Training method of power distribution network voltage prediction model and power distribution network voltage prediction method

Abstract

This invention provides a training method for a distribution network voltage prediction model and a method for predicting distribution network voltage. The method includes: acquiring a training set; training a preset machine learning model based on the training set to obtain a first machine learning model; when the quality of the sample data in the training set does not meet a first preset requirement or the first prediction result of the first machine learning model does not meet a second preset requirement, performing data correction and data recombination processing on the training set based on the probability density function of the training set to obtain a recombined training set; using the recombined training set as a new training set, re-executing the steps of obtaining the first machine learning model based on the training set and subsequent steps until the quality of the sample data meets the first preset requirement and the first prediction result meets the second preset requirement, and then using the first machine learning model as the target distribution network voltage prediction model. This invention can train a high-precision distribution network voltage prediction model, thereby achieving accurate voltage prediction at each node in the distribution network.

Description

Technical Field

[0001] This invention relates to the field of distribution network voltage prediction technology, and in particular to a training method for a distribution network voltage prediction model and a method for predicting distribution network voltage. Background Technology

[0002] With the reform and development of the power system, and the integration of a high proportion of renewable energy into the distribution network, safety issues such as voltage exceeding limits and power flow changes have become increasingly prominent. Therefore, it is urgent to predict voltage change trends within the distribution network area to ensure the safe and stable operation of the distribution network.

[0003] In real-world scenarios, predicting voltage in distribution networks is an extremely challenging task. Typically, the actual topology of a distribution network is complex, and each node lacks high-precision data acquisition devices, making it difficult to establish accurate power flow models. Furthermore, due to the low voltage levels and diverse structures of distribution networks, lines and equipment may have difficult-to-obtain accurate operating parameters after long-term operation, thus hindering the development of effective and precise numerical models.

[0004] Furthermore, considering that the volatility of new energy sources such as photovoltaics and wind turbines is mainly affected by the external environment, how to achieve accurate prediction of distribution network voltage is a technical problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] This invention provides a training method for a distribution network voltage prediction model and a method for predicting distribution network voltage, in order to solve the problem of low accuracy in the prediction of distribution network voltage in the prior art.

[0006] In a first aspect, embodiments of the present invention provide a training method for a distribution network voltage prediction model, comprising:

[0007] Acquire a training set consisting of historical environmental data, historical power grid data, and historical voltage data corresponding to the historical environmental data and historical power grid data for each node in the distribution network;

[0008] The preset machine learning model is trained based on the training set to obtain the first machine learning model;

[0009] When the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, the training set is subjected to data correction and data recombination based on the probability density function of the training set to obtain a recombined training set.

[0010] The recombined training set is used as a new training set, and the steps of obtaining the first machine learning model based on the training set and subsequent steps are repeated until the sample data quality meets the first preset requirement and the first prediction result meets the second preset requirement. Then, the first machine learning model is used as the target distribution network voltage prediction model.

[0011] In one possible implementation, the step of performing data correction and data recombination on the training set based on the probability density function of the training set to obtain a recombined training set includes:

[0012] The optimal probability density function of the training set is determined based on the training set and the preset kernel function;

[0013] Based on the optimal probability density function, data is extracted from the training set to obtain the correction set;

[0014] The modified set and the training set are recombined and randomly mixed to obtain a recombined training set.

[0015] In one possible implementation, determining the optimal probability density function of the training set based on the training set and a preset kernel function includes:

[0016] Based on the training set and at least one preset kernel function, determine the probability density function corresponding to each preset kernel function;

[0017] Based on the optimization function, the optimal smoothing parameter is determined from the smoothing parameters corresponding to the at least one preset kernel function, which makes the function value of the probability density function optimal.

[0018] The optimal probability density function of the training set is determined based on the optimal smoothing parameter and the preset kernel function corresponding to the optimal smoothing parameter.

[0019] In one possible implementation, determining the probability density function corresponding to each preset kernel function based on the training set and at least one preset kernel function includes:

[0020] based on Determine the probability density function corresponding to each preset kernel function; where f h (x) represents the probability density function, and N represents the number of samples in the training set. Let h represent the preset kernel function, h represent the smoothing parameter of the preset kernel function, and x represent the random variable. r Let represent the r-th sample data, and d represent the sample dimension of the training set.

[0021] In one possible implementation, training a preset machine learning model based on the training set to obtain a first machine learning model includes:

[0022] The preset machine learning model is trained based on the training set and the initialization parameters of the preset machine learning model to obtain an intermediate first machine learning model;

[0023] Detect whether the second prediction result of the intermediate first machine learning model meets the second preset requirement;

[0024] When the second prediction result does not meet the second preset requirement, the target parameters in the intermediate first machine learning model are optimized to obtain the first machine learning model.

[0025] In one possible implementation, after detecting whether the second prediction result of the intermediate first machine learning model meets the second preset requirement, the method further includes:

[0026] When the second prediction result meets the second preset requirement, the intermediate first machine learning model is used as the first machine learning model, and the steps of "when the quality of the sample data of the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement" and subsequent steps are re-executed until the target distribution network voltage prediction model is obtained.

[0027] Secondly, embodiments of the present invention provide a method for predicting distribution network voltage, comprising:

[0028] Acquire real-time environmental data and power grid data of each node in the distribution network, and record the environmental data and power grid data as real-time power grid data;

[0029] The real-time power grid data of each node is input into the target distribution network voltage prediction model, and the predicted voltage of the node corresponding to the real-time power grid data is output; wherein, the target distribution network voltage prediction model is trained based on the method described in the first aspect above or any possible implementation of the first aspect.

[0030] Thirdly, embodiments of the present invention provide a training device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method as described in the first aspect or any possible implementation of the first aspect.

[0031] Fourthly, embodiments of the present invention provide a prediction device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method described in the second aspect above.

[0032] Fifthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the method as described in the first aspect, the second aspect, or any possible implementation of the first aspect above.

[0033] This invention provides a training method for a distribution network voltage prediction model and a method for predicting distribution network voltage. The method involves acquiring a training set comprised of historical environmental data, historical power grid data, and historical voltage data corresponding to the historical environmental and power grid data for each node in the distribution network. A preset machine learning model is then trained based on this training set to obtain a first machine learning model. If the quality of the sample data in the training set does not meet the first preset requirement, or the first prediction result of the first machine learning model does not meet the second preset requirement, it indicates that the data used for model training is insufficient to support training a model that meets the requirements, or the prediction quality of the model does not yet meet the requirements of the final model prediction quality. Therefore, the training set is then corrected and reorganized based on its probability density function to obtain a reorganized training set. This reorganized training set is then used as a new training set, and the model training process is repeated until the quality of the sample data in the training set meets the first preset requirement and the first prediction result of the first machine learning model meets the second preset requirement. The current first machine learning model is then used as the target distribution network voltage prediction model, thereby enabling accurate voltage prediction for each node in the distribution network. This invention can train a high-precision distribution network voltage prediction model, which facilitates the rapid and accurate prediction of the voltage at each node in the distribution network, obtains the voltage change trend of the distribution network, and provides certain data support for the safe and stable operation of the distribution network. Attached Figure Description

[0034] To more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0035] Figure 1 This is a flowchart illustrating the implementation of the training method for the distribution network voltage prediction model provided in this embodiment of the invention.

[0036] Figure 2 This is a flowchart illustrating the implementation of predicting distribution network voltage according to an embodiment of the present invention;

[0037] Figure 3 This is a flowchart illustrating the implementation of the power distribution network voltage prediction method provided in this embodiment of the invention.

[0038] Figure 4 This is an example diagram of the power distribution network topology provided in the embodiments of the present invention;

[0039] Figure 5 This is a spatiotemporal distribution diagram of the predicted voltage of the distribution network provided in an embodiment of the present invention;

[0040] Figure 6 This is a schematic diagram of the structure of the training device for the distribution network voltage prediction model provided in an embodiment of the present invention;

[0041] Figure 7 This is a schematic diagram of the structure of the power distribution network voltage prediction device provided in an embodiment of the present invention;

[0042] Figure 8 This is a schematic diagram of the training device provided in an embodiment of the present invention;

[0043] Figure 9 This is a schematic diagram of the prediction device provided in an embodiment of the present invention. Detailed Implementation

[0044] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0045] To make the objectives, technical solutions, and advantages of the present invention clearer, specific embodiments will be described below in conjunction with the accompanying drawings.

[0046] This invention provides a training method for a distribution network voltage prediction model. Figure 1 The flowchart illustrates the implementation of the training method for the distribution network voltage prediction model provided in this embodiment of the invention. Figure 1 As shown.

[0047] The method includes:

[0048] Step 101: Obtain a training set consisting of historical environmental data, historical power grid data, and historical voltage data corresponding to the historical environmental data and historical power grid data for each node in the distribution network.

[0049] In step 101, training a model with high prediction accuracy requires access to historical data that is accurate and correlated with voltage characteristics. In this embodiment, a training set for model training is constructed based on historical environmental data, historical power grid data, and historical voltage data of each node in the distribution network. This training set facilitates subsequent training of the distribution network voltage prediction model. For example, historical environmental data may include meteorological data such as irradiance, temperature, and wind speed. Historical power grid data may include power grid data such as node load, renewable energy installed capacity, and geographical distribution. This application does not limit the specific types of data in the historical environmental data and historical power grid data; those skilled in the art can determine the appropriate types as needed. By training the distribution network voltage prediction model based on the acquired historical environmental data, historical power grid data, and historical voltage data, the voltage of the distribution network after a high proportion of renewable energy is integrated into the distribution network can be accurately perceived.

[0050] In addition, after obtaining the training set, a data preprocessing process can be included. Optionally, the data preprocessing process may include operations such as imputing missing data in the training set and correcting or removing outlier data to obtain a high-quality training set, thereby providing good data support for the subsequent training of the distribution network voltage prediction model.

[0051] Step 102: Train the preset machine learning model based on the training set to obtain the first machine learning model.

[0052] In step 102, a preset machine learning model is trained based on the acquired training set to obtain a first machine learning model. For example, the model can be trained based on the Gradient Boosting Decision Tree (GBDT) algorithm or the XGBoost (Extreme Gradient Boosting) algorithm, and this application does not limit it to this.

[0053] For example, this embodiment uses the XGBoost algorithm to train a machine learning model to illustrate the basic principle:

[0054] First, establish the objective function for the GBDT model:

[0055]

[0056] in, This represents the first prediction result for sample i after the t-th iteration. f represents the second prediction result for the first t-1 trees. t (x i () represents the predicted value of the t-th tree, x i This represents the i-th sample data.

[0057] A regularization term is added to the objective function to reduce model complexity and prevent overfitting by limiting the number and values ​​of leaf nodes in the GBDT model. The formula for calculating the regularization term is as follows:

[0058]

[0059] Where w represents the weight of the leaf node of tree q(x), q(x) represents the structure of the tree, Ω(f) represents the complexity of the tree, γ is a hyperparameter, T is the number of leaf nodes, and w j λ represents the leaf node value, and λ is the penalty coefficient.

[0060] Construct the original objective function (i.e., the original objective function of the pre-defined machine learning model). The original objective function of the pre-defined machine learning model includes a loss function and a regularization term based on tree complexity. The formula for the original objective function is:

[0061]

[0062] Where n represents the number of samples, l represents the loss function, and y i t represents the true value, and t represents the number of iterations.

[0063] In the t-th iteration, the model's prediction equals the sum of the predictions from the previous t-1 iterations plus the prediction from the t-th tree. That is, in the t-th iteration, the model's objective function is:

[0064]

[0065] Wherein, Ω(f t Let t represent the complexity of the t-th tree.

[0066] In the objective function Treat it as a whole, solve for the values ​​of the whole, and in Performing a second-order Taylor expansion on the objective function, we obtain:

[0067]

[0068]

[0069]

[0070] Among them, g i h represents the first derivative of the loss function. i This represents the second derivative of the loss function.

[0071] Then, the constant term in the objective function is removed, and the function f(x) and the regularization term are substituted into the objective function to obtain:

[0072]

[0073] Define the set of samples at each leaf node as: I j ={i|q(x i The sample accumulation operation is transformed into a leaf node operation, resulting in the final objective function:

[0074]

[0075] Among them, G j H represents the sum of the first-order partial derivatives contained in leaf node j. j This represents the sum of the second-order partial derivatives contained in the leaf node j.

[0076] When the structure of the tree is determined, i.e., q(x) i When the objective function is determined, to minimize the objective function, the derivative is set to 0, and the optimal leaf node value can be solved. and the optimal objective function obj * .

[0077]

[0078]

[0079] At this point, the training process for the distribution network voltage prediction model based on the XGBoost algorithm is complete.

[0080] In this embodiment, the distribution network voltage prediction model trained based on the XGBoost algorithm has a fast data processing speed and a huge data processing scale. It also supports multiple base learners and allows for custom loss functions, thereby obtaining a prediction model with high prediction accuracy, which is beneficial for subsequent distribution network voltage prediction.

[0081] In one possible implementation, a first machine learning model is obtained by training a pre-defined machine learning model based on a training set, including:

[0082] The preset machine learning model is trained based on the training set and the initialization parameters of the preset machine learning model to obtain the intermediate first machine learning model.

[0083] Check whether the second prediction result of the first intermediate machine learning model meets the second preset requirement.

[0084] When the second prediction result does not meet the second preset requirement, the target parameters in the intermediate first machine learning model are optimized to obtain the first machine learning model.

[0085] In this embodiment, during the model training preparation stage, the preset machine learning model is configured with initial values ​​for the target parameters (i.e., initialization parameters), so that the preset machine learning model can be trained based on the acquired training set and initialization parameters to obtain the intermediate first machine learning model.

[0086] Then, a test set can be obtained separately (this test set can be obtained at the same time as the training set, or it can be obtained when model accuracy detection is required, this application does not limit this), and the second prediction result of the intermediate first machine learning model obtained by training can be detected, so as to determine whether the intermediate first machine learning model meets the prediction requirements based on the second prediction result and the pre-set second preset requirements.

[0087] When the judgment result is not met, that is, the second prediction result does not meet the second preset requirement, adaptive parameter optimization processing is performed on the target parameters in the intermediate first machine learning model to obtain the first machine learning model. For example, the second preset requirement can be: the first prediction error is less than a preset error threshold and the first prediction time is less than a preset time threshold. For example, the preset error threshold can be: 5*e -5 (Error rate threshold) The preset time threshold can be a reasonable time threshold such as 30 seconds or 1 minute.

[0088] For example, the specific process of adaptive parameter optimization of the target parameters in the intermediate first machine learning model can be detailed as follows:

[0089] An optimization model is established with the target parameters as decision variables and the root mean square error of the model prediction results as the optimization objective. The optimization model is then solved using a small step size traversal search method.

[0090] For example, multiple discrete data can be pre-set for each target parameter. After assigning the target parameter to each discrete data, the root mean square error of the prediction model is calculated accordingly. The target parameter corresponding to the smallest root mean square error is selected as the optimal target parameter of the prediction model, thereby obtaining the first machine learning model.

[0091] The following provides examples of the target parameters for adaptive parameter optimization, along with multiple discrete data points for each target parameter:

[0092]

[0093] Where n_estimators represents the total number of iterations, i.e. the number of decision trees; Gamma represents the minimum reduction in objective function (i.e. the minimum reduction in loss) required to further branch at the leaf nodes of the tree; learn_rate represents the learning rate of the model; Subsample represents the proportion of the subsamples used to train the model to the entire sample set; and max_depth represents the maximum depth of the tree model.

[0094] In this embodiment, the optimal target parameters that minimize the root mean square error of the model are obtained by a small step size search traversal method, thereby effectively ensuring the training accuracy and prediction speed of the first machine learning model.

[0095] In one possible implementation, after detecting whether the second prediction result of the intermediate first machine learning model meets the second preset requirement, the method further includes:

[0096] When the second prediction result meets the second preset requirement, the intermediate first machine learning model is used as the first machine learning model, and the steps of "when the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement" and subsequent steps are re-executed until the target distribution network voltage prediction model is obtained.

[0097] In this embodiment, when the second prediction result meets the second preset requirement, it indicates that the initialization parameters are the optimal target parameters. At this point, the intermediate first machine learning model can be used as the first machine learning model. Optionally, new historical environmental data, historical power grid data, and historical voltage data can be collected to form a test set. Based on this test set, the first machine learning model is evaluated and tested to verify whether it can be used as the final target distribution network voltage prediction model. For example, the evaluation and testing of the model may include: detecting whether the quality of the sample data in the training set used to train the first machine learning model meets the first preset requirement, and detecting whether the first prediction result of the first machine learning model meets the second requirement. In addition, the steps "when the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement" and subsequent steps are re-executed until the target distribution network voltage prediction model is obtained.

[0098] Step 103: When the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, the training set is subjected to data correction and data recombination based on the probability density function of the training set to obtain a recombined training set.

[0099] In step 103, if the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, it indicates that the data used for model training is insufficient to support the training of a model that meets the requirements, or that the prediction quality of the model does not yet meet the requirements of the final model prediction quality. In this case, data correction and data recombination are performed on the training set based on the probability density function of the training set to obtain a recombined training set.

[0100] Optionally, sample data quality can include multiple evaluation metrics such as sample data volume and sample data precision. First, it can be checked whether the sample data quality of the training set meets the first preset requirement regarding sample data volume / precision. If the first preset requirement is not met, it indicates that the amount or precision of data used for model training is insufficient to support training a model that meets the requirements. In this case, data correction and data recombination are performed on the training set based on its probability density function to obtain a reconstructed training set.

[0101] When the first preset requirement is met, the system further checks whether the first prediction result of the first machine learning model meets the second preset requirement. For example, the second preset requirement may include multiple evaluation metrics such as error rate and prediction time. If the first prediction result of the first machine learning model does not meet the error rate requirement and / or the prediction time requirement, it indicates that the prediction quality of the model at this point does not yet meet the requirements of the final model prediction quality. In this case, data correction and data reorganization are performed on the training set based on the probability density function of the training set to obtain a reorganized training set.

[0102] Optionally, the method for detecting whether the first prediction result of the first machine learning model meets the second preset requirement can be detailed as follows:

[0103] A test set is constructed by acquiring new historical environmental data, new historical power grid data, and new historical voltage data corresponding to the new historical environmental data and new historical power grid data for each node in the distribution network. The new historical environmental data and historical power grid data from the test set are then input into a first machine learning model, which outputs the predicted voltage of the corresponding node in the distribution network. Based on the accuracy matching between the predicted voltage of each node and the new historical voltage data of the corresponding node, it is determined whether the first prediction result of the first machine learning model meets the second preset requirement.

[0104] In this embodiment, by performing data correction and data recombination on the training set, a more uniform and accurate data set can be obtained, which is beneficial for subsequent training of a more accurate distribution network voltage prediction model based on the recombined training set, thereby realizing the prediction of distribution network voltage.

[0105] In one possible implementation, the training set is modified and reorganized based on its probability density function to obtain a reorganized training set, including:

[0106] The optimal probability density function for the training set is determined based on the training set and the preset kernel function.

[0107] The corrected set is obtained by extracting data from the training set based on the optimal probability density function.

[0108] The revised set and the training set are recombined and randomly mixed to obtain a recombined training set.

[0109] In this embodiment, for example, the preset kernel function can be: Gaussian kernel function, cosine kernel function, Yepanechnikov kernel function, exponential kernel function, linear kernel function, Tophat kernel function, etc., and this application does not limit it. The optimal probability density function of the training set is determined based on the training set and the preset kernel function, and then data is extracted from the training set using the optimal probability density function to obtain the corrected set.

[0110] Optionally, the optimal probability density function can be used to extract data from the training set to obtain the corrected set, which can be described in detail as follows:

[0111] For each node in the distribution network, data is extracted from the corresponding data set in the training set using the optimal probability density function. For example, based on the dimensions of irradiance, temperature, wind speed, node load, renewable energy installed capacity, and geographical distribution, data extraction operations are performed on the irradiance, temperature, wind speed, node load, renewable energy installed capacity, and geographical distribution data in the node's data set. The extracted discrete data from each dimension are then combined into a new data set for that node, and at least one new data set constitutes the correction set.

[0112] After obtaining the corrected set, the data sets in the corrected set and the data sets in the training set are recombined and randomly mixed to obtain the recombined training set.

[0113] In this embodiment, by performing data correction and data recombination on the original data (i.e., the training set), more data sets that can participate in model training are obtained, which is more conducive to training a distribution network voltage prediction model with higher accuracy and faster prediction speed based on these data sets.

[0114] In one possible implementation, the optimal probability density function of the training set is determined based on the training set and a preset kernel function, including:

[0115] The probability density function corresponding to each preset kernel function is determined based on the training set and at least one preset kernel function.

[0116] Based on the optimization function, the optimal smoothing parameter is determined from the smoothing parameters corresponding to at least one preset kernel function, which makes the function value of the probability density function optimal.

[0117] The optimal probability density function of the training set is determined based on the optimal smoothing parameter and the preset kernel function corresponding to the optimal smoothing parameter.

[0118] In this embodiment, for example, multiple sets of six-dimensional space samples are made to follow an unknown probability distribution. Given a preset kernel function, the probability density function of the training set can be determined. However, the choice of the preset kernel function can significantly affect the accuracy of the estimated probability density function.

[0119] Therefore, for at least one preset kernel function, the probability density function corresponding to that kernel function can be determined based on the training set and each preset kernel function. Then, based on the optimization function, the optimal smoothing parameter is determined from the smoothing parameters corresponding to each of the at least one kernel function, which makes the function value of the probability density function optimal (i.e., the highest accuracy).

[0120] After determining the optimal smoothing parameter, the optimal probability density function of the training set is determined based on the optimal smoothing parameter and the preset kernel function corresponding to the optimal smoothing parameter.

[0121] In this embodiment, using a randomly determined preset kernel function and its smoothing parameters may result in an estimated probability density function that is difficult to approximate the original probability distribution of the training set, potentially leading to inaccuracies in the data extraction results. Therefore, in this embodiment, by determining the optimal smoothing parameters of the preset kernel function, the optimal probability density function is further determined, thereby effectively ensuring the accuracy of the data extraction operation and the accuracy of the training set data during model training.

[0122] Optionally, the optimal smoothing parameters for each kernel function can be determined based on the GridSearchCV function, thereby deriving the optimal probability density function. The GridSearchCV function can be divided into two parts: GridSearch and CV, namely grid search and cross-validation. Grid search searches for parameters; that is, within a specified parameter range, the parameters are adjusted sequentially by step size, and the learner is trained using the adjusted parameters to find the parameters with the highest accuracy on the validation set from all parameters.

[0123] The GridSearchCV function is characterized by its high accuracy. Its required parameters are as follows: `Classsklearn.model_selection.GridSearchCV(estimator, param_grid, scoring = None, fit_params = None, n_jobs = None, iid = 'warn', refit = True, cv = 'warn', verbose = 0, pre_dispatch = '2*n_jobs', rror_score = 'raise-deprecating', return_train_score = 'warn')`.

[0124] In the program, `estimator` selects the classifier to use and passes in parameters other than those needed to determine the optimal ones.

[0125] param_grid: The values ​​of the parameters to be optimized. The values ​​can be a dictionary or a list.

[0126] scoring=None: The model evaluation criterion, which defaults to None. In this case, the score function needs to be used.

[0127] n_jobs = 1: n_jobs: number of parallel jobs, int: number of jobs, -1: same as the number of CPU cores, 1: default value.

[0128] iid=True: iid: defaults to True. When True, it means that the probability distribution of each sample fold is consistent, and the error estimate is the sum of all samples.

[0129] `refit=True`: This defaults to `True`. The program will retrain the model on all possible training and test sets using the best smoothing parameters obtained from the cross-validation training set, and use these as the final best smoothing parameters for performance evaluation. In other words, after searching for parameters, the entire dataset is fitted again using the best parameter result (i.e., the best smoothing parameters).

[0130] `cv=None`: Cross-validation parameter, defaults to None, uses three-fold cross-validation. Specifies the number of folds, defaults to 3, can also be a generator that yields the training / test set.

[0131] verbose=0,scoring=None: verbose: log verbosity, int: verbosity, 0: no output of the training process, 1: output occasionally, >1: output for each sub-model.

[0132] `pre_dispatch = '2 * n_jobs'`: Specifies the total number of parallel tasks to be dispatched. When `n_jobs` is greater than 1, the data will be copied at each run point, which may lead to Out Of Memory (OOM). Setting the `pre_dispatch` parameter can pre-divide the total number of jobs, so that the data is copied at most `pre_dispatch` times.

[0133] In one possible implementation, the probability density function corresponding to each preset kernel function is determined based on the training set and at least one preset kernel function, including:

[0134] based on Determine the probability density function corresponding to each preset kernel function.

[0135] Among them, f h (x) represents the probability density function, and N represents the number of samples in the training set. Let h represent the preset kernel function, h represent the smoothing parameter of the preset kernel function, and x represent the random variable. r Let represent the r-th sample data, and d represent the sample dimension of the training set.

[0136] In this embodiment, for example, the training set includes multiple sets of six-dimensional spatial samples, and the specific sample dimensions include: irradiance, temperature, wind speed, node load, new energy installed capacity and geographical distribution data.

[0137] Therefore, the formula for calculating the probability density function is:

[0138]

[0139] Step 104: Use the recombined training set as the new training set, and re-execute the steps of obtaining the first machine learning model based on the training set and subsequent steps until the sample data quality meets the first preset requirement and the first prediction result meets the second preset requirement. Then, use the first machine learning model as the target distribution network voltage prediction model.

[0140] In step 104, the recombined training set can be used as a new training set, and the steps of "training the preset machine learning model based on the training set to obtain the first machine learning model" and subsequent steps can be repeated until the quality of the sample data in the training set meets the first preset requirement and the first prediction result meets the second preset requirement. Then, the first machine learning model is used as the target distribution network voltage prediction model. In this way, by continuously correcting and recombining input data that does not meet the conditions, a distribution network voltage prediction model with higher prediction accuracy and efficiency is obtained by retraining based on the new training set. This target distribution network voltage prediction model is then used to predict the voltage of the distribution network.

[0141] Figure 2 The flowchart for predicting distribution network voltage provided in this embodiment of the invention is as follows: Figure 2 As shown in this embodiment, the environment and power grid information configuration module acquires historical environmental data, historical power grid data, and historical voltage data (i.e., training set) of each node in the distribution network; then, the model building module based on the XGBoost algorithm trains the model based on this historical data; the XGBoost parameter adaptive correction module is used to adaptively correct the target parameters in the model; then, the training model evaluation module evaluates and tests the trained model (e.g., judging whether the sample data quality, the error rate of the model prediction, and the prediction time meet the corresponding requirements); when the evaluation and testing results do not meet the corresponding conditions, the original data set correction preparation module constructs a probability density function based on the training set and at least one preset kernel function; then, the optimal smoothing parameter determination module determines the optimal smoothing parameter corresponding to a certain kernel function that maximizes the accuracy of the probability density function; after determining the optimal smoothing parameter, the optimal probability density function is further determined; then, the data recombination module corrects and recombines the data based on the optimal probability density function to form a new training set; then, the XGBoost prediction model iteration module iteratively trains the distribution network voltage prediction model based on the new training set until a model that meets the conditions is obtained. When the evaluation and detection results of the training model evaluation module meet the corresponding conditions, and a model that meets the conditions is obtained, the optimal voltage sensing module can predict the voltage of the distribution network after a high proportion of new energy sources are connected to the distribution network based on the real-time collected power grid data and environmental data.

[0142] This invention provides a training method for a distribution network voltage prediction model. The method involves acquiring a training set comprised of historical environmental data, historical power grid data, and corresponding historical voltage data for each node in the distribution network. A preset machine learning model is then trained based on this training set to obtain a first machine learning model. If the quality of the sample data in the training set does not meet the first preset requirement, or the first prediction result of the first machine learning model does not meet the second preset requirement, it indicates that the data used for model training is insufficient to support training a model that meets the requirements, or the prediction quality of the model does not yet meet the requirements for the final model prediction quality. Therefore, the training set is then corrected and reorganized based on its probability density function to obtain a reorganized training set. This reorganized training set is then used as a new training set, and the model training process is repeated until the quality of the sample data in the training set meets the first preset requirement and the first prediction result of the first machine learning model meets the second preset requirement. At this point, the first machine learning model is used as the target distribution network voltage prediction model, thereby enabling accurate voltage prediction for each node in the distribution network. This invention can train a high-precision distribution network voltage prediction model, which facilitates the rapid and accurate prediction of the voltage at each node in the distribution network, obtains the voltage change trend of the distribution network, and provides certain data support for the safe and stable operation of the distribution network.

[0143] Figure 3 The flowchart for the implementation of the distribution network voltage prediction method provided in this embodiment of the invention is as follows: Figure 3 As shown, this embodiment of the invention provides a method for predicting distribution network voltage, including:

[0144] Step 201: Obtain real-time environmental data and power grid data of each node in the distribution network, and record the environmental data and power grid data as real-time power grid data.

[0145] In step 201, environmental data and power grid data of each node in the distribution network are acquired through real-time monitoring. Optionally, the environmental data may include: irradiance data, temperature data, and wind speed data, etc. The power grid data may include: node load data, renewable energy installed capacity data, and geographical distribution data, etc. This application does not limit the specific types of data included in the environmental data and power grid data. The environmental data and power grid data are recorded as real-time power grid data and used as input data for the target distribution network voltage prediction model.

[0146] Step 202: Input the real-time power grid data of each node into the target distribution network voltage prediction model, and output the predicted voltage of the node corresponding to the real-time power grid data; wherein, the target distribution network voltage prediction model is trained based on the method of the first aspect or any possible implementation of the first aspect as described above.

[0147] In step 202, the real-time power grid data of each node is input into the target distribution network voltage prediction model, and the predicted voltage of the corresponding node is output, thereby realizing accurate prediction of the distribution network voltage after a high proportion of new energy sources are connected to the distribution network.

[0148] This invention provides a method for predicting distribution network voltage. By inputting real-time grid data of each node in the distribution network into the target distribution network voltage prediction model, the predicted voltage of the corresponding node is output. This enables accurate prediction of distribution network voltage after a high proportion of new energy sources are connected to the distribution network, obtains the trend of distribution network voltage changes, and provides certain data support for the safe and stable operation of the distribution network.

[0149] Figure 4 This is an example diagram of the power distribution network topology provided in an embodiment of the present invention. Figure 5 Please refer to the spatiotemporal distribution diagram of the predicted voltage of the distribution network provided in the embodiments of the present invention. Figure 4 and Figure 5 The embodiments of the present invention provide as follows Figure 4 The example distribution network shown, containing 33 generating nodes, can be used to predict the voltage of each node in the distribution network based on the target distribution network voltage prediction model, thereby obtaining the following... Figure 5 The diagram shows the spatiotemporal distribution of the predicted voltage at each node in the distribution network, thereby enabling accurate prediction of the voltage at each node in the distribution network.

[0150] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0151] The following are device embodiments of the present invention. For details not described in detail, please refer to the corresponding method embodiments described above.

[0152] Figure 6 This is a schematic diagram of the structure of the training device for the distribution network voltage prediction model provided in an embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below:

[0153] like Figure 6 As shown, the training device 1 for the distribution network voltage prediction model includes:

[0154] The historical data acquisition module 11 is used to acquire historical environmental data, historical power grid data, and training sets consisting of historical voltage data corresponding to historical environmental data and historical power grid data of each node in the distribution network.

[0155] The first model training module 12 is used to train a preset machine learning model based on the training set to obtain the first machine learning model.

[0156] The model accuracy verification module 13 is used to perform data correction and data recombination processing on the training set based on the probability density function of the training set when the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, so as to obtain a recombined training set.

[0157] The target model acquisition module 14 is used to reassemble the training set as a new training set, re-execute the steps of obtaining the first machine learning model based on the training set and subsequent steps until the sample data quality meets the first preset requirement and the first prediction result meets the second preset requirement, and then use the first machine learning model as the target distribution network voltage prediction model.

[0158] This invention provides a training device for a distribution network voltage prediction model, comprising: a historical data acquisition module 11, a first model training module 12, a model accuracy verification module 13, and a target model acquisition module 14. The device acquires a training set consisting of historical environmental data, historical power grid data, and historical voltage data corresponding to the historical environmental data and historical power grid data for each node in the distribution network. Then, a preset machine learning model is trained based on the training set to obtain a first machine learning model. When the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, it indicates that the data used for model training is insufficient to support training a model that meets the requirements, or that the prediction quality of the model does not yet meet the requirements of the final model prediction quality. Therefore, the training set is then subjected to data correction and data recombination based on the probability density function of the training set to obtain a recombined training set. The recombined training set is then used as a new training set, and the model training process is repeated until the quality of the sample data in the training set meets the first preset requirement and the first prediction result of the first machine learning model meets the second preset requirement. At this point, the first machine learning model is used as the target distribution network voltage prediction model, thereby enabling accurate voltage prediction for each node in the distribution network based on this target distribution network voltage prediction model. This invention can train a high-precision distribution network voltage prediction model, which facilitates the rapid and accurate prediction of the voltage at each node in the distribution network, obtains the voltage change trend of the distribution network, and provides certain data support for the safe and stable operation of the distribution network.

[0159] In one possible implementation, the first model training module 12 is specifically used for:

[0160] The preset machine learning model is trained based on the training set and the initialization parameters of the preset machine learning model to obtain the intermediate first machine learning model.

[0161] Check whether the second prediction result of the first intermediate machine learning model meets the second preset requirement.

[0162] When the second prediction result does not meet the second preset requirement, the target parameters in the intermediate first machine learning model are optimized to obtain the first machine learning model.

[0163] In one possible implementation, the first model training module 12 is further specifically used for:

[0164] When the second prediction result meets the second preset requirement, the intermediate first machine learning model is used as the first machine learning model, and the steps of "when the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement" and subsequent steps are re-executed until the target distribution network voltage prediction model is obtained.

[0165] In one possible implementation, the model accuracy verification module 13 is specifically used for:

[0166] The optimal probability density function for the training set is determined based on the training set and the preset kernel function.

[0167] The corrected set is obtained by extracting data from the training set based on the optimal probability density function.

[0168] The revised set and the training set are recombined and randomly mixed to obtain a recombined training set.

[0169] In one possible implementation, the model accuracy verification module 13 is further specifically used for:

[0170] The probability density function corresponding to each preset kernel function is determined based on the training set and at least one preset kernel function.

[0171] Based on the optimization function, the optimal smoothing parameter is determined from the smoothing parameters corresponding to at least one preset kernel function, which makes the function value of the probability density function optimal.

[0172] The optimal probability density function of the training set is determined based on the optimal smoothing parameter and the preset kernel function corresponding to the optimal smoothing parameter.

[0173] In one possible implementation, the model accuracy verification module 13 is further specifically used for:

[0174] based on Determine the probability density function corresponding to each preset kernel function.

[0175] Among them, f h (x) represents the probability density function, and N represents the number of samples in the training set. Let h represent the preset kernel function, h represent the smoothing parameter of the preset kernel function, and x represent the random variable. r Let represent the r-th sample data, and d represent the sample dimension of the training set.

[0176] Figure 7 The schematic diagram shows the structure of the distribution network voltage prediction device provided in the embodiment of the present invention. For ease of explanation, only the parts related to the embodiment of the present invention are shown, and are described in detail below:

[0177] like Figure 7 As shown, the voltage prediction device 2 for the distribution network includes:

[0178] The data acquisition module 21 acquires real-time environmental data and power grid data of each node in the distribution network, and records the environmental data and power grid data as real-time power grid data.

[0179] The voltage prediction module 22 inputs the real-time power grid data of each node into the target distribution network voltage prediction model and outputs the predicted voltage of the node corresponding to the real-time power grid data; wherein, the target distribution network voltage prediction model is trained based on the method of the first aspect or any possible implementation of the first aspect.

[0180] This invention provides a voltage prediction device for a distribution network, comprising a data acquisition module 21 and a voltage prediction module 22. By inputting real-time grid data of each node in the distribution network into a target distribution network voltage prediction model, the predicted voltage of the corresponding node is output, thereby achieving accurate prediction of the distribution network voltage after a high proportion of new energy sources are connected to the distribution network, obtaining the voltage change trend of the distribution network, and providing certain data support for the safe and stable operation of the distribution network.

[0181] Figure 8 This is a schematic diagram of a training device provided in an embodiment of the present invention. Figure 8 As shown, the training device 3 in this embodiment includes: a processor 30, a memory 31, and a computer program 32 stored in the memory 31 and executable on the processor 30. When the processor 30 executes the computer program 32, it implements the steps in the training method embodiments of the various distribution network voltage prediction models described above, for example... Figure 1 Steps 101 to 104 are shown. Alternatively, when the processor 30 executes the computer program 32, it implements the functions of each module in the above-described device embodiments, for example... Figure 6 The functions of modules 11 to 14 are shown.

[0182] For example, the computer program 32 can be divided into one or more modules / units, which are stored in the memory 31 and executed by the processor 30 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 32 in the training device 3. For example, the computer program 32 can be divided into... Figure 6 Modules 11 to 14 are shown.

[0183] The training device 3 can be a desktop computer, laptop, handheld computer, or cloud server, etc. The training device 3 may include, but is not limited to, a processor 30 and a memory 31. Those skilled in the art will understand that... Figure 8 This is merely an example of training device 3 and does not constitute a limitation on training device 3. It may include more or fewer components than shown, or combine certain components, or different components. For example, the training device may also include input / output devices, network access devices, buses, etc.

[0184] The processor 30 may 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. A general-purpose processor may be a microprocessor or any conventional processor.

[0185] The memory 31 can be an internal storage unit of the training device 3, such as a hard disk or memory of the training device 3. The memory 31 can also be an external storage device of the training device 3, such as a plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, etc., equipped on the training device 3. Furthermore, the memory 31 can include both internal and external storage units of the training device 3. The memory 31 is used to store the computer program and other programs and data required by the training device. The memory 31 can also be used to temporarily store data that has been output or will be output.

[0186] Figure 9 This is a schematic diagram of a prediction device provided in an embodiment of the present invention. Figure 9 As shown, the prediction device 4 in this embodiment includes: a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, it implements the steps in the above-described embodiments of the power distribution network voltage prediction methods, for example... Figure 3 Steps 201 to 202 are shown. Alternatively, when the processor 40 executes the computer program 42, it implements the functions of each module in the above-described device embodiments, for example... Figure 7 The functions of modules 21 and 22 shown.

[0187] For example, the computer program 42 can be divided into one or more modules / units, which are stored in the memory 41 and executed by the processor 40 to complete the present invention. The one or more modules / units can be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program 42 in the prediction device 4. For example, the computer program 42 can be divided into... Figure 7 Modules 21 to 22 are shown.

[0188] The prediction device 4 can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. The prediction device 4 may include, but is not limited to, a processor 40 and a memory 41. Those skilled in the art will understand that... Figure 9 This is merely an example of prediction device 4 and does not constitute a limitation on prediction device 4. It may include more or fewer components than illustrated, or combine certain components, or different components. For example, the prediction device may also include input / output devices, network access devices, buses, etc.

[0189] The processor 40 may 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. A general-purpose processor may be a microprocessor or any conventional processor.

[0190] The memory 41 can be an internal storage unit of the prediction device 4, such as a hard disk or RAM of the prediction device 4. The memory 41 can also be an external storage device of the prediction device 4, such as a plug-in hard disk, Smart Media Card (SMC), Secure Digital (SD) card, or Flash Card equipped on the prediction device 4. Furthermore, the memory 41 can include both internal and external storage units of the prediction device 4. The memory 41 is used to store the computer program and other programs and data required by the prediction device. The memory 41 can also be used to temporarily store data that has been output or will be output.

[0191] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this application. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0192] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0193] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0194] In the embodiments provided by this invention, it should be understood that the disclosed apparatus / training / prediction device and method can be implemented in other ways. For example, the apparatus / training / prediction device embodiments described above are merely illustrative. For instance, the division of modules or units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between apparatuses or units may be electrical, mechanical, or other forms.

[0195] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0196] Furthermore, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0197] If the integrated module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of the training methods for the various distribution network voltage prediction models or the distribution network voltage prediction method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately added to or subtracted from the content as required by the legislation and patent practice in the jurisdiction. For example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium may not include electrical carrier signals and telecommunication signals.

[0198] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A training method for a distribution network voltage prediction model, characterized in that, include: Acquire a training set consisting of historical environmental data, historical power grid data, and historical voltage data corresponding to the historical environmental data and historical power grid data for each node in the distribution network; The preset machine learning model is trained based on the training set to obtain the first machine learning model; When the quality of the sample data in the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement, the training set is subjected to data correction and data recombination based on the probability density function of the training set to obtain a recombined training set. This includes: determining the optimal probability density function of the training set according to the training set and a preset kernel function; extracting data from the training set based on the optimal probability density function to obtain a corrected set; and recombinating and randomly mixing the corrected set and the training set to obtain a recombined training set. The reorganized training set is used as a new training set, and the steps of obtaining the first machine learning model based on the training set and subsequent steps are re-executed until the sample data quality meets the first preset requirement and the first prediction result meets the second preset requirement. Then, the first machine learning model is used as the target distribution network voltage prediction model. The step of determining the optimal probability density function of the training set based on the training set and the preset kernel function includes: determining the probability density function corresponding to each preset kernel function based on the training set and at least one preset kernel function; determining the optimal smoothing parameter among the smoothing parameters corresponding to the at least one preset kernel function based on the optimization function, which makes the function value of the probability density function optimal; and determining the optimal probability density function of the training set based on the optimal smoothing parameter and the preset kernel function corresponding to the optimal smoothing parameter. The step of extracting data from the training set based on the optimal probability density function to obtain the correction set includes: for each node in the distribution network, using the optimal probability density function to extract data from the data group corresponding to that node in the training set, including irradiance data, temperature data, wind speed data, node load data, new energy installed capacity data, and geographical distribution data, and combining the extracted discrete data of each dimension into a new data group for that node, and combining at least one new data group to form the correction set.

2. The training method for the distribution network voltage prediction model according to claim 1, characterized in that, The step of determining the probability density function corresponding to each preset kernel function based on the training set and at least one preset kernel function includes: based on Determine the probability density function corresponding to each preset kernel function; in, Denotes the probability density function, This represents the number of samples in the training set. Indicates the preset kernel function. This represents the smoothing parameter of the preset kernel function. Represents a random variable. Indicates the first One sample data, This indicates the sample dimension of the training set.

3. The training method for the distribution network voltage prediction model according to claim 1 or 2, characterized in that, The step of training a preset machine learning model based on the training set to obtain a first machine learning model includes: The preset machine learning model is trained based on the training set and the initialization parameters of the preset machine learning model to obtain an intermediate first machine learning model; Detect whether the second prediction result of the intermediate first machine learning model meets the second preset requirement; When the second prediction result does not meet the second preset requirement, the target parameters in the intermediate first machine learning model are optimized to obtain the first machine learning model.

4. The training method for the distribution network voltage prediction model according to claim 3, characterized in that, After detecting whether the second prediction result of the intermediate first machine learning model meets the second preset requirement, the method further includes: When the second prediction result meets the second preset requirement, the intermediate first machine learning model is used as the first machine learning model, and the steps "when the quality of the sample data of the training set does not meet the first preset requirement or the first prediction result of the first machine learning model does not meet the second preset requirement" and subsequent steps are re-executed until the target distribution network voltage prediction model is obtained.

5. A method for predicting voltage in a distribution network, characterized in that, include: Acquire real-time environmental data and power grid data of each node in the distribution network, and record the environmental data and power grid data as real-time power grid data; The real-time power grid data of each node is input into the target distribution network voltage prediction model, and the predicted voltage of the node corresponding to the real-time power grid data is output; wherein, the target distribution network voltage prediction model is trained based on the training method of the distribution network voltage prediction model according to any one of claims 1 to 4.

6. A training device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in any one of claims 1 to 4 above.

7. A prediction device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method as described in claim 5 above.

8. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1 to 5 above.

Images