Low-voltage transformer area closed-loop voltage control method and system

Abstract

The invention discloses a low-voltage transformer area closed-loop voltage control method and system, and relates to the technical field of low-voltage transformer area self-healing, and the method comprises the steps: obtaining a plurality of training closed-loop voltage data, carrying out the data preprocessing of each training closed-loop voltage data, and obtaining a corresponding voltage feature set; and training a preset initial voltage dynamic prediction model by using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, obtaining closed-loop voltage data of the low-voltage transformer area, predicting the closed-loop voltage data through the target voltage dynamic prediction model to obtain a plurality of voltage time sequence change values, and calculating the voltage time sequence change values of the low-voltage transformer area. And loop closing voltage control is carried out on the low-voltage transformer area based on each voltage time sequence change value. The technical problems that a static threshold judgment mode is mostly adopted in traditional loop closing voltage control, but the adaptability to scenes such as voltage fluctuation caused by new energy access and voltage distortion caused by nonlinear loads is extremely poor, and the operation reliability of a low-voltage transformer area is reduced are solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of low-voltage area self-healing, and in particular to a low-voltage area loop closing voltage control method and system. BACKGROUND

[0002] With the large-scale access of distributed photovoltaic, energy storage and other new energy equipment, and the rapid growth of new types of loads such as electric vehicle charging piles, the operation state of low-voltage areas is becoming more and more complex. The traditional management mode relying on manual inspection and manual operation has been difficult to meet the precise and efficient operation and maintenance needs. At present, the distribution network self-healing technology has become the core direction to improve the reliability of power supply. As a key means of fault rapid isolation and load transfer, the loop closing operation of the low-voltage area line can effectively reduce the fault outage time and range. However, the difference in voltage amplitude and phase angle during the loop closing process can easily cause line impact, equipment damage and other risks, which seriously restricts the large-scale application of loop closing technology. Therefore, how to control the loop closing voltage is crucial.

[0003] At present, the traditional loop closing voltage control mostly adopts a static threshold judgment method, that is, whether to perform loop closing operation is determined by pre-setting a voltage difference threshold (such as 3%-5% of the rated voltage) on both sides of the loop. However, this method can only make judgments based on instantaneous steady-state data and cannot capture the dynamic voltage change law from the loop closing moment to the line entering the new steady-state process. The adaptability of this method to scenarios such as voltage fluctuations caused by new energy access and voltage distortion caused by nonlinear loads is extremely poor, which reduces the reliability of the operation of low-voltage areas. SUMMARY

[0004] The present application provides a low-voltage area loop closing voltage control method and system, which solves the technical problem that the traditional loop closing voltage control mostly adopts a static threshold judgment method, which can only make judgments based on instantaneous steady-state data and cannot capture the dynamic voltage change law from the loop closing moment to the line entering the new steady-state process. The adaptability of this method to scenarios such as voltage fluctuations caused by new energy access and voltage distortion caused by nonlinear loads is extremely poor, which reduces the reliability of the operation of low-voltage areas.

[0005] The present application provides a low-voltage area loop closing voltage control method and system, which solves the technical problem that the traditional loop closing voltage control mostly adopts a static threshold judgment method, which can only make judgments based on instantaneous steady-state data and cannot capture the dynamic voltage change law from the loop closing moment to the line entering the new steady-state process. The adaptability of this method to scenarios such as voltage fluctuations caused by new energy access and voltage distortion caused by nonlinear loads is extremely poor, which reduces the reliability of the operation of low-voltage areas.

[0006] Obtaining a plurality of training loop closing voltage data, and performing data preprocessing on each of the training loop closing voltage data to obtain a corresponding voltage feature set;

[0007] Training the initial voltage dynamic prediction model pre-set by using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises a feature enhancement network and a prediction network;

[0008] Obtaining loop voltage data of a low-voltage transformer area, performing feature enhancement processing on the loop voltage data by using the feature enhancement network to obtain a corresponding enhanced feature vector;

[0009] Performing loop voltage prediction on the enhanced feature vector by using the prediction network to obtain a plurality of voltage time sequence change values;

[0010] Controlling the loop voltage of the low-voltage transformer area based on each voltage time sequence change value.

[0011] Optionally, the step of training a preset initial voltage dynamic prediction model by using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, comprising:

[0012] Inputting the voltage feature set into the preset initial voltage dynamic prediction model for training to obtain corresponding training voltage prediction data;

[0013] Based on a preset loss function, calculating a loss function value of the voltage feature set according to the training voltage prediction data;

[0014] When the loss function value is greater than or equal to a preset loss threshold value, then adjusting the network parameters of the initial voltage dynamic prediction model until the loss function value is less than the loss threshold value;

[0015] When the loss function value is less than the loss threshold value, then generating a target voltage dynamic prediction model.

[0016] Optionally, the feature enhancement network comprises a multi-scale time window module, a physical rule embedding module and a feature attention fusion module, and the step of performing feature enhancement processing on the loop voltage data by using the feature enhancement network to obtain a corresponding enhanced feature vector, comprising:

[0017] Constructing a voltage matrix by using the loop voltage data;

[0018] Performing voltage fluctuation feature extraction on the voltage matrix by using the multi-scale time window module to obtain a plurality of voltage fluctuation feature vectors, wherein the multi-scale time window module comprises a first convolutional layer, a second convolutional layer and a third convolutional layer;

[0019] Performing electrical feature extraction on the voltage matrix by using the physical rule embedding module to obtain a corresponding electrical feature vector, wherein the physical rule embedding module comprises a voltage stability coefficient model, a first fully connected layer and a second fully connected layer connected in sequence;

[0020] The feature attention fusion module is used for attention weighted fusion of each voltage fluctuation feature vector and the electrical feature vector, to obtain a corresponding enhanced feature vector.

[0021] Optionally, the step of extracting electrical features from the voltage matrix by the physical rule embedding module to obtain a corresponding electrical feature vector comprises:

[0022] The voltage matrix is input into the voltage stability coefficient model to obtain a corresponding voltage stability coefficient;

[0023] The voltage stability coefficient is mapped by a first full connection layer and a second full connection layer connected in sequence to obtain a corresponding electrical feature vector.

[0024] Optionally, the prediction network comprises a bidirectional long short-term memory module, a sparse Transformer module, a time-space attention fusion module, and a fourth full connection layer, and the step of predicting the closed-loop voltage of the enhanced feature vector by the prediction network to obtain a plurality of voltage time series change values comprises:

[0025] The enhanced feature vector is short-time feature extracted by the bidirectional long short-term memory module to obtain a corresponding short-time feature vector, wherein the bidirectional long short-term memory module comprises two bidirectional long short-term memory networks and two normalization layers;

[0026] The enhanced feature vector is long-time feature extracted by the sparse Transformer module to obtain a corresponding long-time feature vector;

[0027] The short-time feature vector and the long-time feature vector are attention feature fused by the time-space attention fusion module to obtain a corresponding time-space fusion feature vector;

[0028] The time-space fusion feature vector is mapped by a fourth full connection layer to obtain a plurality of voltage time series change values.

[0029] Optionally, the step of controlling the low-voltage area based on each voltage time series change value comprises:

[0030] It is judged whether each voltage time series change value is in a preset first voltage threshold interval;

[0031] When any voltage time series change value is not in the first voltage threshold interval, it is judged whether each voltage time series change value is in a preset second voltage threshold interval;

[0032] When any of the voltage time series change values is not in the second voltage threshold interval, then the low-voltage area is closed loop operation;

[0033] When each of the voltage time series change values is in the second voltage threshold interval, then the low-voltage area is regulated based on a preset cooperative control strategy, and the step of obtaining the closed loop voltage data of the low-voltage area is executed.

[0034] When each of the voltage time series change values is in the first voltage threshold interval, then the low-voltage area is closed loop operation.

[0035] The second aspect of the present application provides a low-voltage area closed loop voltage control system, comprising:

[0036] A preprocessing module is configured to obtain a plurality of training closed loop voltage data, and perform data preprocessing on each of the training closed loop voltage data to obtain a corresponding voltage feature set.

[0037] A training module is configured to train a preset initial voltage dynamic prediction model using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises a feature enhancement network and a prediction network.

[0038] An acquisition module is configured to obtain closed loop voltage data of a low-voltage area, and perform feature enhancement processing on the closed loop voltage data through the feature enhancement network to obtain a corresponding enhanced feature vector.

[0039] A prediction module is configured to perform closed loop voltage prediction on the enhanced feature vector through the prediction network to obtain a plurality of voltage time series change values.

[0040] A control module is configured to perform closed loop voltage control on the low-voltage area based on each of the voltage time series change values.

[0041] The third aspect of the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the low-voltage area closed loop voltage control method according to any one of the above aspects.

[0042] The fourth aspect of the present application provides a computer readable storage medium having a computer program stored thereon, wherein the computer program is executed to implement the low-voltage area closed loop voltage control method according to any one of the above aspects.

[0043] The fifth aspect of the present application provides a computer program product, the computer program product comprises a computer program stored on a non-transitory computer-readable storage medium, the computer program comprises program instructions, wherein when the program instructions are executed by a computer, the computer executes the low-voltage transformer area closing voltage control method according to any one of the above.

[0044] From the above technical solutions, the present application has the following advantages:

[0045] The present application obtains a plurality of training closing voltage data, pre-processes each training closing voltage data to obtain a corresponding voltage feature set, trains a preset initial voltage dynamic prediction model using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, obtains closing voltage data of a low-voltage transformer area, predicts the closing voltage data by the target voltage dynamic prediction model to obtain a plurality of voltage time sequence change values, and controls the closing voltage of the low-voltage transformer area based on each voltage time sequence change value. The present application overcomes the technical problem that the traditional closing voltage control method mainly uses a static threshold judgment method, but this method can only be based on instantaneous steady-state data for judgment and cannot capture the dynamic voltage change rule from the closing moment to the new steady-state process of the line, has poor adaptability to voltage fluctuations caused by new energy access and voltage distortion caused by nonlinear loads, and reduces the reliability of the operation of the low-voltage transformer area. Compared with the traditional closing voltage control method, the present application predicts the closing voltage data by the target voltage dynamic prediction model to obtain a plurality of voltage time sequence change values in the future time period, and then controls the closing voltage of the low-voltage transformer area according to each voltage time sequence change value, thereby avoiding the voltage out-of-limit risk in the closing transient process and improving the reliability of the operation of the low-voltage transformer area. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can also obtain other drawings according to these drawings without creative labor.

[0047] Figure 1 A step flow chart of a low-voltage transformer area closing voltage control method provided for the first embodiment of the present application;

[0048] Figure 2 A step flow chart of a low-voltage transformer area closing voltage control method provided for the second embodiment of the present application;

[0049] Figure 3 A structure diagram of a target voltage dynamic prediction model provided for the second embodiment of the present application;

[0050] Figure 4 A structural block diagram of a low-voltage area loop voltage control system provided for the third embodiment of the present application is shown in the figure.

[0051] Figure 5 A structural block diagram of an electronic device provided for the fourth embodiment of the present application is shown in the figure. DETAILED DESCRIPTION

[0052] The embodiments of the present application provide a low-voltage area loop voltage control method and system, which are used for solving the technical problem that the traditional loop voltage control is mostly based on static threshold judgment, but this method can only make judgment based on instantaneous steady-state data, cannot capture the dynamic voltage change law from the loop moment to the line entering the new steady-state process, has poor adaptability to the voltage fluctuation caused by new energy access and the voltage distortion caused by nonlinear load, and reduces the reliability of low-voltage area operation.

[0053] In order to make the purpose, technical scheme and advantages of the embodiments of the present application clearer, the technical scheme in the embodiments of the present application will be described clearly and completely below in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the protection scope of the present application. It should be noted that in the optional embodiments of the present application, the object information and other related data involved need to be authorized or agreed by the object when the embodiments of the present application are applied to specific products or technologies, and the collection, use and processing of the related data need to comply with the relevant laws, regulations and standards of the country and region. That is to say, if the embodiments of the present application involve data related to the object, the data needs to be obtained under the condition of authorization and agreement of the object, authorization and agreement of the relevant department, and compliance with the relevant laws, regulations and standards of the country and region. If the embodiments involve personal information, the consent of the individual needs to be obtained for the acquisition of all personal information, and the separate consent of the information subject needs to be obtained for the acquisition of sensitive information, and the embodiments also need to be implemented under the condition of authorization and agreement of the object.

[0054] Please refer to Figure 1 , Figure 1 A step flowchart of a low-voltage area loop voltage control method provided for the first embodiment of the present application is shown in the figure.

[0055] The low-voltage area loop voltage control method provided by the present application comprises the following steps.

[0056] In step 101, a plurality of training loop voltage data are acquired, and each training loop voltage data is preprocessed to obtain corresponding voltage feature sets.

[0057] The training closing-in voltage data refers to sample data in different operation scenarios of a low-voltage transformer substation, including voltage time series data of key monitoring points such as both sides of a closing-in point, a low-voltage bus, and a line end under normal power supply state, fault simulation state, different load levels (light load, heavy load, full load), and distributed photovoltaic output fluctuation state.

[0058] The voltage feature set refers to a set of training closing-in voltage data obtained after data preprocessing.

[0059] In the embodiment of the application, a plurality of training closing-in voltage data are acquired, and each training closing-in voltage data is subjected to data preprocessing (cleaning and standardization operation on the training closing-in voltage data) to obtain a corresponding voltage feature set.

[0060] Step 102, the voltage feature set is used to train a preset initial voltage dynamic prediction model to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises a feature enhancement network and a prediction network.

[0061] In the embodiment of the application, based on a preset loss function, the voltage feature set is used to train the preset initial voltage dynamic prediction model to obtain the corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises the feature enhancement network and the prediction network.

[0062] It should be noted that the loss function is specifically:

[0063]

[0064] wherein, is a loss function value, is a total number of monitoring points, is a closing-in post-prediction time step, is a voltage prediction value of an i-th monitoring point, a t-th time step, and a p-th phase, is a voltage actual value of the i-th monitoring point, the t-th time step, and the p-th phase, p is a voltage phase (A phase, B phase, and C phase), i is a monitoring point index, and t is a time index.

[0065] Step 103, closing-in voltage data of a low-voltage transformer substation are acquired, and the closing-in voltage data are subjected to feature enhancement processing by the feature enhancement network to obtain a corresponding enhanced feature vector.

[0066] The loop voltage data refers to voltage data of each monitoring point before the loop closing of the low-voltage transformer area, including three-phase voltage time sequence series (i.e. voltage amplitude and phase angle data (time step is 1 second) of A, B and C three-phase within 1-2 hours before the loop closing) of both sides of the loop closing point, three-phase voltage difference series (i.e. the corresponding phase voltage of both sides of the loop closing point is subtracted through the same time step) of both sides of the loop closing point, voltage time sequence data (including synchronous voltage amplitude data of at least three key bus monitoring points and two line end monitoring points) of the low-voltage bus and the line end.

[0067] In the embodiment of the application, the loop voltage data of the low-voltage transformer area is obtained, and the loop voltage data is processed by a feature enhancement network to obtain a corresponding enhanced feature vector, wherein the feature enhancement network includes a multi-scale time window module, a physical rule embedding module and a feature attention fusion module.

[0068] In step 104, the enhanced feature vector is processed by a prediction network to obtain a plurality of voltage time sequence change values.

[0069] The voltage time sequence change value refers to the output result of the prediction network, which refers to the predicted voltage of a specific monitoring point at a specific time step in a future time period after the loop operation.

[0070] In the embodiment of the application, the enhanced feature vector is processed by a prediction network to obtain a plurality of voltage time sequence change values, wherein the prediction network includes a bidirectional long short-term memory module, a sparse Transformer module, a time-space attention fusion module and a fourth full connection layer.

[0071] In step 105, the low-voltage transformer area is controlled based on each voltage time sequence change value.

[0072] In the embodiment of the application, when each voltage time sequence change value is in the first voltage threshold interval, the loop operation is performed on the low-voltage transformer area. When each voltage time sequence change value is in the second voltage threshold interval (i.e. any voltage time sequence change value is not in the first voltage threshold interval, but does not exceed the second voltage threshold interval (there is a voltage time sequence change value at the edge of the qualified range)), the low-voltage transformer area is regulated and controlled based on a preset cooperative control strategy, and step 103 is executed. When any voltage time sequence change value is not in the second voltage threshold interval, the loop closing operation is performed on the low-voltage transformer area.

[0073] In the embodiment of the present application, by acquiring a plurality of training closing voltage data, the data of each training closing voltage data is preprocessed to obtain a corresponding voltage feature set, the voltage feature set is used to train a preset initial voltage dynamic prediction model to obtain a corresponding target voltage dynamic prediction model, the closing voltage data of the low-voltage transformer area is acquired, the closing voltage data is predicted by the target voltage dynamic prediction model to obtain a plurality of voltage time sequence change values, and the low-voltage transformer area is controlled based on each voltage time sequence change value. Overcome the technical problem that the traditional closing voltage control mostly adopts a static threshold judgment method, but this method can only be based on instantaneous steady-state data for judgment, and cannot capture the dynamic voltage change law from the closing moment to the new steady-state process of the line, and has poor adaptability to the voltage fluctuation caused by the access of new energy and the voltage distortion caused by nonlinear load, thereby reducing the reliability of the operation of the low-voltage transformer area. Compared with the traditional closing voltage control method, the closing voltage data is predicted by the target voltage dynamic prediction model, a plurality of voltage time sequence change values in the future time period are obtained, and the closing voltage of the low-voltage transformer area is controlled according to each voltage time sequence change value, thereby avoiding the voltage out-of-limit risk in the closing transient process and improving the reliability of the operation of the low-voltage transformer area.

[0074] Please refer to Figure 2 , Figure 2 The step flow chart of a low-voltage transformer area closing voltage control method provided in the second embodiment of the present application.

[0075] The low-voltage transformer area closing voltage control method provided by the present application comprises:

[0076] Step 201, a plurality of training closing voltage data is acquired, and each training closing voltage data is preprocessed to obtain a corresponding voltage feature set.

[0077] In the embodiment of the present application, a plurality of training closing voltage data is acquired, and each training closing voltage data is cleaned and standardized to obtain a corresponding voltage feature set.

[0078] Step 202, the voltage feature set is used to train a preset initial voltage dynamic prediction model to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises a feature enhancement network and a prediction network.

[0079] Further, step 202 comprises the following substeps:

[0080] S11, the voltage feature set is input into the preset initial voltage dynamic prediction model for training to obtain corresponding training voltage prediction data.

[0081] The training voltage prediction data refers to the voltage prediction sequence set output by the voltage feature set in the training process of the initial voltage dynamic prediction model.

[0082] In the embodiment of the present application, the initial voltage dynamic prediction model is trained by inputting the preset voltage feature set, and corresponding training voltage prediction data is obtained.

[0083] S12, based on the preset loss function, the loss function value of the voltage feature set is calculated according to the training voltage prediction data.

[0084] In the embodiment of the present application, the training voltage prediction data and the voltage feature set are input into the preset loss function, and the corresponding loss function value is obtained.

[0085] S13, when the loss function value is greater than or equal to the preset loss threshold, the network parameters of the initial voltage dynamic prediction model are adjusted until the loss function value is less than the loss threshold.

[0086] The loss threshold refers to the loss function critical value preset for judging whether the model training accuracy meets the standard, and the value is 0.0015.

[0087] The network parameters refer to the optimizable variables that constitute the initial voltage dynamic prediction model, including the convolution kernel weight of the convolution layer, the gating weight of the bidirectional LSTM, the weight matrix and the bias term of the full connection layer, etc.

[0088] In the embodiment of the present application, when the loss function value is greater than or equal to the preset loss threshold, the gradient descent method is used to adjust the network parameters of the initial voltage dynamic prediction model until the loss function value is less than the loss threshold.

[0089] S14, when the loss function value is less than the loss threshold, the target voltage dynamic prediction model is generated.

[0090] In the embodiment of the present application, when the loss function value is less than the loss threshold, the prediction accuracy of the initial voltage dynamic prediction model meets the requirements, and the target voltage dynamic prediction model is generated.

[0091] Step 203, obtain the closing loop voltage data of the low-voltage area, and perform feature enhancement processing on the closing loop voltage data through the feature enhancement network to obtain the corresponding enhanced feature vector.

[0092] Further, referring to Figure 3 As shown in the figure, the feature enhancement network includes a multi-scale time window module, a physical rule embedding module and a feature attention fusion module, and step 203 includes the following substeps:

[0093] S21, construct a voltage matrix by using the closing loop voltage data.

[0094] In the embodiment of the present application, the voltage matrix is constructed according to the closed-loop voltage data, for example, the voltage matrix is constructed based on the three-dimensional structure of "time step-monitoring point type-voltage parameter (i.e. amplitude, phase, voltage difference)".

[0095] S22, voltage fluctuation feature extraction is performed on the voltage matrix by a multi-scale time window module to obtain a plurality of voltage fluctuation feature vectors, wherein the multi-scale time window module comprises a first convolution layer, a second convolution layer and a third convolution layer.

[0096] It should be noted that the first convolution layer is a one-dimensional convolution layer (window size 5), the second convolution layer is a one-dimensional convolution layer (window size 10), and the third convolution layer is a one-dimensional convolution layer (window size 20).

[0097] In the embodiment of the present application, the voltage fluctuation feature extraction is performed on the voltage matrix by the first convolution layer, the second convolution layer and the third convolution layer respectively to obtain a plurality of voltage fluctuation feature vectors.

[0098] It is worth mentioning that the multi-scale time window module extracts the voltage fluctuation feature of the voltage matrix by setting three one-dimensional convolution layers with different windows, thereby obtaining voltage fluctuation feature vectors of different time scales (such as short-time voltage mutation and long-time voltage trend).

[0099] S23, electrical feature extraction is performed on the voltage matrix by a physical rule embedding module to obtain a corresponding electrical feature vector, wherein the physical rule embedding module comprises a voltage stability coefficient model, a first fully connected layer and a second fully connected layer connected in sequence.

[0100] Further, S23 comprises the following sub-steps:

[0101] S231, inputting the voltage matrix into the voltage stability coefficient model to obtain a corresponding voltage stability coefficient.

[0102] In the embodiment of the present application, the voltage stability coefficient of the voltage matrix is calculated based on the voltage stability coefficient model.

[0103] It should be noted that the voltage stability coefficient model is specifically:

[0104]

[0105] Wherein, S is the voltage stability coefficient, is the historical average voltage before closing the loop, is the rated voltage of the transformer area, is the maximum voltage difference before closing the loop.

[0106] S232, mapping the voltage stability coefficient by the first fully connected layer and the second fully connected layer connected in sequence to obtain a corresponding electrical feature vector.

[0107] It should be noted that the first fully connected layer is a fully connected layer embedded with a LeakyReLU activation function. The second fully connected layer is a fully connected layer embedded with a Linear activation function.

[0108] In the embodiment of the present application, the voltage stability coefficient is mapped to an 8-dimensional electrical feature vector through the first fully connected layer and the second fully connected layer connected in turn.

[0109] It is worth mentioning that the abstract physical concept of the voltage stability coefficient is quantified into a feature that can be input into the model through the physical rule embedding module, improving the sensitivity of the model to the closing loop risk.

[0110] S24, the feature attention fusion module is used to perform attention weighted fusion on each voltage fluctuation feature vector and electrical feature vector to obtain a corresponding enhanced feature vector, wherein the feature attention fusion module includes a self-attention layer, a feature fusion layer and a third fully connected layer connected in turn.

[0111] In the embodiment of the present application, the self-attention layer is used to perform weight distribution on each voltage fluctuation feature vector and electrical feature vector to obtain a corresponding first fusion weight (i.e., focusing on the features that have a significant impact on the closing loop voltage). Based on the first fusion weight, the feature fusion layer is used to perform weighted feature fusion on each voltage fluctuation feature vector and electrical feature vector to obtain a corresponding initial enhanced feature vector. Then, the third fully connected layer is used to perform dimension reduction processing (i.e., dimension unification) on the initial enhanced feature vector to obtain a corresponding enhanced feature vector.

[0112] It should be noted that the third fully connected layer is a fully connected layer embedded with a GELU activation function.

[0113] Step 204, the enhanced feature vector is subjected to closing loop voltage prediction through the prediction network to obtain a plurality of voltage time series change values.

[0114] Further, referring to Figure 3 , the prediction network includes a bidirectional long short-term memory module, a sparse Transformer module, a time-space attention fusion module and a fourth fully connected layer, and step 204 includes the following sub-steps:

[0115] S31, the enhanced feature vector is subjected to short-time feature extraction through the bidirectional long short-term memory module to obtain a corresponding short-time feature vector, wherein the bidirectional long short-term memory module includes two bidirectional long short-term memory networks and two normalization layers.

[0116] In the embodiment of the present application, the enhanced feature vector is sequentially connected by two bidirectional long short-term memory sub-modules for time sequence feature extraction to obtain a corresponding short-time feature vector, wherein the bidirectional long short-term memory sub-module comprises a bidirectional long short-term memory network and a normalization layer connected in sequence.

[0117] It is worth mentioning that the bidirectional long short-term memory module is used to capture the short-time dependence of the voltage time sequence data (such as the voltage fluctuation trend 10-30 seconds before the loop closing).

[0118] S32, long-time feature extraction is performed on the enhanced feature vector by the sparse Transformer module to obtain a corresponding long-time feature vector.

[0119] It should be noted that the sparse Transformer module comprises an encoding module (i.e. 3 layers of stacked encoders), a sparse multi-head attention layer (i.e. 4 attention heads, only paying attention to local windows with a time step interval ≤ T / 4 and key nodes with an interval = T / 2, T), a residual connection module, a normalization layer, a feedforward network (FFN) and a residual connection module connected in sequence.

[0120] In the embodiment of the present application, long-time feature extraction is performed on the enhanced feature vector by the sparse Transformer module (such as the cumulative influence of load changes on voltage 10-20 minutes before the loop closing), to obtain a corresponding long-time feature vector. The sparse Transformer module comprises an encoding module, a sparse multi-head attention layer, a residual connection module (i.e. using a feature fusion layer to fuse the input and output of the sparse multi-head attention layer), a normalization layer, a feedforward network and a residual connection module (i.e. using a feature fusion layer to fuse the output of the feedforward network and the input of the normalization layer) connected in sequence.

[0121] It is worth mentioning that the sparse Transformer module avoids redundant calculation through the sparse multi-head attention layer, thereby improving the real-time performance of the model.

[0122] S33, attention feature fusion is performed on the short-time feature vector and the long-time feature vector by the time-space attention fusion module to obtain a corresponding time-space fusion feature vector.

[0123] In the embodiment of the present application, the short-time feature vector and the long-time feature vector are fused by the time-space attention fusion module to obtain a corresponding time-space fusion feature vector. For example, the short-time feature vector and the long-time feature vector are added element by element by the time-space attention fusion module to obtain a preliminary fusion feature vector. Then, the spatial attention weight and the time attention weight of the preliminary fusion feature vector are calculated by two attention layers respectively. The corresponding time-space fusion feature vector is obtained by weighted fusion according to the spatial attention weight and the time attention weight.

[0124] It should be noted that the time-space fusion feature vector = A1*(spatial attention weight*preliminary fusion feature vector)+(1-A1)*(time attention weight*preliminary fusion feature vector), wherein A1 is a weight coefficient.

[0125] It is worth mentioning that the time-space attention fusion module dynamically filters the information critical to the closed-loop voltage prediction from the two dimensions of "spatial monitoring points" and "time steps", strengthens the effective features and suppresses the redundant interference, and finally outputs the high-precision voltage time series change value.

[0126] S34, mapping the time-space fusion feature vector by the fourth fully connected layer to obtain a plurality of voltage time series change values.

[0127] It should be noted that the fourth fully connected layer is a fully connected layer with a Tanh activation function.

[0128] In the embodiment of the present application, the time-space fusion feature vector is mapped by the fourth fully connected layer to obtain a plurality of voltage time series change values.

[0129] Step 205, determining whether each voltage time series change value is in a preset first voltage threshold interval.

[0130] The first voltage threshold interval refers to a preset optimal safety voltage range of closed-loop operation, which is set to ±3% of the rated voltage based on the low-voltage area rated voltage (single-phase 220V / three-phase 380V).

[0131] In the embodiment of the present application, it is determined whether each voltage time series change value is in the preset optimal safety voltage range of closed-loop operation.

[0132] Step 206, when any voltage time series change value is not in the first voltage threshold interval, it is determined whether each voltage time series change value is in a preset second voltage threshold interval.

[0133] The second voltage threshold interval refers to a preset acceptable safety boundary of voltage after closed-loop operation, and the amplitude range is ±5% of the rated voltage.

[0134] In the embodiment of the present application, when any voltage time sequence change value is not in the preset optimal safety voltage range of loop closing operation, it is determined whether each voltage time sequence change value is in the preset acceptable safety boundary of voltage after loop closing.

[0135] Step 207, when any voltage time sequence change value is not in the second voltage threshold interval, the low-voltage area is locked for loop closing operation.

[0136] In the embodiment of the present application, when any voltage time sequence change value is not in the acceptable safety boundary of voltage after loop closing, the low-voltage area loop closing action is prohibited, and a pre-warning signal is generated.

[0137] It should be noted that the lock loop closing operation means that the intelligent contact switch and related electrical elements are locked to block the execution of the loop closing instruction, so as to avoid accidents such as line impact and equipment damage caused by voltage out-of-limit.

[0138] Step 208, when each voltage time sequence change value is in the second voltage threshold interval, the low-voltage area is regulated based on the preset cooperative control strategy, and the step of acquiring the loop closing voltage data of the low-voltage area is executed.

[0139] In the embodiment of the present application, when each voltage time sequence change value is in the acceptable safety boundary of voltage after loop closing, the low-voltage area is regulated based on the preset cooperative control strategy, and after waiting for a preset regulation time (i.e. a preset time (30-60 seconds) for the action of the equipment to take effect), the step 204 is executed.

[0140] It should be noted that the cooperative control strategy means that the action of the distributed photovoltaic inverter, SVG and other controllable devices is coordinated to adjust the reactive power output or load distribution to optimize the voltage, so that the voltage is close to the first threshold interval.

[0141] Step 209, when each voltage time sequence change value is in the first voltage threshold interval, the low-voltage area is looped.

[0142] In the embodiment of the present application, when each voltage time sequence change value is in the optimal safety voltage range of loop closing operation, the low-voltage area is looped.

[0143] In the embodiment of the present application, by acquiring a plurality of training closing voltage data, the data of each training closing voltage data is preprocessed to obtain a corresponding voltage feature set, the voltage feature set is used to train a preset initial voltage dynamic prediction model to obtain a corresponding target voltage dynamic prediction model, the closing voltage data of the low-voltage transformer area is acquired, the closing voltage data is predicted by the target voltage dynamic prediction model to obtain a plurality of voltage time sequence change values, and the low-voltage transformer area is controlled based on each voltage time sequence change value. Overcome the technical problem that the traditional closing voltage control mostly adopts a static threshold judgment method, but this method can only be based on instantaneous steady-state data for judgment, and cannot capture the dynamic voltage change law from the closing moment to the new steady-state process of the line, and the adaptability is poor for the voltage fluctuation caused by the access of new energy and the voltage distortion caused by nonlinear load, and the reliability of the low-voltage transformer area operation is reduced. Compared with the traditional closing voltage control method, the closing voltage data is predicted by the target voltage dynamic prediction model, a plurality of voltage time sequence change values of the future time period are obtained, and the closing voltage control of the low-voltage transformer area is performed according to each voltage time sequence change value, thereby avoiding the voltage out-of-limit risk in the closing transient process and improving the reliability of the low-voltage transformer area operation.

[0144] Please refer to Figure 4 , Figure 4 The structure block diagram of a low-voltage transformer area closing voltage control system provided in the third embodiment of the present application.

[0145] The low-voltage transformer area closing voltage control system provided by the present application comprises:

[0146] The preprocessing module 301 is used for acquiring a plurality of training closing voltage data, preprocessing each training closing voltage data, and obtaining a corresponding voltage feature set.

[0147] The training module 302 is used for training a preset initial voltage dynamic prediction model by using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model comprises a feature enhancement network and a prediction network.

[0148] The acquisition module 303 is used for acquiring closing voltage data of a low-voltage transformer area, and performing feature enhancement processing on the closing voltage data by using the feature enhancement network to obtain a corresponding enhanced feature vector.

[0149] The prediction module 304 is used for predicting the closing voltage of the enhanced feature vector by using the prediction network to obtain a plurality of voltage time sequence change values.

[0150] The control module 305 is used for controlling the closing voltage of the low-voltage transformer area based on each voltage time sequence change value.

[0151] Further, the training module 302 comprises:

[0152] a training submodule configured to input the voltage feature set into a preset initial voltage dynamic prediction model to obtain corresponding training voltage prediction data;

[0153] a loss submodule configured to calculate a loss function value of the voltage feature set according to the training voltage prediction data based on a preset loss function;

[0154] an adjustment submodule configured to adjust network parameters of the initial voltage dynamic prediction model when the loss function value is greater than or equal to a preset loss threshold, until the loss function value is less than the loss threshold;

[0155] generate a target voltage dynamic prediction model when the loss function value is less than the loss threshold.

[0156] Further, the feature enhancement network comprises a multi-scale time window module, a physical rule embedding module and a feature attention fusion module, and the acquisition module 303 comprises:

[0157] a construction submodule configured to construct a voltage matrix using the closed-loop voltage data;

[0158] a fluctuation feature extraction submodule configured to extract voltage fluctuation features of the voltage matrix through the multi-scale time window module to obtain a plurality of voltage fluctuation feature vectors, wherein the multi-scale time window module comprises a first convolutional layer, a second convolutional layer and a third convolutional layer;

[0159] an electrical feature extraction submodule configured to extract electrical features of the voltage matrix through the physical rule embedding module to obtain a corresponding electrical feature vector, wherein the physical rule embedding module comprises a voltage stability coefficient model, a first fully connected layer and a second fully connected layer connected in sequence;

[0160] a weighted fusion submodule configured to perform attention weighted fusion on each voltage fluctuation feature vector and electrical feature vector through the feature attention fusion module to obtain a corresponding enhanced feature vector, wherein the feature attention fusion module comprises a self-attention layer, a feature fusion layer and a third fully connected layer connected in sequence.

[0161] Further, the electrical feature extraction submodule comprises:

[0162] a voltage stability analysis unit configured to input the voltage matrix into the voltage stability coefficient model to obtain a corresponding voltage stability coefficient;

[0163] a mapping unit configured to map the voltage stability coefficient through the first fully connected layer and the second fully connected layer connected in sequence to obtain a corresponding electrical feature vector.

[0164] Further, the prediction network comprises a bidirectional long short-term memory module, a sparse Transformer module, a time-space attention fusion module and a fourth full connection layer.

[0165] The short-time feature extraction submodule is configured to perform short-time feature extraction on the enhanced feature vector by the bidirectional long short-term memory module to obtain a corresponding short-time feature vector, wherein the bidirectional long short-term memory module comprises two bidirectional long short-term memory networks and two normalization layers.

[0166] The long-time feature extraction submodule is configured to perform long-time feature extraction on the enhanced feature vector by the sparse Transformer module to obtain a corresponding long-time feature vector.

[0167] The attention feature fusion submodule is configured to perform attention feature fusion on the short-time feature vector and the long-time feature vector by the time-space attention fusion module to obtain a corresponding time-space fusion feature vector.

[0168] The mapping submodule is configured to perform mapping operation on the time-space fusion feature vector by the fourth full connection layer to obtain a plurality of voltage time sequence change values.

[0169] Further, the control module 305 comprises:

[0170] The first analysis submodule is configured to determine whether each voltage time sequence change value is in a preset first voltage threshold interval.

[0171] The second analysis submodule is configured to, when any voltage time sequence change value is not in the first voltage threshold interval, determine whether each voltage time sequence change value is in a preset second voltage threshold interval.

[0172] When any voltage time sequence change value is not in the second voltage threshold interval, the low-voltage transformer area is subjected to a closed loop operation.

[0173] When each voltage time sequence change value is in the second voltage threshold interval, the low-voltage transformer area is subjected to regulation and control based on a preset cooperative control strategy, and the step of obtaining the closed loop voltage data of the low-voltage transformer area is executed.

[0174] When each voltage time sequence change value is in the first voltage threshold interval, the low-voltage transformer area is subjected to a closed loop operation.

[0175] Please refer to Figure 5 , Figure 5 A structural block diagram of an electronic device provided in Embodiment Four of the present application.

[0176] An electronic device according to an embodiment of the present application comprises a memory 401 and a processor 402, the memory 401 storing a computer program, and the computer program, when executed by the processor 402, causing the processor 402 to perform the low-voltage area closed-loop voltage control method according to any of the above-described embodiments.

[0177] The memory 401 can be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk, or a ROM. The memory 401 has a storage space 403 for program codes 413 for performing any of the method steps described above. For example, the storage space 403 for program codes can include individual program codes 413 for implementing respective steps in the above-described methods. These program codes can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards, or floppy disks. The program codes can be compressed in a suitable form, for example. These codes, when executed by a computing processing device, cause the computing processing device to perform the respective steps in the above-described methods. These program codes can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards, or floppy disks. The program codes can be compressed in a suitable form, for example. These codes, when executed by a computing processing device, cause the computing processing device to perform the respective steps in the above-described low-voltage area closed-loop voltage control method.

[0178] The fifth embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon, the computer program, when executed by a processor, implementing the low-voltage area closed-loop voltage control method according to any of the above-described embodiments.

[0179] The sixth embodiment of the present application also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, wherein the program instructions, when executed by a computer, cause the computer to perform the low-voltage area closed-loop voltage control method according to any of the above-described embodiments.

[0180] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described system, device, and unit can refer to the corresponding processes in the above-described method embodiments, which will not be described herein.

[0181] In several embodiments provided in the present application, it should be understood that the disclosed system, device and method can be implemented in other manners. For example, the division of the units is only a logical function division, and there can be another division manner for the actual implementation, for example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed mutual couplings or direct couplings or communication connections between different units, can be indirect couplings or communication connections through some interfaces, devices or units, and can be electrical, mechanical or in other forms.

[0182] The units described as separated components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place, or can be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purposes of the embodiments of the present application.

[0183] In addition, each functional unit in the various embodiments of the present application can be integrated in a processing unit, or each unit can exist physically as a separate unit, or two or more units can be integrated in one unit. The integrated unit can be implemented in the form of hardware, or in the form of a software functional unit.

[0184] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on such an understanding, the technical solutions of the present application essentially or substantially, or all or part of the technical solutions, can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods in the various embodiments of the present application. The foregoing storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), magnetic disk or optical disk, and various media that can store program codes.

[0185] The above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for controlling the closed-loop voltage of a low-voltage distribution area, characterized in that, include: Multiple training loop voltage data are acquired, and each training loop voltage data is preprocessed to obtain the corresponding voltage feature set. The voltage feature set is used to train the preset initial voltage dynamic prediction model to obtain the corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model includes a feature enhancement network and a prediction network. The loop voltage data of the low-voltage distribution area is obtained, and the loop voltage data is subjected to feature enhancement processing through the feature enhancement network to obtain the corresponding enhanced feature vector; The prediction network is used to predict the loop voltage of the enhanced feature vector to obtain multiple voltage time-series change values. The low-voltage distribution area is subjected to closed-loop voltage control based on the various voltage timing changes.

2. The low-voltage distribution area loop-closed voltage control method according to claim 1, characterized in that, The step of training a preset initial voltage dynamic prediction model using the voltage feature set to obtain a corresponding target voltage dynamic prediction model includes: The voltage feature set is input into a preset initial voltage dynamic prediction model for training to obtain the corresponding training voltage prediction data. Based on a preset loss function, the loss function value of the voltage feature set is calculated according to the training voltage prediction data; When the loss function value is greater than or equal to a preset loss threshold, the network parameters of the initial voltage dynamic prediction model are adjusted until the loss function value is less than the loss threshold. When the loss function value is less than the loss threshold, a dynamic prediction model for the target voltage is generated.

3. The low-voltage distribution area loop-closed voltage control method according to claim 1, characterized in that, The feature enhancement network includes a multi-scale time window module, a physical rule embedding module, and a feature attention fusion module. The step of performing feature enhancement processing on the closed-loop voltage data through the feature enhancement network to obtain the corresponding enhanced feature vector includes: A voltage matrix is ​​constructed using the closed-loop voltage data; The voltage fluctuation features of the voltage matrix are extracted by the multi-scale time window module to obtain multiple voltage fluctuation feature vectors. The multi-scale time window module includes a first convolutional layer, a second convolutional layer and a third convolutional layer. The voltage matrix is ​​subjected to electrical feature extraction by the physical rule embedding module to obtain the corresponding electrical feature vector. The physical rule embedding module includes a voltage stability coefficient model, a first fully connected layer and a second fully connected layer connected in sequence. The feature attention fusion module performs attention-weighted fusion on each voltage fluctuation feature vector and the electrical feature vector to obtain the corresponding enhanced feature vector. The feature attention fusion module includes a self-attention layer, a feature fusion layer and a third fully connected layer connected in sequence.

4. The low-voltage distribution area loop-closed voltage control method according to claim 3, characterized in that, The step of extracting electrical features from the voltage matrix through the physical rule embedding module to obtain the corresponding electrical feature vector includes: Input the voltage matrix into the voltage stability coefficient model to obtain the corresponding voltage stability coefficient; The voltage stability coefficient is mapped by sequentially connecting the first fully connected layer and the second fully connected layer to obtain the corresponding electrical feature vector.

5. The low-voltage distribution area loop-closed voltage control method according to claim 1, characterized in that, The prediction network includes a bidirectional long short-term memory module, a sparse Transformer module, a temporal-spatial attention fusion module, and a fourth fully connected layer. The step of using the prediction network to predict the loop voltage of the enhanced feature vector to obtain multiple voltage time-series change values ​​includes: The enhanced feature vector is extracted using the bidirectional long short-term memory module to obtain the corresponding short-term feature vector. The bidirectional long short-term memory module includes two bidirectional long short-term memory networks and two normalization layers. The sparse Transformer module is used to extract long-term features from the enhanced feature vector to obtain the corresponding long-term feature vector. The temporal-space attention fusion module performs attention feature fusion on the short-time feature vector and the long-time feature vector to obtain the corresponding temporal-space fused feature vector. The time-space fusion feature vector is mapped through the fourth fully connected layer to obtain multiple voltage time-series change values.

6. The low-voltage distribution area loop-closed voltage control method according to claim 1, characterized in that, The step of performing loop-closed voltage control on the low-voltage distribution area based on each of the voltage time-series change values ​​includes: Determine whether each of the voltage timing change values ​​is within a preset first voltage threshold range; If any of the voltage time-series change values ​​is not within the first voltage threshold range, then it is determined whether each of the voltage time-series change values ​​is within a preset second voltage threshold range; When any of the voltage timing change values ​​is not within the second voltage threshold range, a lockout loop-closing operation is performed on the low-voltage zone. When all the voltage timing change values ​​are within the second voltage threshold range, the low-voltage zone is regulated based on the preset collaborative control strategy, and the step of obtaining the loop voltage data of the low-voltage zone is executed. When all the voltage timing change values ​​are within the first voltage threshold range, a loop closing operation is performed on the low-voltage zone.

7. A low-voltage distribution area closed-loop voltage control system, characterized in that, include: The preprocessing module is used to acquire multiple training loop voltage data, perform data preprocessing on each training loop voltage data, and obtain the corresponding voltage feature set. The training module is used to train a preset initial voltage dynamic prediction model using the voltage feature set to obtain a corresponding target voltage dynamic prediction model, wherein the target voltage dynamic prediction model includes a feature enhancement network and a prediction network. The acquisition module is used to acquire the loop voltage data of the low-voltage distribution area, and to perform feature enhancement processing on the loop voltage data through the feature enhancement network to obtain the corresponding enhanced feature vector. The prediction module is used to predict the closed-loop voltage of the enhanced feature vector through the prediction network to obtain multiple voltage time-series change values; The control module is used to perform closed-loop voltage control on the low-voltage zone based on the various voltage timing change values.

8. An electronic device, characterized in that, The system includes a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the low-voltage zone loop-closed voltage control method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed, it implements the low-voltage zone loop-closed voltage control method as described in any one of claims 1-6.

10. A computer program product, characterized in that, The computer program product includes a computer program stored on a non-transitory computer-readable storage medium, the computer program including program instructions, wherein when the program instructions are executed by a computer, the computer performs the low-voltage zone loop-closed voltage control method as described in any one of claims 1-6.

Images