation of the grid and the quality of power supply. Therefore, it is very important to reasonably configure the transformer capacity, because it needs to match the construction scale and power demand of the photovoltaic power generation system. Too large capacity may lead to waste of resources, while too small capacity cannot meet the power demand of distributed photovoltaic power stations, thus affecting the stability and operation efficiency of the entire power system. Summary of the invention

[0005] The present application provides a method, device and electronic equipment for optimizing the capacity of a transformer for photovoltaic access to a power grid, aiming to solve the problems of resource waste caused by unreasonable configuration of transformer capacity in the prior art.

[0006] In a first aspect, the present application provides a method for optimizing the capacity of a transformer for photovoltaic access to a power grid, the method comprising:

[0007] Acquire historical target data; the target data includes weather data and power grid load data;

[0008] Call the trained prediction model to predict the future equipment power generation according to the historical target data; wherein the prediction model includes an intermediate quantity prediction model and a power generation prediction model, the historical target data is input into the intermediate quantity prediction model to output the future time target data, and the future time target data is input into the power generation prediction model to output the corresponding future equipment power generation.

[0009] In a second aspect, the present application provides a photovoltaic device power generation prediction device, the method comprising:

[0010] An acquisition module is used to acquire historical target data; the target data includes weather data and power grid load data;

[0011] A calling module is used to call a trained prediction model to predict the power generation of the equipment at future time based on the historical target data; wherein the prediction model includes an intermediate quantity prediction model and a power generation prediction model, the historical target data is input into the intermediate quantity prediction model to output the future time target data, and the future time target data is input into the power generation prediction model to output the corresponding future time equipment power generation.

[0012] In a third aspect, the present application provides an electronic device, comprising a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the above-mentioned method for optimizing the capacity of a transformer for photovoltaic grid access or the above-mentioned model training method is implemented.

[0013] In a fourth aspect, a storage medium stores computer-readable instructions thereon, wherein the computer-readable instructions are executed by one or more processors to implement the transformer capacity optimization method for photovoltaic grid access as described above or to implement the model training method as described above.

[0014] In a fifth aspect, a computer program product is provided, the computer program product comprising computer-readable instructions, the computer-readable instructions being stored in a storage medium, one or more processors of an electronic device reading the computer-readable instructions from the storage medium, loading and executing the computer-readable instructions, so that the electronic device implements the transformer capacity optimization method for photovoltaic grid access as described above or implements the above-mentioned model training method.

[0015] Compared with the prior art, the method of the present application is based on the use of supervised learning technology, and through the analysis and learning of historical data, the future output data of photovoltaic power generation and the future load data of the power grid are predicted; further, the optimized distributed photovoltaic access method is calculated based on the predicted future output data of photovoltaic power generation and the future load data of the power grid, and the optimized transformer capacity of the distributed photovoltaic access to the power grid is obtained. By reasonably predicting and optimizing photovoltaic processing and power grid load, the optimized photovoltaic access method and transformer capacity are obtained, thereby avoiding the waste of resources and further improving the stability and operation efficiency of the entire power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A schematic diagram of the main process of the method for optimizing the capacity of a transformer for photovoltaic power grid access provided in an embodiment of the present application;

[0018] Figure 2 for Figure 1 A schematic diagram of a method sub-flow diagram of the illustrated embodiment;

[0019] Figure 3 for Figure 2 A schematic diagram of a flow chart of a training method for a second network in the illustrated embodiment;

[0020] Figure 4 for Figure 1 A schematic diagram of a flow chart of a training method for a second prediction model in the illustrated embodiment;

[0021] Figure 5 for Figure 2 A schematic diagram of a structure of a specific example of the second network in the illustrated embodiment;

[0022] Figure 6 for Figure 1 A schematic structural diagram of a specific example of the second prediction model in the illustrated embodiment;

[0023] Figure 7 for Figure 6 A schematic diagram of the input results of the generator in the model of the illustrated embodiment;

[0024] Figure 8 for Figure 6 A schematic diagram of the input result of the discriminator in the model of the illustrated embodiment;

[0025] Fig. 9A schematic diagram of the loads at each node in the power grid topology structure implemented in this application;

[0026] Fig.10 A structural block diagram of a transformer capacity optimization device for photovoltaic grid access provided in an embodiment of the application;

[0027] Fig.11 A structural block diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0028] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0029] It will be understood by those skilled in the art that, unless expressly stated, the singular forms "one", "said", and "the" used herein may also include plural forms. It should be further understood that the term "comprising" used in the specification of the present disclosure refers to the presence of the features, integers, steps, operations, elements, and / or components, but does not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof. It should be understood that when we refer to an element as being "connected" or "coupled" to another element, it may be directly connected or coupled to the other element, or there may be intermediate elements. In addition, the "connection" or "coupling" used herein may include wireless connection or wireless coupling. The term "and / or" used herein includes all or any unit and all combinations of one or more associated listed items.

[0030] Figure 1 The main flow diagram of a method for optimizing the capacity of a transformer connected to a photovoltaic power grid provided in an embodiment of the present application is shown in FIG. The execution subject of the method for optimizing the capacity of a transformer connected to a photovoltaic power grid in an embodiment of the present application is a terminal device, which includes but is not limited to mobile devices such as smart phones, tablet computers, and computers, and may also include devices such as desktop computers. Figure 1 As shown, the method includes the following steps S100 to S400:

[0031] S100, obtaining photovoltaic historical output related information and grid historical load data.

[0032] The data acquisition equipment is used to collect the output information of each known distributed photovoltaic access point in the target area and the power parameters of the power grid load in the target area. The specific data to be collected include the historical photovoltaic output related information: such as the light intensity at each time point of each day in a short period of time after each distributed photovoltaic in the target area is connected to the distribution network, the photovoltaic power generation, the photovoltaic power generation error value, and the seasonal information, date information and temperature information in the corresponding time; the historical load data of the power grid: such as the basic power consumption of each load point of the power grid, the voltage loss of the unit length line, the power loss of the power grid per unit length line, and the seasonal information, date information and temperature information in the corresponding time.

[0033] S200 , inputting photovoltaic historical output related information into a trained first prediction model to predict photovoltaic future output data.

[0034] After the first model is trained with the information related to the historical photovoltaic output, the trained first prediction model can be used to predict the future photovoltaic output data based on the collected historical output information of the distributed photovoltaic access points.

[0035] In an exemplary embodiment, the first prediction model includes a first network and a second network; Figure 2 As shown, step S200 specifically includes the following steps S210 to S230:

[0036] S210: Inputting the photovoltaic historical output related information into the first network to output the photovoltaic output prediction value in the future target period.

[0037] In an exemplary embodiment, the first network is constructed using one or more Long Short-Term Memory (LSTM) networks. Figure 5 The LSTM structure diagram shown in FIG. 1 shows an example of an LSTM network structure. In one embodiment, the LSTM network structure includes an input gate, multiple hidden layers, and a forget gate inside an output gate. The input value and output value of the LSTM network can be expressed by the following formula:

[0038] Forget Gate:

[0039] f t =σ(W hf h t-1 +W xf x t +b f );

[0040] Input Gate:

[0041] i t =σ(W hi h t-1 +W xix t +b i );

[0042]

[0043]

[0044] Output Gate:

[0045] o t =σ(W ho h t-1 +W xo x t +b o );

[0046] h t =o t ⊙tanh(c t );

[0047] In the formula, f t ,i t , o t , are the output values ​​of the forget gate, input gate and output gate respectively; [W hf W xf ],[W hi W xi ],[W hg W xg ],[W ho W xo ], b f , b i , b g , b o are the network weights and biases of the four neural network layers; ⊙ is the Hadamard product operator of the matrix.

[0048] S220: Input the photovoltaic output prediction value in the future target period into the second network to output the photovoltaic output scenario in the future target period.

[0049] In an exemplary embodiment, the second network is constructed using a generative adversarial network; Figure 6 As shown, the generative adversarial network in this embodiment is a conditional generative adversarial network (CGAN), which includes a generator and a discriminator, and a CGAN model is used to construct a distributed photovoltaic output scenario.

[0050] In this embodiment, a model training method for the network is also provided, which is used to train the CGAN network; specifically, Figure 3As shown, the network training method includes the following steps S221 to S224:

[0051] S221, obtaining training samples; the training samples include photovoltaic historical output measured data used as actual samples and photovoltaic historical output prediction data used as corresponding condition labels.

[0052] Based on the actual measured data of photovoltaic historical output and the predicted data of photovoltaic historical output, the actual measured data of distributed photovoltaic output is used as the actual sample x=[x1, x2, ..., x k ], and the corresponding photovoltaic historical output forecast data is used as the conditional label, where k is the number of sequence points for each sample.

[0053] S222, fusing the actual sample with the random noise vector and inputting the resultant into the generator to output a generated sample.

[0054] See also Figures 6 to 8 , Figure 6 The input and output variables of the CGAN model are shown. The noise vector z is randomly sampled from a Gaussian distribution with a mean of 1 and a standard deviation of 1, and is consistent with the historical photovoltaic output forecast data. After horizontal splicing, they are input to the generator to output the generated samples; Figure 7 As shown, the task of the generator is to learn the photovoltaic processing characteristics and output photovoltaic output scenarios mapped with the photovoltaic historical output forecast data.

[0055] S223, fusing the output generated sample with the conditional label and fusing the actual sample with the conditional label, and then inputting them into the discriminator together to output the probability of the actual sample.

[0056] like Figure 8 As shown in Figure 2, the input of the discriminator comes from both actual samples and generated samples. The actual samples and generated samples are respectively compared with the historical photovoltaic output forecast data. After horizontal splicing, the data is input to the discriminator. The task of the discriminator is to determine the probability that its input sample is an actual sample under the photovoltaic historical output forecast data. Through adversarial training between the generator and the discriminator, the generator finally learns the mapping relationship between the photovoltaic historical output forecast data and the actual photovoltaic output. By inputting random noise vectors and photovoltaic historical output forecast data, it can generate photovoltaic output day-ahead scenarios that conform to the real scenario. These scenarios contain the possible range of changes in the predicted photovoltaic output, which can provide effective information for interval prediction. After that, the prediction interval can be constructed by interval estimation at different confidence levels.

[0057] S224. Obtain a trained second network by minimizing the losses of the generator and the discriminator.

[0058] During the training process of CGAN, the generator and discriminator are trained synchronously, and the overall training objective function is:

[0059]

[0060] Where D(·) represents the probability that the input sample determined by the discriminator is the actual sample; E represents the calculation expectation; z~P z (z) represents the Gaussian distribution P z (z) is the noise vector z sampled in; x~P r (x) represents the actual sample x sampled from the actual distribution.

[0061] S230. Determine photovoltaic output data for the future target period according to the photovoltaic output prediction value for the future target period and the photovoltaic output scenario for the future target period; the photovoltaic output data is used to indicate a variation range of the photovoltaic output.

[0062] In this embodiment, suppose that for the i-th group of test samples (y i , x i ), use the trained generator to generate M photovoltaic output scenarios Q = {x′ i，1 , x′ i，2 , ..., x′ i，M}. The average value of M photovoltaic scenes can be used as the true value u i Estimates:

[0063]

[0064] Assume that the generated PV scenario can be estimated unbiasedly to the true value u i , then the model variance estimate can be expressed as:

[0065]

[0066] Existing methods are difficult to directly and accurately estimate the noise variance Using the interval expansion method, directly from Expand It is expressed as:

[0067]

[0068] Therefore, the upper and lower bounds of the prediction interval are expressed as:

[0069] U f(α) (y i )=m(y i )+β·σ G (y i );

[0070] L f(α) (yi )=m(y i )-β·σ G (y i );

[0071] Where β = z 1-α / 2 d represents the expansion factor, z 1-α / 2 It is the function value corresponding to the α significance level in the standard normal distribution. The expansion factor needs to be calculated based on the samples in the validation set. Assuming there are N samples in the validation set, the factor of the i-th sample is obtained by the following formula:

[0072]

[0073] Where k is the number of sample sequence points. i Sorted in descending order, the N×αth factor is selected as the expansion factor (if it is not an integer, it is rounded down). After the expansion factor is determined based on the samples of the validation set, the upper and lower bounds of the prediction interval of the test set samples under the 1-α confidence interval can be determined.

[0074] S300, inputting the historical load data of the power grid into the trained second prediction model to predict the future load data of the power grid;

[0075] In an exemplary embodiment, the second prediction model includes a plurality of sub-models constructed based on each grid node, and each sub-model is constructed using an LSTM network; in this embodiment, the grid topology is as follows: Fig. 9 As shown, specifically, in the distribution network, the loads at level 2 and level 3 are dominant. Fig. 9 In the example, L1, L2, and L3 have n, h, and 1 nodes, respectively. Based on the collected grid load information in the partition, n model data are independently trained on the grid load data before the prediction time point based on the LSTM model. In power load prediction, LSTM can learn the time dependency and seasonality patterns in historical load data to predict future loads. In a specific implementation, Figure 4 As shown, the training method of the second prediction model includes the following steps S310 to S330:

[0076] S310, obtaining historical load data of each node power grid;

[0077] The historical load data of each node in the power grid is obtained as a training sample pair, which is used to independently train the sub-model of each node in the power grid.

[0078] S320, inputting the historical power grid load data of each node into the sub-model of the corresponding node to output the power grid load forecast data of each node;

[0079] According to the measured data of historical load of the power grid, the corresponding forecast data is predicted.

[0080] S330, training each of the sub-models by minimizing the loss between the grid load prediction data and the grid load measured data of each node to obtain a trained second prediction model.

[0081] In an exemplary embodiment, the training method for each sub-model based on the loss of power grid load prediction data and power grid load measured data can use the alternating direction method of multipliers (ADMM) to update the parameters of each sub-model; in the update iterative process, the federated averaging algorithm (FedAvg) is used to aggregate and update the parameters of each sub-model until the model converges or reaches the set number of iterations to obtain a trained second prediction model.

[0082] It can be known that for the load conditions of each node in the power grid, Fig. 9 It can be seen from the figure that the load is accumulated layer by layer. Without considering the network loss, the sum of the child node load should be equal to the parent node load, that is, the upper and lower layer loads should have aggregation consistency in value, which can be regarded as the correlation between multiple load objects and hierarchical load data. For the bottom load node, the objective function of collaborative prediction is shown as follows:

[0083]

[0084] Further,

[0085]

[0086] In the formula, f i (·) represents the load forecast value of L1 power supply area i; θ i and x i Represents model parameters and model input features; represents the load forecast value of L1 power supply area i; They represent the load forecast values ​​of h nodes in L2 respectively; Indicates the load forecast value of the L3 node.

[0087] For this distributed optimization problem, the ADMM algorithm is used to obtain the augmented Lagrangian function:

[0088]

[0089] In the formula, [θ1, θ2,..., θ n ] is the parameter to be optimized of n models; λ∈R r×1is the Lagrange multiplier; ρ is the penalty factor, ρ>0, ψ(·) means to find the mean of all elements in the vector. The execution steps of the iterative process of the obtained model are as follows:

[0090]

[0091] Where k and k+1 represent the kth and k+1th iteration numbers respectively.

[0092] In each iteration, n load forecasting models are independently trained in a distributed manner, and the models only need to exchange information through the alternating direction multiplier algorithm (ADMM). In the collaborative training process of multi-layer load forecasting models, the FedAvg method is used. The clients perform a round of training on their local data sets and upload the updated new models to the server. The server then solves the global model by averaging the model parameters, and then sends the new global model to the client for the next round of training until the set number of federated training iterations is reached, and the model training is completed. The training method combining the alternating direction multiplier algorithm (ADMM) and the FedAvg method has many advantages in data privacy protection, optimization efficiency, model robustness, system flexibility, convergence and accuracy, and scalability for training each sub-model based on the loss of power grid load forecast data and power grid load measured data.

[0093] S400: input the predicted future photovoltaic output data and future grid load data into a preset optimization model to calculate and obtain the optimized transformer capacity for distributed photovoltaic access to the grid.

[0094] Based on the predicted future photovoltaic output data and future grid load data, the access mode of distributed photovoltaics can be obtained, and after optimization according to the optimization model, the optimized access mode of distributed photovoltaics can be further determined, and the transformer capacity of the optimized photovoltaic access can be further calculated.

[0095] In an exemplary embodiment, based on the objective function of the optimization model and the grid constraints, the optimized transformer capacity of the distributed photovoltaic access to the grid is calculated; wherein the objective function is used to indicate the minimization of voltage deviation, network loss and investment cost. A multi-objective optimization model for the access of distributed photovoltaic to the distribution network is constructed, and voltage deviation, network loss and investment cost are used as indicators to achieve the minimum voltage drop, the lowest network loss and the minimum investment cost as the optimization objectives to construct a multi-objective optimization model, and the optimal transformer capacity of the distributed photovoltaic access to the weak structure grid is determined according to various constraints in the grid.

[0096] The optimization model in this embodiment sets voltage deviation, network loss and investment cost as optimization indicators. By reasonably selecting the weights of physical indicators and economic indicators and optimizing various constraints in the power grid through intelligent algorithms, the optimization of photovoltaic grid-connected capacity is achieved.

[0097] Specifically, the optimization objective function of investment cost can be:

[0098]

[0099] Where: r is the fixed annual interest rate; m is the service life of the photovoltaic equipment; c1 is the dynamic investment per kilowatt; c2 is the static investment per kilowatt; P k is the number of candidate point installations; N k The number of photovoltaic equipment to be installed.

[0100] The objective function for minimizing network loss can be:

[0101]

[0102] Where: L is the network loss function; N branch is the number of branches; I j is the current value of branch j; R j is the resistance value of branch j.

[0103] The goal of optimizing voltage quality can be:

[0104]

[0105] Where: Q is the voltage deviation function; U n , U nr is the rated voltage and actual voltage of node n; N node is the total number of nodes.

[0106] Since the three optimization objectives of lowest investment cost, smallest network loss and best voltage quality have different dimensions, the dimensions of the three optimization objectives are now unified, and the weight coefficient method is used to convert the three sub-objectives into a single objective, as shown in the following formula:

[0107]

[0108] Where: L is the network loss of the distribution system after the photovoltaic equipment is connected to the grid; L0 is the network loss in the original line; Q is the node voltage deviation; N node is the number of nodes in the power grid; C is the total investment cost after the photovoltaic equipment is installed and put into use; C min , C maxis the cost of connecting the photovoltaic equipment to the grid with the minimum and maximum permitted capacity; a, b, c are the penalty factors for the grid loss, voltage deviation and investment cost indicators; w1, w2, w3 are the weight factors of each indicator. The weights of grid loss, voltage deviation and investment cost can be set to 0.3, 0.3, 0.4 respectively, or adjusted according to actual usage needs.

[0109] It can be understood that the penalty factors a, b, and c are used to eliminate poor results during the optimization process and have no effect on the final optimization result. That is, when the optimized grid loss is greater than the grid loss before the photovoltaic equipment is connected to the grid, the parameter a is set to a larger positive number N, that is, when the grid loss increases, this solution is abandoned, otherwise, a=1; similarly, They are used to judge whether the solution results meet the requirements of optimal voltage quality and lowest investment cost.

[0110] In addition, the constraints of the power grid need to be met in the constructed multi-objective optimization model, which can include the following series of equations and constraints:

[0111] Power balance constraints:

[0112]

[0113] Where P s , Q s That is, the active and reactive power injected by the power supply at the head end of the distribution network; P PVi , Q PVi Inject active power and reactive power into photovoltaic; P load , Q load is the total system load; P loss , Q loss is the total network loss of the system.

[0114] Capacity constraints:

[0115] 0≤P i ≤P i.MAX ;

[0116] Where P i.MAX The maximum capacity of photovoltaic power that the grid allows to be connected.

[0117] Node voltage constraints:

[0118] U min ≤U nr ≤U max ;

[0119] Where U min , U max It is the minimum and maximum value of the voltage allowed in the power grid line.

[0120] Line current constraints:

[0121] I i ≤I imax ;

[0122] In the formula, I max It is the maximum value of the current allowed in the power grid line.

[0123] Permeability Constraint:

[0124]

[0125] Where η is the ratio of the total photovoltaic access capacity to the total grid load.

[0126] The improved immune algorithm is used to solve the function of the multi-objective optimization model, and the optimal solution is obtained according to the output of the objective function. The optimal access method of distributed photovoltaic access to the weak-structured power grid is obtained, and the optimal transformer capacity is calculated. The improved immune algorithm simulates the mechanism of the biological immune system and can search for the global optimal solution more efficiently to avoid falling into the local optimum. Compared with traditional optimization algorithms, such as particle swarm optimization and genetic algorithm, the improved immune algorithm has higher flexibility and adaptability in dealing with multi-objective optimization problems.

[0127] like Fig.10 As shown, an embodiment of the present application also provides a transformer capacity optimization device 160 for distributed photovoltaic access to the power grid, and the device includes an acquisition module 161, a first processing module 162, a second processing module 163 and a third processing module 164; wherein the acquisition module 161 is used to obtain information related to photovoltaic historical output and historical load data of the power grid; the first processing module 162 is used to input information related to photovoltaic historical output into a trained first prediction model to predict future photovoltaic output data; the second processing module 163 is used to input historical load data of the power grid into a trained second prediction model to predict future load data of the power grid; the third processing module 164 is used to input the predicted future photovoltaic output data and future load data of the power grid into a preset optimization model to calculate and obtain the optimized transformer capacity for distributed photovoltaic access to the power grid.

[0128] Optionally, the first processing module 162 includes a first processing unit, a second processing unit and a third processing unit, wherein the first processing unit is used to input information related to the historical photovoltaic output into the first network to output a photovoltaic output prediction value for the future target period; the second processing unit is used to input the photovoltaic output prediction value for the future target period into the second network to output a photovoltaic output scenario for the future target period; the third processing unit is used to determine the photovoltaic output data for the future target period based on the photovoltaic output prediction value for the future target period and the photovoltaic output scenario for the future target period; the photovoltaic output data is used to indicate a change range of the photovoltaic output.

[0129] Optionally, the second processing unit also includes a first model training unit, which is used to obtain training samples; the training samples include photovoltaic historical output measured data used as actual samples and photovoltaic historical output prediction data used as corresponding conditional labels; the actual samples are fused with random noise vectors and input into the generator to output generated samples; the output generated samples are fused with conditional labels and the actual samples are fused with conditional labels and then input into the discriminator together to output the probability of the actual samples; and the trained second network is obtained by minimizing the losses of the generator and the discriminator.

[0130] Optionally, the second processing module includes a second model training unit, which is used to obtain historical power grid load data of each node; input the historical power grid load data of each node into the sub-model of the corresponding node to output power grid load prediction data of each node; train each of the sub-models by minimizing the loss between the power grid load prediction data of each node and the power grid load actual measured data to obtain a trained second prediction model.

[0131] Optionally, the second model training unit includes a model training sub-unit, which is used to update the parameters of each sub-model using an alternating direction multiplier algorithm; during the update iteration process, the parameters of each sub-model are aggregated and updated using a federal averaging method until the model converges or reaches a set number of iterations to obtain a trained second prediction model.

[0132] Optionally, the third processing module includes an optimization unit, which is used to calculate the optimized transformer capacity of the distributed photovoltaic access to the power grid based on the objective function of the optimization model and the grid constraints; wherein the objective function is used to indicate the minimization of voltage deviation, network loss and investment cost.

[0133] See also Fig.11 In an embodiment of the present application, an electronic device 4000 is provided, and the electronic device 4000 may include: a desktop computer, a laptop computer, a server, etc.

[0134] exist Fig.11 In the embodiment, the electronic device 4000 includes at least one processor 4001 and at least one memory 4003.

[0135] The data interaction between the processor 4001 and the memory 4003 can be realized through at least one communication bus 4002. The communication bus 4002 may include a path for transmitting data between the processor 4001 and the memory 4003. The communication bus 4002 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. The communication bus 4002 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, only one thick line is used in the figure, but it does not mean that there is only one bus or one type of bus.

[0136] Optionally, the electronic device 4000 may further include a transceiver 4004, which may be used for data interaction between the electronic device and other electronic devices, such as data transmission and / or data reception. It should be noted that in actual applications, the transceiver 4004 is not limited to one, and the structure of the electronic device 4000 does not constitute a limitation on the embodiments of the present application.

[0137] Processor 4001 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array) or other programmable logic devices, transistor logic devices, hardware components or any combination thereof. It may implement or execute various exemplary logic blocks, modules and circuits described in conjunction with the disclosure of this application. Processor 4001 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.

[0138] The memory 4003 can be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program instructions or codes in the form of instructions or data structures and can be accessed by the electronic device 4000, but is not limited to this.

[0139] Computer-readable instructions are stored in the memory 4003 , and the processor 4001 can read the computer-readable instructions stored in the memory 4003 through the communication bus 4002 .

[0140] The computer-readable instructions are executed by one or more processors 4001 to implement the transformer capacity optimization method for photovoltaic grid access in the above-mentioned embodiments.

[0141] In addition, a storage medium is provided in an embodiment of the present application, on which computer-readable instructions are stored, and the computer-readable instructions are executed by one or more processors to implement the transformer capacity optimization method for photovoltaic grid access as described above.

[0142] In an embodiment of the present application, a computer program product is provided. The computer program product includes computer-readable instructions. The computer-readable instructions are stored in a storage medium. One or more processors of an electronic device read the computer-readable instructions from the storage medium, load and execute the computer-readable instructions, so that the electronic device implements the transformer capacity optimization method for photovoltaic grid access as described above.

[0143] Compared with the related art, the method of the present application is based on the use of supervised learning technology, through the analysis and learning of historical data, to predict the future output data of photovoltaic power and the future load data of the power grid; further, through the predicted future output data of photovoltaic power and the future load data of the power grid, and according to the optimization model, the optimized distributed photovoltaic access method is calculated, and the optimized distributed photovoltaic access to the power grid transformer capacity is obtained. By reasonably predicting and optimizing photovoltaic processing and power grid load, the optimized photovoltaic access method and transformer capacity are obtained, thereby avoiding the waste of resources and further improving the stability and operation efficiency of the entire power system. Furthermore, the prediction model of the present application uses LSTM network and CCAN network as the basic network, and combines the training method of alternating direction multiplier algorithm (ADMM) and federated average method (FedAvg) to train the prediction model, so as to have higher optimization efficiency, model robustness, system flexibility, convergence and accuracy, and scalability. In addition, in terms of optimization models, by reasonably selecting the weights of physical and economic indicators, and optimizing through intelligent algorithms for various constraints in the power grid, and combining multi-level optimization under supervised learning under random changes in photovoltaic power generation or extreme conditions, the optimization of photovoltaic grid-connected capacity under different scenarios is achieved.

[0144] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0145] The present application is described by some embodiments, and it is known to those skilled in the art that various changes or equivalent substitutions may be made to these features and embodiments without departing from the spirit and scope of the present application. In addition, under the teachings of the present application, these features and embodiments may be modified to adapt to specific circumstances and materials without departing from the spirit and scope of the present application. Therefore, the present application is not limited by the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of the present application are within the scope protected by the present application.

Claims

1. A method for optimizing the capacity of a transformer for photovoltaic access to a power grid, characterized in that: The method comprises: Obtain information related to the historical photovoltaic output and the historical grid load; Inputting the photovoltaic historical output related information into the trained first prediction model to predict the photovoltaic future output data; Inputting the information related to the historical load of the power grid into the trained second prediction model to predict the future load data of the power grid; The predicted future photovoltaic output data and future grid load data are input into the preset optimization model to calculate the optimized transformer capacity for distributed photovoltaic access to the grid.

2. The method according to claim 1, characterized in that The first prediction model includes a first network and a second network; the step of inputting photovoltaic historical output related information into the trained first prediction model to predict photovoltaic future output data includes: Inputting the photovoltaic historical output related information into the first network to output the photovoltaic output forecast value in the future target period; Inputting the photovoltaic output prediction value in the future target period into the second network to output the photovoltaic output scenario in the future target period; The photovoltaic output data for the future target period is determined according to the photovoltaic output forecast value for the future target period and the photovoltaic output scenario for the future target period; the photovoltaic output data is used to indicate a variation range of the photovoltaic output.

3. The method according to claim 2, characterized in that The second network is constructed based on a generative adversarial network, which includes a generator and a discriminator. The training method of the second network includes: Acquire training samples; the training samples include photovoltaic historical output measured data used as actual samples and photovoltaic historical output prediction data used as corresponding condition labels; The actual sample is fused with the random noise vector and then input into the generator to output the generated sample; The generated sample output is fused with the conditional label, and the actual sample is fused with the conditional label, and then inputted into the discriminator to output the probability of the actual sample; By minimizing the loss of the generator and the discriminator, a trained second network is obtained.

4. The method according to claim 1, characterized in that: The second prediction model includes a plurality of sub-models constructed based on each power grid node, and the training method of the second prediction model includes: Obtain historical load data of each node power grid; Input the historical load data of the power grid of each node into the sub-model of the corresponding node to output the predicted load data of the power grid of each node; Each of the sub-models is trained by minimizing the loss between the grid load prediction data and the grid load measured data of each node to obtain a trained second prediction model.

5. The method according to claim 4, characterized in that The step of training each of the sub-models by minimizing the loss between the power grid load prediction data and the power grid load measured data of each node to obtain a trained second prediction model includes: Each sub-model is trained independently by minimizing the actual data of the historical power grid load of each node and the corresponding historical power grid load forecast data; during the model training process, the alternating direction multiplier algorithm is used to update the parameters of each sub-model; during the update iteration process, the federal averaging method is used to aggregate and update the parameters of each sub-model until the model converges or reaches the set number of iterations to obtain a trained second prediction model.

6. The method according to any one of claims 1 to 5, characterized in that: The method of inputting the predicted future photovoltaic output data and future grid load data into a preset optimization model to calculate and obtain the optimized transformer capacity for distributed photovoltaic access to the grid includes: Based on the objective function of the optimization model and the grid constraints, the optimized transformer capacity of the distributed photovoltaic grid is calculated; wherein the objective function is used to indicate the minimization of voltage deviation, grid loss and investment cost.

7. The method according to claim 6, characterized in that The objective function formula is: Among them, L is the network loss of the distribution system after the photovoltaic equipment is connected to the grid; L0 is the network loss in the original line; Q is the node voltage deviation; N node is the number of nodes in the power grid; C is the total investment cost after the photovoltaic equipment is installed and put into use; C min , C max are the costs incurred by connecting photovoltaic equipment with the minimum and maximum permitted capacities to the grid; a, b, c are the penalty factors for network loss, voltage deviation and investment cost indicators; w1, w2, w3 are the weight factors of each indicator.

8. A transformer capacity optimization device for distributed photovoltaic access to a power grid, characterized in that: The device comprises: An acquisition module is used to obtain historical photovoltaic output information and historical grid load data; A first processing module is used to input the photovoltaic historical output related information into the trained first prediction model to predict the photovoltaic future output data; A second processing module is used to input the historical load data of the power grid into the trained second prediction model to predict the future load data of the power grid; The third processing module is used to input the predicted future photovoltaic output data and future grid load data into a preset optimization model to calculate and obtain the optimized transformer capacity for distributed photovoltaic access to the grid.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the transformer capacity optimization method for photovoltaic power grid access as described in any one of claims 1 to 7 when executing the computer program.

10. A storage medium having computer-readable instructions stored thereon, characterized in that: The computer-readable instructions are executed by one or more processors to implement the transformer capacity optimization method for photovoltaic grid access according to any one of claims 1 to 7.

Images