New energy power generation scene reduction method and device, electronic equipment, medium and product

Abstract

The invention relates to the technical field of electricity, in particular to a new energy power generation scene reduction method and device, electronic equipment, a medium and a product, and the method comprises the steps: obtaining an original new energy scene set and the occurrence probability of each scene in the original new energy scene set; inputting the original new energy scene set into a pre-trained twin network to obtain an embedded vector of each scene, and calculating a deep learning distance between any two scenes based on the embedded vector of each scene; and based on the occurrence probability of each scene and the deep learning distance between any two scenes, deleting the target scene meeting the preset deletion condition in the original new energy scene set to obtain a reduced representative scene, thereby solving the problems of low scene classification precision and low scene combination accuracy in the related technology, and improving the scene classification accuracy. The scene distinguishing capability is improved, and the scene merging accuracy is improved.

Description

Technical Field

[0001] This invention relates to the field of electrical technology, and in particular to a method, apparatus, electronic device, medium, and product for reducing new energy power generation scenarios. Background Technology

[0002] With the rapid expansion of wind and solar power installed capacity, the randomness, intermittency and strong fluctuation of new energy power output are constantly increasing. The access of a large number of new energy power plants makes the power balance characteristics and power flow distribution of the power system exhibit high uncertainty. Due to the large number of original scenarios, if they are directly used for power flow calculation, security constraint verification and scheduling optimization, it will result in a huge calculation scale and excessive solution time, which will be difficult to meet the real-time and engineering application requirements of large power grids. Therefore, it is necessary to reduce the scale of scenarios by scenario reduction while ensuring statistical representativeness.

[0003] In related technologies, the synchronous back-substitution reduction method is usually used to reduce the number of scenarios. Euclidean distance or Kantorovich distance-based probability metrics are typically used to measure the similarity between scenarios. Then, a small number of representative scenarios are selected and weights are redistributed through clustering, iterative matching, etc., which can maintain the probability distribution and statistical characteristics of new energy power generation scenarios to a certain extent.

[0004] However, the relevant technologies are mainly based on hourly power value differences, which makes it difficult to fully characterize the differences in dynamic forms such as ramp rate and fluctuation intensity of new energy power time series. This can easily lead to the misclassification of similar scenarios and the over-merging of scenarios with significant differences in form, resulting in the underestimation of potential risks, which urgently needs to be addressed. Summary of the Invention

[0005] This invention provides a method, device, electronic equipment, medium, and product for reducing new energy power generation scenarios, in order to solve the problems of low scenario differentiation accuracy and low scenario merging accuracy in related technologies, thereby improving scenario differentiation capability and scenario merging accuracy.

[0006] To achieve the above objectives, a first aspect of the present invention provides a method for reducing new energy power generation scenarios, comprising the following steps: obtaining an original set of new energy scenarios and the probability of occurrence of each scenario in the original set of new energy scenarios; inputting the original set of new energy scenarios into a pre-trained Siamese network to obtain an embedding vector for each scenario, and calculating a deep learning distance between any two scenarios based on the embedding vector of each scenario; and deleting target scenarios in the original set of new energy scenarios that meet preset deletion conditions based on the probability of occurrence of each scenario and the deep learning distance between any two scenarios to obtain a reduced representative scenario set.

[0007] Further, in some embodiments, the step of deleting target scenarios that meet preset deletion conditions from the original new energy scenario set to obtain a reduced representative scenario set based on the occurrence probability of each scenario and the deep learning distance between any two scenarios includes: initializing the original new energy scenario set according to the occurrence probability of each scenario based on the deep learning distance between any two scenarios to obtain an initialized new energy scenario set; determining whether the number of scenarios in the initialized new energy scenario set reaches a preset reduction scale; if the number of scenarios in the initialized new energy scenario set does not reach the preset reduction scale, then using the initialized new energy scenario set as a new original new energy scenario set, and re-executing the step of initializing the original new energy scenario set according to the occurrence probability of each scenario based on the deep learning distance between any two scenarios to obtain an initialized new energy scenario set, until the number of scenarios in the new initialized new energy scenario set reaches the preset reduction scale; and obtaining the reduced representative scenario set based on each representative scenario in the new initialized new energy scenario set and the probability of each representative scenario.

[0008] Furthermore, in some embodiments, after determining whether the number of scenarios in the initialized new energy scenario set has reached a preset reduction scale, the method further includes: if the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, then obtaining the reduced representative scenario set by assigning each representative scenario in the initialized new energy scenario set and the probability of each representative scenario.

[0009] Furthermore, in some embodiments, the step of initializing the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario to obtain the initialized new energy scenario set includes: determining the scenario with the least impact in the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario; and deleting the scenario with the least impact to obtain the initialized new energy scenario set.

[0010] Furthermore, in some embodiments, after obtaining the initialized new energy scenario set by deleting the scenario with the least impact, the method further includes: determining the scenario in the initialized new energy scenario set that is closest to the scenario with the least impact in terms of depth metric; and transferring the probability of the scenario with the least impact to the scenario that is closest to the scenario with the least impact in terms of depth metric.

[0011] Furthermore, in some embodiments, before inputting the original new energy scenario set into the pre-trained Siamese network, the method further includes: obtaining the original new energy scenario set and constructing a training sample set based on the original new energy scenario set; dividing the training sample set into a training set, a validation set, and a test set based on a preset partitioning ratio; constructing a target Siamese deep neural network, inputting the training set into the target Siamese deep neural network for training to obtain initial model parameters; based on the initial model parameters, inputting the validation set into the target Siamese deep neural network for performance evaluation, and adjusting the initial model parameters according to the performance evaluation results until the contrastive loss function of the validation set converges to obtain optimal model parameters; based on the optimal model parameters, inputting the test set into the target Siamese deep neural network for model testing, and obtaining the preset pre-trained Siamese network when the test results meet preset requirements.

[0012] According to the new energy power generation scenario reduction method provided in this embodiment of the invention, an original set of new energy scenarios and the occurrence probability of each scenario are obtained. This scenario set is then input into a pre-trained Siamese network to obtain an embedding vector for each scenario. Based on the embedding vectors, the deep learning distance between any two scenarios is calculated. Finally, combining the occurrence probability of each scenario with the deep learning distance, target scenarios in the original new energy scenario set that meet preset deletion conditions are deleted, resulting in a reduced representative scenario. This solves the problems of low scenario differentiation accuracy and low scenario merging accuracy in related technologies, improving scenario differentiation capability and scene merging accuracy.

[0013] To achieve the above objectives, a second aspect of the present invention provides a new energy power generation scenario reduction device, comprising: an acquisition module, configured to acquire an original set of new energy scenarios and the probability of occurrence of each scenario in the original set of new energy scenarios; a calculation module, configured to input the original set of new energy scenarios into a pre-trained Siamese network to obtain an embedding vector for each scenario, and calculate a deep learning distance between any two scenarios based on the embedding vector of each scenario; and a generation module, configured to delete target scenarios in the original set of new energy scenarios that meet preset deletion conditions based on the probability of occurrence of each scenario and the deep learning distance between any two scenarios to obtain a reduced representative scenario set.

[0014] Further, in some embodiments, the generation module is specifically used for: initializing the original new energy scene set based on the deep learning distance between any two scenes and the occurrence probability of each scene to obtain an initialized new energy scene set; determining whether the number of scenes in the initialized new energy scene set reaches a preset reduction scale; if the number of scenes in the initialized new energy scene set does not reach the preset reduction scale, then using the initialized new energy scene set as a new original new energy scene set, and re-executing the step of initializing the original new energy scene set based on the deep learning distance between any two scenes and the occurrence probability of each scene to obtain an initialized new energy scene set, until the number of scenes in the new initialized new energy scene set reaches the preset reduction scale; and obtaining the reduced representative scene set based on each representative scene in the new initialized new energy scene set and the probability of each representative scene.

[0015] Furthermore, in some embodiments, after determining whether the number of scenarios in the initialized new energy scenario set has reached a preset reduction scale, the generation module is further configured to: if the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, then obtain the reduced representative scenario set by taking each representative scenario in the initialized new energy scenario set and the probability of each representative scenario.

[0016] Furthermore, in some embodiments, the generation module is also used to: determine the scenario with the least impact in the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario; and delete the scenario with the least impact to obtain the initialized new energy scenario set.

[0017] Furthermore, in some embodiments, after obtaining the initialized new energy scenario set by deleting the scenario with the least impact, the generation module is further configured to: determine the scenario in the initialized new energy scenario set that is closest to the scenario with the least impact in terms of depth metric; and transfer the probability of the scenario with the least impact to the scenario that is closest to the scenario with the least impact in terms of depth metric.

[0018] Furthermore, in some embodiments, before inputting the original new energy scenario set into the pre-trained Siamese network, the computing module is specifically used for: acquiring the original new energy scenario set and constructing a training sample set based on the original new energy scenario set; dividing the training sample set into a training set, a validation set, and a test set based on a preset partitioning ratio; constructing a target Siamese deep neural network, inputting the training set into the target Siamese deep neural network for training to obtain initial model parameters; based on the initial model parameters, inputting the validation set into the target Siamese deep neural network for performance evaluation, and adjusting the initial model parameters according to the performance evaluation results until the contrastive loss function of the validation set converges to obtain optimal model parameters; based on the optimal model parameters, inputting the test set into the target Siamese deep neural network for model testing, and obtaining the preset pre-trained Siamese network when the test results meet preset requirements.

[0019] According to the new energy power generation scenario reduction device provided in this embodiment of the invention, an original set of new energy scenarios and the occurrence probability of each scenario are obtained. This scenario set is then input into a pre-trained Siamese network to obtain an embedding vector for each scenario. Based on the embedding vectors, a deep learning distance between any two scenarios is calculated. Finally, combining the occurrence probability of each scenario with the deep learning distance, target scenarios in the original new energy scenario set that meet preset deletion conditions are deleted, resulting in a reduced representative scenario. This solves the problems of low scenario differentiation accuracy and low scenario merging accuracy in related technologies, improving scenario differentiation capability and scene merging accuracy.

[0020] To achieve the above objectives, a third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the new energy power generation scenario reduction method as described in the above embodiments.

[0021] To achieve the above objectives, a fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the new energy power generation scenario reduction method as described in the above embodiments.

[0022] A fifth aspect of the present invention provides a computer program product, including a computer program that is executed to implement the new energy power generation scenario reduction method as described in the above embodiments.

[0023] Therefore, the present invention has the following beneficial effects: (1) This invention learns the embedded representation of the power generation time series of new energy through a twin 1-D CNN-Bi-LSTM network. The distance between scenes is improved from "time-by-time numerical difference" to "fluctuation pattern difference". It can simultaneously perceive dynamic features such as climbing rate, duration, and amplitude, making scene clustering and representative scene selection more in line with the physical reality of new energy output.

[0024] (2) Based on the overall process and ideas of the traditional scene reduction method, this invention only replaces the distance calculation link between scenes, without changing the existing processes such as scene generation, probability assignment and weight redistribution after reduction. It is easy to directly connect with existing simulation platforms, stochastic power flow and stochastic scheduling models, and the engineering implementation cost is low.

[0025] (3) This invention automatically partitions the scene by means of the power fluctuation intensity index of new energy power generation, and uses it as the physical label for training twin network, so that the depth measurement model corresponds to the new energy fluctuation mechanism, thereby improving the interpretability and controllability of the scene reduction process. Attached Figure Description

[0026] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 A flowchart of a method for reducing new energy power generation scenarios according to an embodiment of the present invention; Figure 2 A schematic diagram of a one-dimensional convolutional layer and a bidirectional long short-term memory network structure provided according to a specific embodiment of the present invention; Figure 3 A schematic diagram of a Siamese neural network architecture provided according to a specific embodiment of the present invention; Figure 4 A schematic diagram of an improved power generation scenario reduction process based on a Siamese deep neural network according to a specific embodiment of the present invention; Figure 5 This is a time-series broken line diagram of reduced typical scenario power generation output according to an embodiment of the present invention; Figure 6 This is a block diagram of a new energy power generation scenario reduction device provided according to an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of an electronic device provided according to an embodiment of the present invention. Detailed Implementation

[0027] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0028] The following describes the method, apparatus, electronic equipment, medium, and product for reducing new energy power generation scenarios according to embodiments of the present invention with reference to the accompanying drawings. First, the method for reducing new energy power generation scenarios according to embodiments of the present invention will be described with reference to the accompanying drawings.

[0029] Figure 1 A flowchart illustrating a method for reducing new energy power generation scenarios according to an embodiment of the present invention.

[0030] like Figure 1 As shown, the method for reducing the scope of this new energy power generation scenario includes the following steps: In step S101, the original set of new energy scenarios and the probability of occurrence of each scenario in the original set of new energy scenarios are obtained.

[0031] The original new energy scenario set refers to the set of new energy power generation scenarios generated based on historical measured data or probability models, and the occurrence probability of each scenario refers to the occurrence probability of each new energy power generation scenario.

[0032] Specifically, the embodiments of the present invention assume co-generation N Each of the following new energy power generation scenarios has a length of [length missing]. T At the nth time step, the nth The scene is denoted as: ; in, For the scene In time The active power output of wind power is expressed in MW. To eliminate dimensional differences, the power output is normalized using range normalization. ; in, , , The normalized dimensionless power is then used to assign an occurrence probability to each scene based on the scene generation method or historical statistical results. ,satisfy .

[0033] Furthermore, to characterize the fluctuation characteristics of each scene, the power increment between adjacent time steps is defined: ； Based on this, the fluctuation intensity for each scenario is calculated: ; in, For the first The scene in time step arrive Normalized power variation between For the first The volatility intensity index corresponding to each scenario The larger the value, the more frequent and larger the power ramp-up in the scenario, and the more drastic the fluctuation.

[0034] As one possible implementation method, embodiments of the present invention cover all scenarios. Statistical analysis was performed, and the 25th, 50th, and 75th percentiles were calculated and denoted as follows: Q 25 , Q 50 , Q 75 Based on this, the scene is automatically divided into four fluctuation intensity ranges: if J i ≤ Q 25 Then record the scene. i For low volatility range, label l i =1; if Q 25 < J i ≤ Q 50 Then record the scene. i The range is between low and medium fluctuations. l i =2; if Q 50 < J i ≤ Q 75 Then record the scene. i The range is characterized by moderate to strong fluctuations. l i =3; if J i > Q 75 This is then recorded as a high-fluctuation range, labeled as such. l i =4.

[0035] Furthermore, in some embodiments, before inputting the original new energy scenario set into the pre-trained Siamese network, the method further includes: obtaining the original new energy scenario set and constructing a training sample set based on the original new energy scenario set; dividing the training sample set into a training set, a validation set, and a test set based on a preset partitioning ratio; constructing a target Siamese deep neural network, inputting the training set into the target Siamese deep neural network for training to obtain initial model parameters; based on the initial model parameters, inputting the validation set into the target Siamese deep neural network for performance evaluation, and adjusting the initial model parameters according to the performance evaluation results until the contrastive loss function of the validation set converges to obtain the optimal model parameters; based on the optimal model parameters, inputting the test set into the target Siamese deep neural network for model testing, and obtaining the preset pre-trained Siamese network when the test results meet preset requirements.

[0036] For example, the normalized scene sequence As input to a Siamese network, the Siamese network structure consists of two subnetworks sharing parameters. Each subnetwork implements a mapping from a one-dimensional time series to a low-dimensional embedding vector, i.e.

[0037] in, Indicates having parameters Deep network mapping, For the scene The feature embedding vector, with dimension . , It consists of a series of one-dimensional convolutional layers and bidirectional long short-term memory (Bi-LSTM) network layers, where the one-dimensional convolutional layers extract local temporal window features as follows: ; in For convolution output features, The kernel length is 1. For convolution kernel weights, For bias, As a nonlinear activation function, the Bi-LSTM layer performs bidirectional temporal modeling on the sequence to obtain the final temporal features. Then, the embedding vector is obtained by mapping it to the embedding space through a fully connected layer:

[0038] in, This is the weight matrix. This is the bias vector.

[0039] For example, Figure 2 This is a schematic diagram of a one-dimensional convolutional layer and a bidirectional long short-term memory network structure according to a specific embodiment of the present invention. Figure 3This is a schematic diagram of a Siamese neural network architecture provided according to a specific embodiment of the present invention, combined with... Figure 2 Figure 3 After receiving time-series data of new energy power generation scenarios, the Siamese neural network first extracts the fluctuation features within the local time window of the power sequence through the convolutional and pooling layers of the 1-D CNN module. Then, it performs bidirectional time-series modeling through the Bi-LSTM module to capture the dynamic dependencies of the sequence, and finally outputs a low-dimensional embedding vector that can represent the scene features. The two sub-networks of the Siamese network share parameters and are respectively adapted to the scene. With Scene After performing the above feature extraction process and obtaining the corresponding embedding vectors, the L2 norm distance between the two is calculated. Then, the network parameters are optimized by comparing the loss function, so that the network can learn a deep distance metric that can accurately characterize the similarity of the fluctuation patterns of new energy scenarios.

[0040] In step S102, the original set of new energy scenarios is input into a pre-trained Siamese network to obtain the embedding vector of each scenario, and the deep learning distance between any two scenarios is calculated based on the embedding vector of each scenario.

[0041] For example, fluctuation intensity labels obtained based on scene segmentation l i Construct sample set pairs, for example, for any pair of scenarios ( i , j ):like l i = l j If the sample pair is positive, then the similarity label is determined as follows. y ij =1; if l i ≠ l j If the sample pair is negative, then the similarity label is determined as follows: y ij =0, defining the depth distance output by the Siamese network as:

[0042] in, Embedded vector The Each component; in this embodiment of the invention, a contrastive loss function is also selected to train the Siamese network. The contrastive loss function is: ; in, y ij∈{0,1} is the similarity label of the sample pair (i,j); m is the margin parameter (with a value of 1), which is used to constrain the minimum distance of negative sample pairs; L ij For sample pairs ( i , j The loss value.

[0043] In step S103, based on the occurrence probability of each scenario and the deep learning distance between any two scenarios, the target scenarios that meet the preset deletion conditions in the original new energy scenario set are deleted to obtain the reduced representative scenario set.

[0044] In some embodiments, based on the occurrence probability of each scenario and the deep learning distance between any two scenarios, target scenarios that meet preset deletion conditions in the original new energy scenario set are deleted to obtain a reduced representative scenario set. This includes: initializing the original new energy scenario set according to the occurrence probability of each scenario based on the deep learning distance between any two scenarios to obtain an initialized new energy scenario set; determining whether the number of scenarios in the initialized new energy scenario set reaches a preset reduction scale; if the number of scenarios in the initialized new energy scenario set does not reach the preset reduction scale, then the initialized new energy scenario set is used as a new original new energy scenario set, and the step of initializing the original new energy scenario set according to the occurrence probability of each scenario based on the deep learning distance between any two scenarios to obtain an initialized new energy scenario set is repeated until the number of scenarios in the new initialized new energy scenario set reaches the preset reduction scale; and obtaining the reduced representative scenario set based on each representative scenario in the new initialized new energy scenario set and the probability of each representative scenario.

[0045] Furthermore, in some embodiments, after determining whether the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, the method further includes: if the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, then the reduced representative scenario set is obtained by taking each representative scenario in the initialized new energy scenario set and the probability of each representative scenario.

[0046] The preset reduction scale refers to the number of typical scenarios that need to be retained in the final stage of engineering applications, as predetermined in the process. Furthermore, in some embodiments, based on the deep learning distance between any two scenarios, the original new energy scenario set is initialized according to the occurrence probability of each scenario to obtain the initialized new energy scenario set, including: based on the deep learning distance between any two scenarios, determining the scenario with the least impact in the original new energy scenario set according to the occurrence probability of each scenario; deleting the scenario with the least impact to obtain the initialized new energy scenario set.

[0047] As one possible implementation, in this embodiment of the invention, after training is completed, for each scene... Perform forward computation to obtain its embedding vector. For any two scenarios (i,j), the formula is used. As a deep learning distance between scenes, an N×N distance matrix is ​​constructed. Using deep learning distance matrix Based on this, the initial probability of each scenario is set to... ,in Let be the number of scenes within the cluster, at the th In the next iteration, for each candidate deletion scenario in the current retention set... The impact index of its deletion is calculated as follows: ; This means using the weighted distance between the scene and other scenes to measure the impact of deleting the scene on the overall probability distribution. It should be understood that the smaller the depth metric distance, the more similar the scene is to its neighboring scenes; that is, the more similar the scene is to its neighboring scenes. The smaller the size, the more suitable it is as a priority for deletion, minimizing the impact on the scenario. Delete to obtain the initialized set of new energy scenarios.

[0048] Furthermore, in some embodiments, after obtaining the initialized set of new energy scenarios by deleting the scenario with the least impact, the method further includes: determining the scenario in the initialized set of new energy scenarios that is closest to the scenario with the least impact in terms of depth metric; and transferring the probability of the scenario with the least impact to the scenario that is closest to the scenario with the least impact in terms of depth metric.

[0049] Find the scene that is closest to it in terms of depth metric in the remaining scene:

[0050] The probability of the deleted scene is transferred entirely to the nearest neighbor scene:

[0051] The probabilities of other scenarios remain unchanged. At this point, the total probability is still 1. The information of the deleted scenario is taken over by its most similar scenario in the feature space, thus realizing the view of new energy power generation scenarios and obtaining a reduced representative scenario set.

[0052] To enable those skilled in the art to better understand the new energy power generation scenario reduction method of the present invention, the following explanation will be provided in conjunction with specific embodiments.

[0053] Figure 4 This is a schematic diagram of an improved power generation scenario reduction process based on a Siamese deep neural network according to a specific embodiment of the present invention, as shown below. Figure 4As shown, the process first reads the original new energy scenarios and their occurrence probabilities, calculates the embedding vectors of each scenario using a trained Siamese network, obtains the depth distance matrix between scenarios based on the embedding vectors, and then initializes the representative scenario set according to the scenario probabilities. Subsequently, it is determined whether the number of representative scenarios reaches the preset scale M. If it does, the original scenarios are assigned to the nearest representative scenario according to the minimum depth distance. Then, the probabilities are aggregated in each cluster to calculate the weight of the representative scenario. Finally, the reduced representative scenario set is output, completing the entire scenario reduction process.

[0054] Furthermore, the effectiveness and rationality of the new energy power generation scenario reduction method of the present invention can also be verified through the accompanying drawings. Figure 5 This is a time-series broken line diagram of reduced typical scenario power generation output according to an embodiment of the present invention, as shown below. Figure 5 As shown, the power output of each scenario exhibits obvious fluctuation characteristics within the time series. In the early stage (such as time steps 0-20), the power output of some scenarios is relatively high. Subsequently, the power output of each scenario shows different fluctuations with time steps. Overall, it reflects the dynamic differences in power output under different new energy power generation scenarios and can intuitively reflect the differences in power change patterns of each scenario. This provides a theoretical basis for twin networks to extract features and calculate the similarity between scenarios.

[0055] According to the new energy power generation scenario reduction method provided in this embodiment of the invention, an original set of new energy scenarios and the occurrence probability of each scenario are obtained. This scenario set is then input into a pre-trained Siamese network to obtain an embedding vector for each scenario. Based on the embedding vectors, the deep learning distance between any two scenarios is calculated. Finally, combining the occurrence probability of each scenario with the deep learning distance, target scenarios in the original new energy scenario set that meet preset deletion conditions are deleted, resulting in a reduced representative scenario. This solves the problems of low scenario differentiation accuracy and low scenario merging accuracy in related technologies, improving scenario differentiation capability and scene merging accuracy.

[0056] Next, the new energy power generation scenario reduction device provided according to an embodiment of the present invention is described with reference to the accompanying drawings.

[0057] Figure 6 This is a block diagram of a new energy power generation scenario reduction device provided according to an embodiment of the present invention.

[0058] like Figure 6 As shown, the new energy power generation scenario reduction device 10 includes: an acquisition module 100, a calculation module 200, and a generation module 300.

[0059] The acquisition module 100 is used to acquire the original new energy scenario set and the occurrence probability of each scenario in the original new energy scenario set; the calculation module 200 is used to input the original new energy scenario set into a pre-trained Siamese network to obtain the embedding vector of each scenario, and calculate the deep learning distance between any two scenarios based on the embedding vector of each scenario; the generation module 300 is used to delete target scenarios in the original new energy scenario set that meet the preset deletion conditions based on the occurrence probability of each scenario and the deep learning distance between any two scenarios to obtain a reduced representative scenario set.

[0060] Further, in some embodiments, the generation module 300 is specifically used for: initializing an original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario to obtain an initialized new energy scenario set; determining whether the number of scenarios in the initialized new energy scenario set reaches a preset reduction scale; if the number of scenarios in the initialized new energy scenario set does not reach the preset reduction scale, then using the initialized new energy scenario set as a new original new energy scenario set, and re-executing the step of initializing the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario to obtain an initialized new energy scenario set, until the number of scenarios in the new initialized new energy scenario set reaches the preset reduction scale; and obtaining a reduced representative scenario set based on each representative scenario in the new initialized new energy scenario set and the probability of each representative scenario.

[0061] Furthermore, in some embodiments, after determining whether the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, the generation module 300 is further configured to: if the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, then obtain the reduced representative scenario set by taking each representative scenario and the probability of each representative scenario in the initialized new energy scenario set.

[0062] Furthermore, in some embodiments, the generation module 300 is also used to: determine the scenario with the least impact in the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario; and delete the scenario with the least impact to obtain the initialized new energy scenario set.

[0063] Furthermore, in some embodiments, after obtaining the initialized new energy scenario set by deleting the scenario with the least impact, the generation module 300 is further configured to: determine the scenario closest to the scenario with the least impact in the depth metric within the initialized new energy scenario set; and transfer the probability of the scenario with the least impact to the scenario closest to the scenario with the least impact in the depth metric.

[0064] Furthermore, in some embodiments, before inputting the original new energy scenario set into the pre-trained Siamese network, the computing module 200 is specifically used for: acquiring the original new energy scenario set and constructing a training sample set based on the original new energy scenario set; dividing the training sample set into a training set, a validation set, and a test set based on a preset partitioning ratio; constructing a target Siamese deep neural network, inputting the training set into the target Siamese deep neural network for training to obtain initial model parameters; based on the initial model parameters, inputting the validation set into the target Siamese deep neural network for performance evaluation, and adjusting the initial model parameters according to the performance evaluation results until the contrastive loss function of the validation set converges to obtain the optimal model parameters; based on the optimal model parameters, inputting the test set into the target Siamese deep neural network for model testing, and obtaining the preset pre-trained Siamese network when the test results meet the preset requirements.

[0065] It should be noted that the foregoing explanation of the embodiment of the method for reducing new energy power generation scenarios also applies to the new energy power generation scenario reduction device of this embodiment, and will not be repeated here.

[0066] According to the new energy power generation scenario reduction device provided in this embodiment of the invention, an original set of new energy scenarios and the occurrence probability of each scenario are obtained. This scenario set is then input into a pre-trained Siamese network to obtain an embedding vector for each scenario. Based on the embedding vectors, a deep learning distance between any two scenarios is calculated. Finally, combining the occurrence probability of each scenario with the deep learning distance, target scenarios in the original new energy scenario set that meet preset deletion conditions are deleted, resulting in a reduced representative scenario. This solves the problems of low scenario differentiation accuracy and low scenario merging accuracy in related technologies, improving scenario differentiation capability and scene merging accuracy.

[0067] Figure 7 This is a schematic diagram of an electronic device provided according to an embodiment of the present invention. The electronic device may include: The memory 701, the processor 702, and the computer program stored on the memory 701 and executable on the processor 702.

[0068] When the processor 702 executes the program, it implements the new energy power generation scenario reduction method provided in the above embodiments.

[0069] Furthermore, electronic devices also include: Communication interface 703 is used for communication between memory 701 and processor 702.

[0070] The memory 701 is used to store computer programs that can run on the processor 702.

[0071] The memory 701 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0072] If the memory 701, processor 702, and communication interface 703 are implemented independently, then the communication interface 703, memory 701, and processor 702 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 7 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0073] Optionally, in a specific implementation, if the memory 701, processor 702, and communication interface 703 are integrated on a single chip, then the memory 701, processor 702, and communication interface 703 can communicate with each other through an internal interface.

[0074] The processor 702 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement embodiments of the present invention.

[0075] In addition, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the above-described method for reducing new energy power generation scenarios.

[0076] In addition, embodiments of the present invention also provide a computer program product, including a computer program, which is executed to implement the above-mentioned method for reducing new energy power generation scenarios.

[0077] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "a plurality of" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0078] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0079] Although embodiments of the present invention have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for reducing new energy power generation scenarios, characterized in that, Includes the following steps: Obtain the original set of new energy scenarios and the probability of occurrence of each scenario in the original set of new energy scenarios; The original set of new energy scenarios is input into a pre-trained Siamese network to obtain the embedding vector of each scenario, and the deep learning distance between any two scenarios is calculated based on the embedding vector of each scenario. Based on the occurrence probability of each scenario and the deep learning distance between any two scenarios, the target scenarios that meet the preset deletion conditions in the original new energy scenario set are deleted to obtain a reduced representative scenario set.

2. The method according to claim 1, characterized in that, The process of deleting target scenarios that meet preset deletion conditions from the original new energy scenario set based on the occurrence probability of each scenario and the deep learning distance between any two scenarios to obtain a reduced representative scenario set includes: Based on the deep learning distance between any two scenarios, the original new energy scenario set is initialized according to the occurrence probability of each scenario to obtain the initialized new energy scenario set; Determine whether the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale; If the number of scenarios in the initialized new energy scenario set does not reach the preset reduction scale, the initialized new energy scenario is used as a new original new energy scenario set, and the step of initializing the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario to obtain the initialized new energy scenario set is executed again until the number of scenarios in the new initialized new energy scenario set reaches the preset reduction scale. The reduced representative scenario set is obtained based on each representative scenario in the newly initialized new energy scenario set and the probability of each representative scenario.

3. The method according to claim 2, characterized in that, After determining whether the number of scenarios in the initialized new energy scenario set has reached the preset reduction scale, the method further includes: If the number of scenarios in the initialized new energy scenario set reaches the preset reduction scale, then the reduced representative scenario set is obtained by taking each representative scenario in the initialized new energy scenario set and the probability of each representative scenario.

4. The method according to claim 2, characterized in that, The process of initializing the original new energy scenario set based on the deep learning distance between any two scenarios and the occurrence probability of each scenario to obtain the initialized new energy scenario set includes: Based on the deep learning distance between any two scenarios, the scenario with the least impact in the original new energy scenario cluster is determined according to the occurrence probability of each scenario; The initialized set of new energy scenarios is obtained by deleting the scenario with the least impact.

5. The method according to claim 4, characterized in that, After obtaining the initialized set of new energy scenarios by deleting the scenarios with the least impact, the process also includes: In the initialized set of new energy scenarios, determine the scenario that is closest to the scenario with the least impact under depth measurement; The probability of the scenario with the least impact is transferred to the scenario that is closest to the scenario with the least impact in terms of depth metric.

6. The method according to claim 1, characterized in that, Before inputting the original set of new energy scenarios into the pre-trained Siamese network, the following steps are also included: Obtain the original set of new energy scenarios and construct a training sample set based on the original set of new energy scenarios; Based on a preset partitioning ratio, the training sample set is divided into a training set, a validation set, and a test set; a target Siamese deep neural network is constructed, and the initial model parameters are obtained by inputting the training set into the target Siamese deep neural network for training. Based on the initial model parameters, the validation set is input into the target Siamese deep neural network for performance evaluation, and the initial model parameters are adjusted according to the performance evaluation results until the contrastive loss function of the validation set converges to obtain the optimal model parameters. Based on the optimal model parameters, the test set is input into the target Siamese deep neural network for model testing, and when the test results meet the preset requirements, the preset pre-trained Siamese network is obtained.

7. A device for reducing the scenario of new energy power generation, characterized in that, include: The acquisition module is used to acquire the original set of new energy scenarios and the probability of occurrence of each scenario in the original set of new energy scenarios; The calculation module is used to input the original set of new energy scenarios into a pre-trained Siamese network to obtain the embedding vector of each scenario, and to calculate the deep learning distance between any two scenarios based on the embedding vector of each scenario. The generation module is used to delete target scenarios that meet preset deletion conditions from the original new energy scenario set based on the occurrence probability of each scenario and the deep learning distance between any two scenarios to obtain a reduced representative scenario set.

8. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and capable of running on the processor, wherein the processor executes the program to implement the method for reducing new energy power generation scenarios as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method for reducing new energy power generation scenarios as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the new energy power generation scenario reduction method as described in any one of claims 1-6.

Images