Probabilistic power flow analysis method and system based on source-load interaction mechanism

Abstract

The application discloses a kind of probability power flow analysis method and system based on source-load interaction mechanism, the method includes: using source-load interaction relationship identification model based on the source-load data corresponding to target power grid identifies source-load interaction relationship, the model is obtained by iterative training to big model;Probability power flow analysis is carried out based on source-load interaction relationship and source-load data;Each round training includes: obtaining sample source-load data and its label;Sample source-load characteristics of sample source-load data are determined by big model and sample source-load characteristics are mapped into interactive relationship characteristics;In database, target prompt information matched with interactive relationship characteristics is obtained;Second sample source-load interaction relationship is generated using big model based on target prompt information and interactive relationship characteristics;Big model is trained in current round based on the similarity between first sample source-load interaction relationship and second sample source-load interaction relationship, can improve analysis result precision, provide more accurate reference information for power system planning and dispatching.

Description

Technical Field

[0001] This application relates to the field of electrical technology, and in particular to a probabilistic power flow analysis method and system based on a source-load interaction mechanism. Background Technology

[0002] In recent years, renewable energy generation has accounted for an increasingly larger proportion of total power generation. With the large-scale integration of renewable energy sources, which have intermittent generation characteristics, the uncertainty of the power system has increased dramatically. This increased uncertainty can easily impact the power grid, threatening its safe and stable operation. In the field of modern power system uncertainty analysis, probabilistic power flow analysis is widely used as an important analytical tool. Probabilistic power flow analysis can fully consider various random factors, and the analysis results can be used to provide comprehensive and important reference information for power system planning and operation scheduling, facilitating subsequent planning and scheduling of the power system.

[0003] In related technologies, probabilistic power flow analysis is typically performed directly using collected historical source-load data. This process involves randomly simulating and generating numerous uncertainties based on the historical source-load data for further analysis. However, these randomly simulated uncertainties may not accurately reflect the actual conditions of the power grid under analysis, resulting in low accuracy of the analysis results for that grid. Summary of the Invention

[0004] To address the aforementioned technical issues, this application proposes a probabilistic power flow analysis method and system based on a source-load interaction mechanism. This method enables the probabilistic power flow analysis results to better reflect the actual situation of the power grid under analysis, thereby improving the accuracy of the analysis results.

[0005] In a first aspect, embodiments of this application provide a probabilistic power flow analysis method based on a source-load interaction mechanism, including:

[0006] Based on the source-load data corresponding to the target power grid, a pre-trained source-load interaction relationship identification model is used to identify the source-load interaction relationship. The source-load interaction relationship identification model is obtained by iteratively training a large model.

[0007] Based on the source-load interaction relationship and the source-load data, probabilistic power flow analysis is performed;

[0008] Each round of training in the iterative training includes:

[0009] Obtain sample source charge data and its labels, wherein the labels indicate the first sample source charge interaction relationship;

[0010] Using the large model, the sample source load features of the sample source load data are determined, and the sample source load features are mapped into interaction relationship features;

[0011] Retrieve target prompt information that matches the interaction relationship features from the prompt information database;

[0012] Based on the target prompt information and the interaction relationship features, the large model is used to generate the second sample source-load interaction relationship;

[0013] Based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample, the large model is trained in the current round.

[0014] Optionally, the large model includes a pre-trained source charge feature extraction network, and the sample source charge features are obtained by using the source charge feature extraction network to extract features from the sample source charge data;

[0015] The training method of the source payload feature extraction network includes:

[0016] Obtain sample features and their corresponding data, wherein the modality of the data corresponding to the sample features is the same as the modality of the sample source load data;

[0017] The source payload feature extraction network to be trained is used to extract features from the data corresponding to the sample features to obtain the first feature;

[0018] The source payload feature extraction network to be trained is trained based on the difference between the first feature and the sample feature.

[0019] Optionally, the large model includes a mapping relationship network to be trained;

[0020] Specifically, the large model maps the sample source load features into interaction relationship features, including:

[0021] The sample source load features are input into the mapping relationship network, so that the mapping relationship network maps the sample source load features into the interaction relationship features according to its current network parameters;

[0022] The step of training the large model for the current round based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample includes:

[0023] Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, the current network parameters of the mapping relationship network are adjusted in the current round.

[0024] Optionally, the step of inputting the sample source charge features into the mapping relationship network, such that the mapping relationship network maps the sample source charge features into the interaction relationship features according to its current network parameters, includes:

[0025] The sample source load features are input into the mapping relationship network;

[0026] The mapping network maps the sample source load features to at least one interaction relationship sub-feature according to its current network parameters. The at least one interaction relationship sub-feature corresponds to at least one source load interaction relationship type. The at least one source load interaction relationship type is included in a plurality of preset source load interaction relationship types. An interaction relationship sub-feature represents the interaction relationship of the sample source load data under the corresponding source load interaction relationship type.

[0027] The at least one interaction relationship sub-feature is fused by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature.

[0028] Optionally, the current network parameters of the mapping relationship network include the current weights of each of the multiple source-load interaction relationship types. The process of fusing the at least one interaction relationship sub-feature by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature includes:

[0029] The mapping relationship network performs weighted fusion processing on the at least one interaction relationship sub-feature according to the current weight of each of the at least one source-load interaction relationship types to obtain the interaction relationship feature.

[0030] Optionally, the prompt information database contains multiple prompt information items that are associated one-to-one with the multiple source-load interaction relationship types. Each prompt information item includes a text representation indicating the corresponding source-load interaction relationship type. The target prompt information includes the prompt information associated with each of the at least one source-load interaction relationship types.

[0031] The step of generating a second sample source-load interaction relationship using the large model based on the target prompt information and the interaction relationship features includes:

[0032] The target prompt information and the interaction relationship features are input into the large model, so that the large model uses the target text representation as a reasoning prompt and the interaction relationship features as the reasoning basis to perform reasoning, thereby generating the second sample source-load interaction relationship. The target text representation refers to the text representation contained in the prompt information associated with each of the at least one source-load interaction relationship types.

[0033] Optionally, the similarity is determined by the similarity between the second feature and the third feature, wherein the second feature is obtained by extracting features from the source-load interaction relationship of the first sample, and the third feature is obtained by extracting features from the source-load interaction relationship of the second sample.

[0034] Optionally, the probabilistic power flow analysis based on the source-load interaction relationship and the source-load data includes:

[0035] Based on the source load data, a probability distribution model is constructed;

[0036] Based on the probability distribution model and the source-load interaction relationship, a probabilistic power flow analysis is performed using a pre-configured probabilistic power flow calculation method, wherein the probabilistic power flow calculation method includes any of the following: Monte Carlo simulation method, analytical method, and semi-invariant method.

[0037] Optionally, the probabilistic power flow calculation method includes at least the Monte Carlo simulation method, and the probabilistic power flow analysis based on the probability distribution model and the source-load interaction relationship, using a pre-configured probabilistic power flow calculation method, includes:

[0038] Random samples are generated using the probability distribution model, wherein the random samples indicate source-load correlation scenarios;

[0039] The random sample is corrected based on the source-load interaction relationship;

[0040] Probabilistic power flow analysis is performed using Monte Carlo simulation based on modified random samples.

[0041] Secondly, embodiments of this application provide a probabilistic power flow analysis system based on a source-load interaction mechanism, comprising:

[0042] The relationship identification module is used to identify the source-load interaction relationship based on the source-load data corresponding to the target power grid using a pre-trained source-load interaction relationship identification model, which is obtained by iteratively training a large model.

[0043] The probabilistic power flow analysis module is used to perform probabilistic power flow analysis based on the source-load interaction relationship and the source-load data;

[0044] Each round of training in the iterative training includes:

[0045] Obtain sample source charge data and its labels, wherein the labels indicate the first sample source charge interaction relationship;

[0046] Using the large model, the sample source load features of the sample source load data are determined, and the sample source load features are mapped into interaction relationship features;

[0047] Retrieve target prompt information that matches the interaction relationship features from the prompt information database;

[0048] Based on the target prompt information and the interaction relationship features, the large model is used to generate the second sample source-load interaction relationship;

[0049] Based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample, the large model is trained in the current round.

[0050] In summary, the embodiments of this application have at least the following beneficial effects:

[0051] In this embodiment of the application, source-load interaction relationships are identified using a pre-trained source-load interaction relationship identification model based on source-load data corresponding to the target power grid. This model is obtained through iterative training of a large model. Based on the source-load interaction relationships and the source-load data, probabilistic power flow analysis is performed. Each round of training in the iterative training includes: acquiring sample source-load data and its labels, where the labels indicate a first sample source-load interaction relationship; determining the sample source-load features of the sample source-load data using the large model; and mapping the sample source-load features to interaction relationship features; and retrieving the interaction relationship features from the prompt information database. The system generates target prompts based on the target prompts and the interaction relationship features. Then, it generates a second sample source-load interaction relationship using the large model. Based on the similarity between the first and second sample source-load interaction relationships, the large model is trained for the current round. Thus, the identified source-load interaction relationships accurately reflect the actual situation between the load side and the generation side of the target power grid. This allows for full consideration of the actual situation during probabilistic power flow analysis, making the probabilistic power flow analysis results closer to the actual situation of the power grid under analysis, improving the accuracy of the analysis results, and providing more accurate reference information for power system planning and operation scheduling. Attached Figure Description

[0052] Figure 1 This is a flowchart illustrating the probabilistic power flow analysis method based on the source-load interaction mechanism provided in the embodiments of this application;

[0053] Figure 2 This is a flowchart illustrating the training process of the source-load interaction relationship recognition model provided in the embodiments of this application;

[0054] Figure 3 This is a schematic diagram of the structure of the probabilistic power flow analysis system based on the source-load interaction mechanism provided in the embodiments of this application;

[0055] Figure 4 This is a schematic diagram of the structure of the computer device provided in the embodiments of this application. Detailed Implementation

[0056] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0057] In the description of this application, the terms "first," "second," "third," etc., are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first," "second," "third," etc., may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "a plurality of" means two or more. In the description of this application, the term "comprising" and its variations are open-ended, meaning "including but not limited to." The term "based on" means "at least partially based on." The term "according to" means "at least partially according to." The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments."

[0058] In the description of this application, it should be noted that, unless otherwise expressly specified and limited, the terms "installation," "connection," and "linking" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection between two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0059] In the description of this application, it should be noted that, unless otherwise defined, all technical and scientific terms used in this application have the same meaning as commonly understood by one of ordinary skill in the art. The terminology used in this application is for the purpose of describing specific embodiments only and is not intended to limit the application. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0060] Firstly, see [the following] Figure 1 The diagram illustrates a flowchart of a probabilistic power flow analysis method based on a source-load interaction mechanism provided in this application. The method includes steps S101-S102, as follows:

[0061] S101, Based on the source-load data corresponding to the target power grid, the source-load interaction relationship is identified using a pre-trained source-load interaction relationship identification model, which is obtained by iteratively training a large model.

[0062] In one example, S101 may include: inputting the source-load data corresponding to the target power grid into a pre-trained source-load interaction relationship recognition model for recognition, so as to obtain the recognized source-load interaction relationship.

[0063] It should be noted that the source-load interaction relationship / sample source-load interaction relationship described in the various embodiments of this application is suitable to indicate at least one of the following: the interaction relationship between load and generator output within the target area corresponding to the target power grid, the interaction relationship between load and generator output within any sub-area contained in the target area, the interaction relationship between a type of load and different types of generator output within the target area or any sub-area, and the interaction relationship between a type of generator output and different types of load within the target area or any sub-area.

[0064] S102, Based on the source-load interaction relationship and the source-load data, perform probabilistic power flow analysis.

[0065] In one example, S102 may include: constructing a probability distribution model based on the source-load interaction relationship and the source-load data; and performing probability power flow analysis using a pre-configured probability power flow calculation method according to the constructed probability distribution model, wherein the probability power flow calculation method includes any of the following: Monte Carlo simulation method, analytical method, and semi-invariant method.

[0066] Furthermore, based on the source-load interaction relationship and the source-load data, a probability distribution model can be constructed, which may include:

[0067] Based on the source data in the source load data, a first probability distribution model is constructed using a first distribution function.

[0068] Based on the load data in the source load data, a second probability distribution model is constructed using a second distribution function.

[0069] Based on the source-load interaction relationship and the source-load data, a third probability distribution model is constructed using a third distribution function.

[0070] Based on the first probability distribution model, the second probability distribution model, and the third probability distribution model, a joint probability distribution model is constructed as the constructed probability distribution model.

[0071] For example, at least one of the first distribution function, the second distribution function, and the third distribution function can be determined according to at least one of the following: normal distribution, Weibull distribution, log-normal distribution, exponential distribution, and uniform distribution.

[0072] Among them, see Figure 2 The diagram illustrates the flowchart of each round of training in the iterative training provided in this embodiment of the application. Each round of training includes steps S201-S205, as follows:

[0073] S201, Obtain sample source load data and its labels, wherein the labels indicate the first sample source load interaction relationship.

[0074] In one example, the sample source load data may include historical load data, generator output data, weather data, holiday information, etc. The label can be manually added to the sample source load data.

[0075] In one example, prior to S202, the acquired historical source load data can be preprocessed to serve as new historical source load data. This preprocessing can include at least one of the following: data cleaning, data standardization, and feature engineering. Data cleaning is suitable for handling missing values, outliers, and noise to ensure data quality and integrity. Data standardization is suitable for standardizing or normalizing data to facilitate model training. Feature engineering is suitable for extracting useful features, such as time features (hours, dates, seasons), weather features (temperature, humidity), and holiday markers.

[0076] In one example, before S202, a suitable large language model can be selected as the large model to be trained. For example, existing large language models such as Transformer, BERT (Bidirectional Encoder Representations from Transformers), and GPT (Generative Pre-trained Transformer) can be chosen. Then, the acquired / preprocessed historical source payload data is converted into a text format matching the selected large language model, and the converted data is serialized to obtain a long sequence, which serves as the sample source payload data for training. The sample source payload data for training is then divided into a training set and a validation set; for example, the ratio of the training set to the validation set can be 80%:20%.

[0077] S202, using the large model, determine the sample source load features of the sample source load data, and map the sample source load features into interaction relationship features.

[0078] In one example, S202 may include: encoding the sample source load data through an encoder contained in the large model to generate corresponding sample source load features, and mapping the sample source load features to interaction relationship features through a mapping relationship table pre-configured in the large model, wherein the mapping relationship table may record the mapping relationship between various types of sample source load features and various types of interaction relationship features.

[0079] In one example, the sample source payload data can be pre-converted into text format so that the large model can extract features from the text-format sample source payload data.

[0080] S203, retrieve target prompt information that matches the interaction relationship features from the prompt information database.

[0081] In one example, the prompt information database may contain multiple prompt information corresponding to various interaction relationship features, wherein one interaction relationship feature corresponds to at least one of the multiple prompt information. Thus, the target prompt information is all the prompt information corresponding to the mapped interaction relationship features.

[0082] In one example, all the prompts contained in the prompt database can be stored in text format.

[0083] S204, Based on the target prompt information and the interaction relationship features, the large model is used to generate the second sample source-load interaction relationship.

[0084] Understandably, adding target hints can make large models generate second sample source-load interaction relationships more efficiently and accurately.

[0085] In one example, the target cue information can be textual information suitable for describing the inference process related to interaction relationship features, so as to suggest how the large model can perform inference related to interaction relationship features more efficiently. Thus, S204 may include: inputting the target cue information and interaction relationship features into the large model, so that the large model first extracts the textual features of the target cue information, then uses the extracted textual features as inference cues, and performs inference based on the mapped interaction relationship features to obtain the second sample source-load interaction relationship.

[0086] In one example, prior to S205, the method may further include setting training parameters for the current epoch of training of a large model, wherein the training parameters to be set may include at least one of the following: hyperparameters, loss functions, and optimizers.

[0087] The hyperparameters may include at least one of the following: learning rate, batch size, and training epochs.

[0088] The loss function may include at least one of the following: mean squared error, cross-entropy loss.

[0089] The optimizer may include at least one of the following: Adam (Adaptive Moment Estimation) or SGD (Stochastic Gradient Descent).

[0090] S205, Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, the large model is trained in the current round.

[0091] In one example, the current training iteration in S205 may include: calculating a loss function based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, updating model parameters through backpropagation, and / or periodically (i.e., during a certain iteration of iterative training) evaluating model performance on a validation set to adjust hyperparameters and model structure based on the evaluation results, thereby completing the current training iteration. The evaluation result can be determined by calculating at least one of the following metrics: mean squared error, root mean square error, and coefficient of determination.

[0092] In one example, both the first and second sample source-load interactions can be represented in text form. Thus, a general text similarity model can be used to determine the similarity between the first and second sample source-load interactions to characterize the similarity. This similarity can be calculated by weighting the corresponding cosine similarity, Euclidean distance, Manhattan distance, and / or Pearson correlation coefficient.

[0093] In one example, the large model can also be integrated with any number of models from the above embodiments to improve the robustness and accuracy of predictions.

[0094] In one alternative implementation, the large model includes a pre-trained source charge feature extraction network, wherein the sample source charge features are obtained by extracting features from the sample source charge data using the source charge feature extraction network.

[0095] It should be noted that the source-load feature extraction network in this embodiment is pre-trained before iterative training of the large model, that is, before the current training round of the large model. Typically, a general feature extraction network with the same modality as the sample source-load data can be directly configured into the large model for direct use. Therefore, it is not necessary to train the source-load feature extraction network during iterative training of the large model, making the training of the large model more efficient. This is because the format of the sample source-load data is usually fixed and independent of the region where the power grid is located.

[0096] The training method of the source payload feature extraction network includes:

[0097] Obtain sample features and their corresponding data, wherein the modality of the data corresponding to the sample features is the same as the modality of the sample source load data.

[0098] The source-charge feature extraction network to be trained is used to extract features from the data corresponding to the sample features to obtain the first feature.

[0099] The source payload feature extraction network to be trained is trained based on the difference between the first feature and the sample feature.

[0100] In one example, the source payload feature extraction network to be trained can be composed of an encoder of the corresponding modality.

[0101] In one example, the difference between the first feature and the sample feature can be determined by the similarity between the first feature and the sample feature, which can be calculated by weighting the corresponding cosine similarity, Euclidean distance, Manhattan distance and / or Pearson correlation coefficient.

[0102] In one alternative implementation, the large model includes a mapping network to be trained.

[0103] Specifically, the large model maps the sample source load features into interaction relationship features, including:

[0104] The sample source load features are input into the mapping relationship network, so that the mapping relationship network maps the sample source load features into the interaction relationship features according to its current network parameters.

[0105] The step of training the large model for the current round based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample includes:

[0106] Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, the current network parameters of the mapping relationship network are adjusted in the current round.

[0107] In one example, the current network parameters of the mapping relationship network refer to the network parameters before the start of the current training round. These network parameters may include a pre-defined mapping relationship table, which may record the mapping relationships between various sample source load features and various interaction relationship features. It should be understood that the pre-defined various sample source load features may not include the extracted sample source load features. Therefore, the mapping process can be completed by matching the sample source load features with the highest similarity to the extracted sample source load features from the pre-defined various sample source load features. Thus, adjusting the current network parameters of the mapping relationship network in the current round may include adjusting the mapping relationship.

[0108] In one example, the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship can be determined based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship. This similarity can be obtained by weighting the corresponding cosine similarity, Euclidean distance, Manhattan distance and / or Pearson correlation coefficient.

[0109] In an optional implementation, the step of inputting the sample source charge features into the mapping relationship network, such that the mapping relationship network maps the sample source charge features into the interaction relationship features according to its current network parameters, includes:

[0110] The sample source load features are input into the mapping relationship network.

[0111] The mapping relationship network maps the sample source-load features to at least one interaction relationship sub-feature according to its current network parameters. The at least one interaction relationship sub-feature corresponds to at least one source-load interaction relationship type. The at least one source-load interaction relationship type is included in a plurality of preset source-load interaction relationship types. An interaction relationship sub-feature characterizes the interaction relationship of the sample source-load data under the corresponding source-load interaction relationship type.

[0112] The at least one interaction relationship sub-feature is fused by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature.

[0113] In one example, the pre-defined mapping table in the above embodiments may also record the mapping relationship between each type of sample source load feature and at least one type of interaction relationship sub-feature among various types of interaction relationship sub-features.

[0114] In one example, the fusion process may include splicing.

[0115] In one example, the source-load interaction relationship type can be used to indicate at least one of the following: different sub-regions in the target area corresponding to the target power grid, and the relationship between the various types of source devices and the various types of load devices connected to the target power grid.

[0116] In one optional implementation, the current network parameters of the mapping relationship network include the current weights of each of the plurality of source-load interaction relationship types. The process of fusing the at least one interaction relationship sub-feature by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature includes:

[0117] The mapping relationship network performs weighted fusion processing on the at least one interaction relationship sub-feature according to the current weight of each of the at least one source-load interaction relationship types to obtain the interaction relationship feature.

[0118] It is understandable that, when the current network parameters of the mapping relationship network include the current weights of each of the multiple source-load interaction relationship types, the process of adjusting the current network parameters of the mapping relationship network in the current round may also include adjusting the current weights in the current round, so that the adjusted current weights can be used as the current weights in the next round.

[0119] In one optional implementation, the prompt information database contains multiple prompt information items that are associated one-to-one with the plurality of source-load interaction relationship types. Each prompt information item includes a text representation indicating the corresponding source-load interaction relationship type. The target prompt information includes the prompt information associated with each of the at least one source-load interaction relationship type.

[0120] The step of generating a second sample source-load interaction relationship using the large model based on the target prompt information and the interaction relationship features includes:

[0121] The target prompt information and the interaction relationship features are input into the large model, so that the large model uses the target text representation as a reasoning prompt and the interaction relationship features as the reasoning basis to perform reasoning, thereby generating the second sample source-load interaction relationship. The target text representation refers to the text representation contained in the prompt information associated with each of the at least one source-load interaction relationship types.

[0122] In one optional implementation, the similarity is determined by the similarity between a second feature and a third feature, wherein the second feature is obtained by feature extraction of the source-load interaction relationship of the first sample, and the third feature is obtained by feature extraction of the source-load interaction relationship of the second sample.

[0123] In one example, the similarity between the second and third features can be calculated by weighting the corresponding cosine similarity, Euclidean distance, Manhattan distance, and / or Pearson correlation coefficient.

[0124] In one optional implementation, the probabilistic power flow analysis based on the source-load interaction relationship and the source-load data includes:

[0125] Based on the source load data, a probability distribution model is constructed.

[0126] Based on the probability distribution model and the source-load interaction relationship, a probabilistic power flow analysis is performed using a pre-configured probabilistic power flow calculation method, wherein the probabilistic power flow calculation method includes any of the following: Monte Carlo simulation method, analytical method, and semi-invariant method.

[0127] In one example, probabilistic power flow analysis is performed using a pre-configured probabilistic power flow calculation method based on the probability distribution model and the source-load interaction relationship. This may include: performing calculations using the probabilistic power flow calculation method according to the probability distribution model, and correcting the calculation results using the source-load interaction relationship to complete the probabilistic power flow analysis.

[0128] In one optional implementation, the probabilistic power flow calculation method includes at least the Monte Carlo simulation method, and the probabilistic power flow analysis based on the probability distribution model and the source-load interaction relationship, using a pre-configured probabilistic power flow calculation method, includes:

[0129] Random samples are generated using the probability distribution model, wherein the random samples indicate source-load correlation scenarios.

[0130] The random sample is corrected based on the source-load interaction relationship.

[0131] Probabilistic power flow analysis is performed using Monte Carlo simulation based on modified random samples.

[0132] Secondly, correspondingly, the embodiments of this application also provide a probabilistic power flow analysis system based on a source-load interaction mechanism, which can implement all the processes of the probabilistic power flow analysis method based on a source-load interaction mechanism provided in the above embodiments.

[0133] See Figure 3 The diagram illustrates the structure of a probabilistic power flow analysis system based on a source-load interaction mechanism provided in an embodiment of this application. The system includes:

[0134] The relationship identification module 301 is used to identify the source-load interaction relationship based on the source-load data corresponding to the target power grid using a pre-trained source-load interaction relationship identification model, wherein the source-load interaction relationship identification model is obtained by iteratively training a large model.

[0135] The probabilistic power flow analysis module 302 is used to perform probabilistic power flow analysis based on the source-load interaction relationship and the source-load data.

[0136] Each round of training in the iterative training includes:

[0137] Obtain sample source load data and its labels, wherein the labels indicate the first sample source load interaction relationship.

[0138] The large model is used to determine the sample source load features of the sample source load data, and to map the sample source load features into interaction relationship features.

[0139] Retrieve target prompt information that matches the interaction relationship features from the prompt information database.

[0140] Based on the target prompt information and the interaction relationship features, the large model is used to generate the second sample source-load interaction relationship.

[0141] Based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample, the large model is trained in the current round.

[0142] In one alternative implementation, the large model includes a pre-trained source charge feature extraction network, wherein the sample source charge features are obtained by extracting features from the sample source charge data using the source charge feature extraction network.

[0143] The training method of the source payload feature extraction network includes:

[0144] Obtain sample features and their corresponding data, wherein the modality of the data corresponding to the sample features is the same as the modality of the sample source load data.

[0145] The source-charge feature extraction network to be trained is used to extract features from the data corresponding to the sample features to obtain the first feature.

[0146] The source payload feature extraction network to be trained is trained based on the difference between the first feature and the sample feature.

[0147] In one alternative implementation, the large model includes a mapping network to be trained.

[0148] Specifically, the large model maps the sample source load features into interaction relationship features, including:

[0149] The sample source load features are input into the mapping relationship network, so that the mapping relationship network maps the sample source load features into the interaction relationship features according to its current network parameters.

[0150] The step of training the large model for the current round based on the similarity between the source-load interaction relationship of the first sample and the source-load interaction relationship of the second sample includes:

[0151] Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, the current network parameters of the mapping relationship network are adjusted in the current round.

[0152] In an optional implementation, the step of inputting the sample source charge features into the mapping relationship network, such that the mapping relationship network maps the sample source charge features into the interaction relationship features according to its current network parameters, includes:

[0153] The sample source load features are input into the mapping relationship network.

[0154] The mapping relationship network maps the sample source-load features to at least one interaction relationship sub-feature according to its current network parameters. The at least one interaction relationship sub-feature corresponds to at least one source-load interaction relationship type. The at least one source-load interaction relationship type is included in a plurality of preset source-load interaction relationship types. An interaction relationship sub-feature characterizes the interaction relationship of the sample source-load data under the corresponding source-load interaction relationship type.

[0155] The at least one interaction relationship sub-feature is fused by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature.

[0156] In one optional implementation, the current network parameters of the mapping relationship network include the current weights of each of the plurality of source-load interaction relationship types. The process of fusing the at least one interaction relationship sub-feature by the mapping relationship network according to its current network parameters to obtain the interaction relationship feature includes:

[0157] The mapping relationship network performs weighted fusion processing on the at least one interaction relationship sub-feature according to the current weight of each of the at least one source-load interaction relationship types to obtain the interaction relationship feature.

[0158] In one optional implementation, the prompt information database contains multiple prompt information items that are associated one-to-one with the plurality of source-load interaction relationship types. Each prompt information item includes a text representation indicating the corresponding source-load interaction relationship type. The target prompt information includes the prompt information associated with each of the at least one source-load interaction relationship type.

[0159] The step of generating a second sample source-load interaction relationship using the large model based on the target prompt information and the interaction relationship features includes:

[0160] The target prompt information and the interaction relationship features are input into the large model, so that the large model uses the target text representation as a reasoning prompt and the interaction relationship features as the reasoning basis to perform reasoning, thereby generating the second sample source-load interaction relationship. The target text representation refers to the text representation contained in the prompt information associated with each of the at least one source-load interaction relationship types.

[0161] In one optional implementation, the similarity is determined by the similarity between a second feature and a third feature, wherein the second feature is obtained by feature extraction of the source-load interaction relationship of the first sample, and the third feature is obtained by feature extraction of the source-load interaction relationship of the second sample.

[0162] In one optional implementation, the probabilistic power flow analysis based on the source-load interaction relationship and the source-load data includes:

[0163] Based on the source load data, a probability distribution model is constructed.

[0164] Based on the probability distribution model and the source-load interaction relationship, a probabilistic power flow analysis is performed using a pre-configured probabilistic power flow calculation method, wherein the probabilistic power flow calculation method includes any of the following: Monte Carlo simulation method, analytical method, and semi-invariant method.

[0165] In one optional implementation, the probabilistic power flow calculation method includes at least the Monte Carlo simulation method, and the probabilistic power flow analysis based on the probability distribution model and the source-load interaction relationship, using a pre-configured probabilistic power flow calculation method, includes:

[0166] Random samples are generated using the probability distribution model, wherein the random samples indicate source-load correlation scenarios.

[0167] The random sample is corrected based on the source-load interaction relationship.

[0168] Probabilistic power flow analysis is performed using Monte Carlo simulation based on modified random samples.

[0169] Thirdly, embodiments of this application provide a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, it implements the steps of the probabilistic power flow analysis method based on the source-load interaction mechanism described above.

[0170] Fourthly, embodiments of this application provide a computer program product, including computer instructions, which, when executed by a processor, implement the steps of the probabilistic power flow analysis method based on the source-load interaction mechanism described above.

[0171] Fifthly, embodiments of this application provide a computer device including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor executes the computer program to implement the steps of the probabilistic power flow analysis method based on the source-load interaction mechanism described above.

[0172] See Figure 4 The computer device in this embodiment includes a processor 401, a memory 402, and a computer program stored in the memory 402 and executable on the processor 401, such as a probabilistic power flow analysis program based on a source-load interaction mechanism. When the processor 401 executes the computer program, it implements the steps in the various probabilistic power flow analysis method embodiments based on the source-load interaction mechanism described above, for example... Figure 1 The steps S101-S102 are shown.

[0173] For example, the computer program may be divided into one or more modules / units, which are stored in the memory 402 and executed by the processor 401 to complete this application. The one or more modules / units may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the computer device.

[0174] The computer device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The computer device may include, but is not limited to, a processor 401 and a memory 402. Those skilled in the art will understand that the schematic diagram is merely an example of a computer device and does not constitute a limitation on the computer device. It may include more or fewer components than shown, or combine certain components, or different components. For example, the computer device may also include input / output devices, network access devices, buses, etc.

[0175] The processor 401 can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor, or processor 401 can be any conventional processor. The processor 401 is the control center of the computer device, connecting various parts of the entire computer device through various interfaces and lines.

[0176] The memory 402 can be used to store the computer programs and / or modules. The processor 401 implements various functions of the computer device by running or executing the computer programs and / or modules stored in the memory 402 and calling the data stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created according to the use of the mobile phone (such as audio data, phonebook, etc.). In addition, the memory 402 may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other volatile solid-state storage device.

[0177] Wherein, if the modules / units integrated into the computer device are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by the processor 401, it can implement the steps of the various method embodiments described above. Wherein, the computer program includes computer program code, which can be in the form of source code, object code, executable file, or some intermediate form, etc. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, portable hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0178] In summary, the embodiments of this application have at least the following beneficial effects:

[0179] In this embodiment of the application, source-load interaction relationships are identified using a pre-trained source-load interaction relationship identification model based on source-load data corresponding to the target power grid. This model is obtained through iterative training of a large model. Based on the source-load interaction relationships and the source-load data, probabilistic power flow analysis is performed. Each round of training in the iterative training includes: acquiring sample source-load data and its labels, where the labels indicate a first sample source-load interaction relationship; determining the sample source-load features of the sample source-load data using the large model; and mapping the sample source-load features to interaction relationship features; and retrieving the interaction relationship features from the prompt information database. The system generates target prompts based on the target prompts and the interaction relationship features. Then, it generates a second sample source-load interaction relationship using the large model. Based on the similarity between the first and second sample source-load interaction relationships, the large model is trained for the current round. Thus, the identified source-load interaction relationships accurately reflect the actual situation between the load side and the generation side of the target power grid. This allows for full consideration of the actual situation during probabilistic power flow analysis, making the probabilistic power flow analysis results closer to the actual situation of the power grid under analysis, improving the accuracy of the analysis results, and providing more accurate reference information for power system planning and operation scheduling.

[0180] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary hardware platforms, or it can be implemented entirely by hardware. Based on this understanding, all or part of the technical solutions of this application that contribute to the background technology can be embodied in the form of a software product. This computer software product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0181] The above description is the preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications are also considered to be within the scope of protection of this application.

Claims

1. A probability power flow analysis method based on source-load interaction mechanism, characterized in that, The method comprises the following steps: Based on the source-load data corresponding to the target power grid, a source-load interaction relationship is identified using a pre-trained source-load interaction relationship identification model, wherein the source-load interaction relationship identification model is obtained by iteratively training a large model; Based on the source-load interaction relationship and the source-load data, probabilistic power flow analysis is performed; Each round of training in the iterative training comprises the following steps: Obtain sample source-load data and its label, wherein the label indicates the first sample source-load interaction relationship; Determine the sample source-load characteristics of the sample source-load data by the large model, and map the sample source-load characteristics into interaction relationship characteristics; Obtain target prompt information matching the interaction relationship characteristics in the prompt information database; Based on the target prompt information and the interaction relationship characteristics, generate a second sample source-load interaction relationship using the large model; Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, perform current round training on the large model; The large model contains a mapping relationship network to be trained; Mapping the sample source-load characteristics into interaction relationship characteristics by the large model comprises the following steps: Input the sample source-load characteristics into the mapping relationship network, so that the mapping relationship network maps the sample source-load characteristics into the interaction relationship characteristics according to its current network parameters; Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, perform current round training on the large model comprises the following steps: Based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship, adjust the current network parameters of the mapping relationship network in the current round; The input of the sample source-load characteristics into the mapping relationship network comprises the following steps: Input the sample source-load characteristics into the mapping relationship network; Map the sample source-load characteristics into at least one interaction relationship sub-feature according to the current network parameters of the mapping relationship network, wherein the at least one interaction relationship sub-feature one-to-one corresponds to at least one source-load interaction relationship type, the at least one source-load interaction relationship type is included in a plurality of preset source-load interaction relationship types, and an interaction relationship sub-feature represents the interaction relationship embodied by the sample source-load data under the corresponding source-load interaction relationship type; Fuse the at least one interaction relationship sub-feature according to the current network parameters of the mapping relationship network to obtain the interaction relationship characteristics; The source-load interaction relationship is suitable for indicating at least one of the following: the interaction relationship between the load and the generator output in the target region corresponding to the target power grid, the interaction relationship between the load and the generator output in any sub-region included in the target region, the interaction relationship between a type of load and different types of generator output in the target region or any sub-region, and the interaction relationship between a type of generator output and different types of load in the target region or any sub-region. The similarity is determined by a similarity between a second feature and a third feature, the second feature is obtained by feature extraction on the first sample source-load interaction relationship, and the third feature is obtained by feature extraction on the second sample source-load interaction relationship.

2. The probabilistic power flow analysis method based on source-load interaction mechanism according to claim 1, wherein, The large model comprises a pre-trained source-load feature extraction network, and the sample source-load feature is obtained by feature extraction on the sample source-load data by using the source-load feature extraction network. The training manner of the source-load feature extraction network comprises: obtaining sample features and corresponding data, wherein the modalities of the data corresponding to the sample features are the same as the modalities of the sample source-load data; performing feature extraction on the data corresponding to the sample features by using a source-load feature extraction network to be trained to obtain first features; training the source-load feature extraction network to be trained based on the difference between the first features and the sample features.

3. The probabilistic power flow analysis method based on source-load interaction mechanism according to claim 1, wherein, The current network parameters of the mapping relationship network comprise current weights of the plurality of source-load interaction relationship types, and the fusion processing of the at least one interaction relationship sub-feature by the mapping relationship network according to the current network parameters thereof to obtain the interaction relationship feature comprises: weighted fusion processing of the at least one interaction relationship sub-feature by the mapping relationship network according to the current weights of the at least one source-load interaction relationship type to obtain the interaction relationship feature.

4. The probabilistic power flow analysis method based on source-load interaction mechanism according to claim 1, wherein, The prompt information database comprises a plurality of pieces of prompt information associated with the plurality of source-load interaction relationship types in one-to-one manner, one piece of prompt information comprises a text representation for indicating a corresponding source-load interaction relationship type, and the target prompt information comprises prompt information associated with the at least one source-load interaction relationship type. The generation of the second sample source-load interaction relationship by using the large model based on the target prompt information and the interaction relationship feature comprises: inputting the target prompt information and the interaction relationship feature into the large model, so that the large model performs reasoning with a target text representation as a reasoning prompt and with the interaction relationship feature as a reasoning basis, thereby generating the second sample source-load interaction relationship, wherein the target text representation refers to a text representation contained in the prompt information associated with the at least one source-load interaction relationship type.

5. The probabilistic power flow analysis method based on source-load interaction mechanism according to claim 1, wherein, The probability flow analysis based on the source-load interaction relationship and the source-load data comprises: constructing a probability distribution model according to the source-load data; performing probability flow analysis by using a pre-configured probability flow calculation method based on the probability distribution model and the source-load interaction relationship, wherein the probability flow calculation method comprises any one of a Monte Carlo simulation method, an analytical method, and a semi-invariant method.

6. The probabilistic power flow analysis method based on source-load interaction mechanism according to claim 5, wherein, The probability flow calculation method at least comprises the Monte Carlo simulation method, and the probability flow analysis by using the pre-configured probability flow calculation method based on the probability distribution model and the source-load interaction relationship comprises: generating random samples by using the probability distribution model, wherein the random samples indicate source-load correlation scenarios; correcting the random samples according to the source-load interaction relationship; and performing probability flow analysis by using a pre-configured probability flow calculation method based on the probability distribution model and the source-load interaction relationship. Based on the revised random sample, a probabilistic power flow analysis is performed using a Monte Carlo simulation method.

7. A probabilistic power flow analysis system based on source-load interaction mechanism, characterized in that, The method comprises the following steps: The relationship identification module is configured to identify source-load interaction relationships based on source-load data corresponding to a target power grid using a pre-trained source-load interaction relationship identification model, wherein the source-load interaction relationship identification model is obtained by iteratively training a large model. The probabilistic power flow analysis module is configured to perform probabilistic power flow analysis based on the source-load interaction relationships and the source-load data. Each round of training in the iterative training comprises the following steps: Obtain sample source-load data and its label, wherein the label indicates a first sample source-load interaction relationship. Determine sample source-load features of the sample source-load data using the large model, and map the sample source-load features into interaction relationship features. Obtain target prompt information matching the interaction relationship features from a prompt information database. Generate a second sample source-load interaction relationship using the large model based on the target prompt information and the interaction relationship features. Perform current round training of the large model based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship. The large model contains a mapping relationship network to be trained. Mapping the sample source-load features into interaction relationship features using the large model comprises the following steps: Input the sample source-load features into the mapping relationship network, so that the mapping relationship network maps the sample source-load features into the interaction relationship features according to its current network parameters. Perform current round adjustment of the current network parameters of the mapping relationship network based on the similarity between the first sample source-load interaction relationship and the second sample source-load interaction relationship. Input the sample source-load features into the mapping relationship network. Map the sample source-load features into at least one interaction relationship sub-feature according to the current network parameters of the mapping relationship network, wherein the at least one interaction relationship sub-feature one-to-one corresponds to at least one source-load interaction relationship type, the at least one source-load interaction relationship type is included in a plurality of preset source-load interaction relationship types, and an interaction relationship sub-feature represents the interaction relationship embodied by the sample source-load data under the corresponding source-load interaction relationship type. Fuse the at least one interaction relationship sub-feature according to the current network parameters of the mapping relationship network to obtain the interaction relationship features. ​ ​ The source-load interaction relationship is suitable for indicating at least one of an interaction relationship between a load and a generator output in a target region corresponding to a target power grid, an interaction relationship between a load and a generator output in any sub-region included in the target region, an interaction relationship between a load of a type and a generator output of a different type in the target region or any sub-region, and an interaction relationship between a generator output of a type and a load of a different type in the target region or any sub-region. The similarity is determined by a similarity between the second feature and the third feature, the second feature is obtained by performing feature extraction on the first sample source-load interaction relationship, and the third feature is obtained by performing feature extraction on the second sample source-load interaction relationship.

Images