Power distribution network management method and device, computer equipment, readable storage medium and program product

Abstract

The invention relates to a power distribution network management method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the following steps: acquiring a plurality of candidate parameter sets corresponding to a power distribution network at a target moment; each candidate parameter set comprises candidate active power and candidate reactive power corresponding to a plurality of power grid nodes contained in the power distribution network; each candidate parameter set corresponds to the same power grid node; for each power grid node, determining a voltage amplitude corresponding to the power grid node based on the candidate active power and the candidate reactive power of the power grid node; for each candidate parameter set, determining fitness corresponding to the candidate parameter set under the condition that the voltage amplitude of each power grid node in the candidate parameter set meets a voltage constraint condition; and taking each candidate active power in the candidate parameter set of which the fitness meets the adaptation condition as a threshold value of a dynamic operation domain corresponding to each power grid node at the target moment. By adopting the method, the efficiency can be improved.

Description

Technical Field

[0001] The present application relates to the field of electric power technology, and in particular to a distribution network management method, apparatus, computer equipment, computer-readable storage medium, and computer program product. Background Art

[0002] With the development of technologies such as photovoltaics, wind power, batteries, and electric vehicles, these distributed resources are becoming increasingly popular in our daily lives and are being integrated into distribution networks on a large scale. However, the large-scale integration of distributed resources also brings new challenges to the operation of distribution networks. For example, if distributed resources export or import electricity at the same time, it may violate the voltage and thermal constraints in the network, causing voltage to exceed the limit, thus posing a safety hazard.

[0003] In traditional technology, the dynamic operation domain is usually calculated based on the optimal power flow calculation method, which means that distribution network operators need to obtain a complete and accurate distribution network power flow model. However, in practical applications, such power grid power flow models are often difficult to obtain because it is difficult to observe the circuit topology and structure of complex networks, and it is also difficult to measure circuit impedance, load and other parameters. Therefore, for power grid operators, if they want to calculate the dynamic operation domain, they first need a lot of time and resources to obtain the corresponding power flow model, and they also need to perform a lot of calculations and solutions on the power flow model, which is inefficient. Summary of the invention

[0004] Based on this, it is necessary to provide a distribution network management method, device, computer equipment, computer-readable storage medium and computer program product that can improve efficiency in response to the above technical problems.

[0005] In a first aspect, the present application provides a distribution network management method, comprising:

[0006] Acquire multiple candidate parameter sets corresponding to the distribution network at the target time; each of the candidate parameter sets includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes included in the distribution network; each of the candidate parameter sets corresponds to the same grid node;

[0007] For each of the grid nodes, determining a voltage amplitude corresponding to the grid node based on a candidate active power and a candidate reactive power of the grid node;

[0008] For each of the candidate parameter sets, when the voltage amplitudes of the respective power grid nodes in the candidate parameter set satisfy the voltage constraint condition, determining the fitness corresponding to the candidate parameter set;

[0009] Each candidate active power in the candidate parameter set whose fitness satisfies the adaptation condition is used as a threshold value of the dynamic operation domain corresponding to each of the grid nodes at the target moment.

[0010] In one of the embodiments, the voltage amplitude is determined based on the output of a neural network model; the input of the neural network model is the candidate active power and candidate reactive power of the grid node;

[0011] The process of establishing the neural network model includes:

[0012] Obtaining historical active power, historical reactive power, and historical voltage amplitude of the distribution network at a historical moment;

[0013] An initial model is obtained, and the historical active power, the historical reactive power, and the historical voltage amplitude are used as training data for the initial model. The initial model is trained to obtain a neural network model.

[0014] In one embodiment, the obtaining of the historical active power of the distribution network at a historical moment includes:

[0015] Obtaining the initial historical active power of the distribution network at a historical moment;

[0016] When an abnormality exists in the initial historical active power, power compensation is performed on the active power at the historical moment based on the historical active power at the previous historical moment and the historical active power at the next historical moment to obtain the historical active power of the distribution network at the historical moment.

[0017] In one of the embodiments, when the voltage amplitudes of the respective power grid nodes in the candidate parameter set satisfy the voltage constraint condition, determining the fitness corresponding to the candidate parameter set includes:

[0018] When the voltage amplitudes of the respective grid nodes in the candidate parameter set satisfy the voltage constraint conditions, performing power superposition on the respective active powers of the respective grid nodes to obtain the total power corresponding to the candidate parameter set;

[0019] Based on the total power, a fitness corresponding to the candidate parameter set is determined.

[0020] In one embodiment, the step of obtaining multiple candidate parameter sets corresponding to the power distribution network at the target time includes:

[0021] Obtain the current active power and current reactive power corresponding to each of multiple grid nodes in the distribution network at the target time;

[0022] Each of the current active powers and each of the current reactive powers are adjusted respectively to obtain a plurality of candidate parameter sets.

[0023] In one embodiment, the target time includes a future time; and the method further includes:

[0024] Acquire multiple prediction parameter sets corresponding to the distribution network at a future moment; each of the prediction parameter sets includes predicted active power and predicted reactive power corresponding to multiple grid nodes; and the multiple grid nodes corresponding to each of the prediction parameter sets are the same;

[0025] For each of the grid nodes, based on the predicted active power and predicted reactive power of the grid node, determining a predicted voltage amplitude corresponding to the grid node;

[0026] For each of the prediction parameter sets, when the voltage amplitudes of the respective power grid nodes in the prediction parameter set satisfy the voltage constraint condition, determining the prediction fitness corresponding to the prediction set;

[0027] Each predicted active power in the prediction parameter set whose prediction fitness satisfies the adaptation condition is used as the prediction threshold of the dynamic operation domain corresponding to each of the power grid nodes at the future moment.

[0028] In a second aspect, the present application also provides a distribution network management device, including:

[0029] A parameter set acquisition module, used to acquire multiple candidate parameter sets corresponding to the distribution network at the target time; each of the candidate parameter sets includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes included in the distribution network; each of the candidate parameter sets corresponds to the same grid node;

[0030] A voltage amplitude determination module, configured to determine, for each of the grid nodes, a voltage amplitude corresponding to the grid node based on the candidate active power and candidate reactive power of the grid node;

[0031] A fitness determination module, configured to determine, for each of the candidate parameter sets, the fitness corresponding to the candidate parameter set when the voltage amplitudes of the respective power grid nodes in the candidate parameter set all satisfy the voltage constraint condition;

[0032] The dynamic operation domain determination module is used to use each candidate active power in the candidate parameter set whose fitness meets the adaptation condition as the threshold of the dynamic operation domain corresponding to each of the grid nodes at the target moment.

[0033] In a third aspect, the present application further provides a computer device, wherein the computer device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of the above method are implemented.

[0034] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0035] In a fifth aspect, the present application further provides a computer program product, wherein the computer program product comprises a computer program, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0036] The above-mentioned distribution network management method, device, computer equipment, computer-readable storage medium and computer program product obtain multiple candidate parameter sets corresponding to the distribution network at the target time, wherein each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes contained in the distribution network, and each candidate parameter set corresponds to the same grid node. The operating parameter conditions under multiple operating conditions can be pre-determined. For each grid node, the voltage amplitude corresponding to the grid node is determined based on the candidate active power and candidate reactive power of the grid node, and the actual voltage operating condition corresponding to each grid node under the condition can be determined. For each candidate parameter set, when the voltage amplitudes of each grid node in the candidate parameter set meet the voltage constraint conditions, the fitness corresponding to the candidate parameter set is determined, and the allowable injection power of the grid under the condition can be determined. The candidate active power in the candidate parameter set whose fitness meets the adaptation condition is used as the threshold of the dynamic operating domain corresponding to each grid node at the target time. The active power corresponding to the condition with the minimum allowable injection power can be used as the threshold of the dynamic operating domain, thereby improving the efficiency of determining the dynamic operating domain. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the drawings required for use in the embodiments of the present application or related technical descriptions will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without paying creative work.

[0038] Figure 1 An application environment diagram of a distribution network management method in an embodiment;

[0039] Figure 2 A schematic diagram of a flow chart of a distribution network management method in one embodiment;

[0040] Figure 3 A schematic diagram of a flow chart of a distribution network management method in another embodiment;

[0041] Figure 4 is a structural block diagram of a distribution network management device in one embodiment;

[0042] Figure 5 FIG. 4 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0043] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0044] The distribution network management method provided in the embodiment of the present application can be applied to Figure 1 In the application environment shown. Among them, the terminal 102 communicates with the server 104 through the network. The data storage system can store the data that the server 104 needs to process. The data storage system can be integrated on the server 104, or it can be placed on the cloud or other network servers. Among them, the terminal 102 can be, but is not limited to, various personal computers, laptops, smart phones, tablets, Internet of Things devices and portable wearable devices. The Internet of Things devices can be smart speakers, smart TVs, smart air conditioners, smart car devices, projection devices, etc. Portable wearable devices can be smart watches, smart bracelets, head-mounted devices, etc. The head-mounted device can be a virtual reality (VR) device, an augmented reality (AR) device, smart glasses, etc. The server 104 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. Specifically, during the process of managing the distribution network, the server 104 obtains multiple candidate parameter sets corresponding to the distribution network at the target time from the terminal 102; each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes contained in the distribution network; each candidate parameter set corresponds to the same grid node; for each grid node, the voltage amplitude corresponding to the grid node is determined based on the candidate active power and candidate reactive power of the grid node; for each candidate parameter set, when the voltage amplitudes of each grid node in the candidate parameter set meet the voltage constraint condition, the fitness corresponding to the candidate parameter set is determined; each candidate active power in the candidate parameter set whose fitness meets the adaptation condition is used as the threshold of the dynamic operation domain corresponding to each grid node at the target time.

[0045] In an exemplary embodiment, Figure 2 As shown, a distribution network management method is provided, which is applied to Figure 1 The server 104 in the example is used as an example to illustrate the method, which includes the following steps S202 to S208. Among them:

[0046] Step S202: Acquire multiple candidate parameter sets corresponding to the distribution network at the target time.

[0047] Among them, the target moment refers to one or more key time points set by an individual, group or organization when pursuing a certain goal or achieving a certain plan. The distribution network refers to a power grid that receives electric energy from the transmission network or regional power plant, and distributes it locally through distribution facilities or distributes it step by step to various types of users according to voltage levels. It is a network that plays an important role in distributing electric energy in the power grid. It is mainly composed of overhead lines, cables, towers, distribution transformers, switchgear, reactive compensation capacitors and other distribution equipment and ancillary facilities. The main function of the distribution network is to provide electric energy to various users and ensure the stability and reliability of the power supply. The candidate parameter set is a set of one or more sets of parameter values ​​prepared in advance in order to find the optimal or appropriate parameter configuration during scientific experiments, model construction, system optimization, etc.

[0048] Each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes contained in the distribution network; each candidate parameter set corresponds to the same grid node. The nodes in the distribution network refer to the key points in the power grid that connect multiple circuits or devices. These nodes play the role of connecting, distributing and converting electric energy in the distribution network, and are crucial to ensuring the stable operation of the power grid. In the distribution network, nodes usually refer to the key points in the power grid that connect multiple circuits or devices. These nodes can be electrical equipment such as transformers, switchgear, busbars, or a certain intersection in the power grid. Together, they constitute the basic framework of the distribution network, providing important guarantees for the transmission and distribution of electric energy.

[0049] Specifically, in order to manage the distribution network, multiple candidate parameter sets corresponding to the distribution network at the target time can be obtained in advance, each candidate parameter set includes candidate active power and candidate reactive power corresponding to multiple grid nodes contained in the distribution network, and each candidate parameter set corresponds to the same grid node, so as to facilitate the subsequent voltage analysis of each candidate parameter set, and finally determine the optimal solution of the distribution network at the target time. Exemplarily, the method of obtaining multiple candidate parameter sets can be random selection, or through the whale optimization algorithm, simulating the unique search method and hunting mechanism of humpback whales. The position of each humpback whale in WOA (Whale Optimization Algorithm) represents a potential solution. By continuously updating the position of the whale in the solution space, the global optimal solution is finally obtained.

[0050] Step S204: for each grid node, based on the candidate active power and candidate reactive power of the grid node, determine the voltage amplitude corresponding to the grid node.

[0051] The voltage amplitude refers to the maximum value of the positive and negative half-cycle of the AC voltage, also known as the peak-to-peak value. It reflects the peak value of the voltage signal and is an important parameter for judging the strength and stability of the voltage waveform. In electrical engineering, the voltage amplitude is often used to measure the strength of the voltage signal, and the performance and health of the circuit are evaluated by measuring and analyzing the voltage amplitude.

[0052] Specifically, for each grid node, based on the candidate active power and candidate reactive power of the grid node, the voltage amplitude corresponding to the grid node is determined, so as to facilitate the subsequent evaluation of the voltage amplitude. Exemplarily, the process of determining the voltage amplitude corresponding to the grid node can be calculated by a preset neural network model or by establishing a simulated electrical model.

[0053] Step S206 , for each candidate parameter set, when the voltage amplitudes of the respective power grid nodes in the candidate parameter set all satisfy the voltage constraint condition, determine the fitness corresponding to the candidate parameter set.

[0054] Among them, the voltage constraint refers to a certain limit or range that the voltage variable needs to meet in the power system. These constraints are crucial to the stable operation of the power system. Fitness is an indicator to measure the quality of the solution, usually calculated by the objective function. The fitness reflects the performance of the solution on a specific problem and can be used to measure the pros and cons of the solution corresponding to the candidate parameter set. It can be understood that in this embodiment, the desired goal of distribution network management is to allow the sum of injected powers to be minimized, that is, the smaller the fitness, the closer the candidate parameter set is to the desired goal.

[0055] Specifically, since the voltage amplitude during the operation of the distribution network must meet the voltage constraint conditions to ensure the safe operation of the power grid and power grid equipment, after determining the voltage amplitude corresponding to each power grid node, it is also necessary to match the voltage constraint conditions of each voltage amplitude of each power grid node in the candidate parameter set for each candidate parameter set. Only when the voltage amplitude of each power grid node meets the voltage constraint conditions can the next step be performed to determine the fitness corresponding to the candidate parameter set.

[0056] For example, the voltage constraint can be expressed as:

[0057] In the formula is the voltage amplitude of the i grid node at the required calculation time point t, is the root node voltage.

[0058] In some specific embodiments, when the voltage amplitude of each grid node satisfies the voltage constraint condition, the active power of each grid node in the candidate parameter set can be directly superimposed to obtain the total power corresponding to the candidate parameter set, and the fitness corresponding to the candidate parameter set is determined based on the total power. In other specific embodiments, a fitness determination model can also be established, and the fitness corresponding to the candidate parameter set can be determined based on the model.

[0059] Step S208: taking each candidate active power in the candidate parameter set whose fitness satisfies the adaptation condition as a threshold value of the dynamic operation domain corresponding to each grid node at the target moment.

[0060] Among them, the adaptation condition refers to the suitable matching condition corresponding to the distribution network management. In this embodiment, the adaptation condition refers to the minimum value of fitness. The dynamic operation domain describes the output range of distributed power sources (such as distributed generators DG) and microgrids in the distribution network under the premise of ensuring the safety of the distribution network, which can be conveniently used for active scheduling of distributed power sources and microgrids. Compared with the static operation domain, the dynamic operation domain will dynamically calculate the "operation constraints" of the power grid on distributed energy based on the latest forecast or operation data of the distribution network, so as to achieve the purpose of maximizing the consumption of distributed energy.

[0061] Specifically, after determining the fitness corresponding to each candidate parameter set, the candidate parameter set whose fitness meets the adaptation condition can be determined, and each candidate active power in the candidate parameter set is used as the threshold of the dynamic operation domain corresponding to each grid node at the target time.

[0062] The above-mentioned distribution network management method obtains multiple candidate parameter sets corresponding to the distribution network at the target time, wherein each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes contained in the distribution network, and each candidate parameter set corresponds to the same grid node. The operating parameter conditions under multiple operating conditions can be pre-determined. For each grid node, the voltage amplitude corresponding to the grid node is determined based on the candidate active power and candidate reactive power of the grid node, and the actual voltage operating condition corresponding to each grid node under the condition can be determined. For each candidate parameter set, when the voltage amplitudes of each grid node in the candidate parameter set meet the voltage constraint conditions, the fitness corresponding to the candidate parameter set is determined, and the allowable injection power of the power grid under the condition can be determined. The candidate active power in the candidate parameter set whose fitness meets the adaptation condition is used as the threshold of the dynamic operating domain corresponding to each grid node at the target time. The active power corresponding to the condition with the minimum allowable injection power can be used as the threshold of the dynamic operating domain, thereby improving the efficiency of determining the dynamic operating domain.

[0063] In an exemplary embodiment, the voltage amplitude is determined based on the output of the neural network model; the input of the neural network model is the candidate active power and candidate reactive power of the power grid node; the process of establishing the neural network model includes: obtaining the historical active power, historical reactive power, and historical voltage amplitude of the distribution network at historical moments; obtaining an initial model, using the historical active power, historical reactive power, and historical voltage amplitude as training data for the initial model, training the initial model, and obtaining a neural network model.

[0064] The historical moment refers to the moment in the historical state, that is, the moment that has occurred. The historical active power, historical reactive power, and historical voltage amplitude are all the active power, reactive power, and voltage amplitude at the historical moment. The initial model refers to the untrained neural network model in the initial state.

[0065] Specifically, the process of determining the voltage amplitude corresponding to the grid node based on the candidate active power and candidate reactive power of the grid node can be implemented through a neural network model, that is, the voltage amplitude is determined based on the output of the neural network model, and the input of the neural network model is the candidate active power and candidate reactive power of the grid node. The process of establishing the neural network is: obtaining the historical active power, historical reactive power, historical voltage amplitude, and initial model of the distribution network at historical moments, and training the initial model based on the above historical data to obtain the neural network model.

[0066] In some specific embodiments, the constructed neural network model is as follows: In the input layer, the input data set is processed by the sequence folding layer seqfold. The significance of the sequence folding layer is to fold the input sequence data according to the specified window size to form multiple subsequences, and then arrange these subsequences into a new matrix or tensor, which can not only reduce the data dimension, but also enable the convolution layer to effectively perform convolution operations and extract local features. After the data passes through the sequence folding layer, it enters the first convolution layer conv_1, in which a convolution kernel of size 3x1 is used to perform a convolution operation on the data to generate a feature map with a channel number of 32 to extract the features of the input sequence.

[0067] After the initial feature extraction, the data will be processed in parallel through two channels. One of the channels continues to perform convolution operations to extract features. The feature map passes through the activation layer relu_1, which uses the ReLU function for nonlinear mapping to alleviate the gradient vanishing problem and accelerate the convergence speed of the model. Next, it enters the second convolution layer conv_2, which will further extract features based on con_1 to increase the expressiveness and complexity of the model. The convolution kernel size of conv_2 is also 3x1, that is, convolution is performed along the height of the feature map, but no convolution operation is performed in the width direction of the feature map. The purpose of this design is to increase the ability to extract spatial features while maintaining the time series information of the input feature map. After the second convolution layer, a feature map with 64 channels will be generated similarly, and activated by relu_2 to help the model capture more advanced and abstract features, making the model more discriminative and generalizable. While performing convolution, the second data channel is the related operations of the Attention mechanism performed in parallel. After the convolution layer, the feature map is globally averaged pooled through the global average pooling layer gapool to obtain a feature vector. The global average pooling layer does not use a fixed-size pooling window to pool the feature map like a regular pooling layer, but uses the average value of the entire feature map as the output. This means that regardless of the size of the feature map, the global average pooling layer will generate a fixed-size output, thereby reducing the number of parameters and computational complexity, and enhancing the robustness of the model. Next, the feature vector passes through the fully connected layer fc_2, flattens the feature map into a one-dimensional vector, and further processes and abstracts the features extracted by the global average pooling layer to better capture the high-level features of the input data and provide a more meaningful representation for subsequent tasks. After being activated by relu_3, the data enters the second fully connected layer fc_3. Compared with fc_2, fc_3 has more output channels, which can be regarded as a further mapping and abstraction of features for higher-level feature representation and extraction. Next, after the feature extraction is completed, the sigmoid layer is activated to limit the output to the range of (0, 1) and generate attention weights for weighting the feature vector. After the weighting is completed, the dot product layer is used to fuse the parallel operation of the convolution operation and the attention mechanism, multiplying one tensor with another tensor as the attention weight, so that the attention weight is applied to the convolution layer output, strengthening the influence of important features and generating a fused feature representation. The attention mechanism calculation formula is as follows:

[0068]

[0069] The fused feature representation is restored from the folded state to the original sequence form through the sequence unfolding layer sequnfold and the data flattening layer flatten, and flattened into one-dimensional data to meet the input requirements of LSTM. Then it is input into the bidirectional LSTM layer bilstm, which takes the feature sequence as input and performs bidirectional LSTM calculations in the hidden layer to learn the long-term dependencies of the sequence data. The output of the BiLSTM layer is connected to the regression layer through the fully connected layer fc for regression prediction, calculates the loss between the predicted value and the true value, and uses the back propagation algorithm to adjust the network parameters to reduce the value of the loss function. The parameter update is completed through the optimizer, and the weights and biases in the network are updated according to the gradient of the loss function.

[0070] In this embodiment, the initial model is trained by historical data to obtain a neural network model that can use candidate active power and candidate reactive power of the grid node as input and voltage amplitude as output, which can improve the accuracy of voltage amplitude determination.

[0071] In an exemplary embodiment, the historical active power of the distribution network at a historical moment is obtained, including: obtaining the initial historical active power of the distribution network at the historical moment; when there is an abnormality in the initial historical active power, based on the historical active power at the previous historical moment and the historical active power at the next moment, power compensation is performed on the active power at the historical moment to obtain the historical active power of the distribution network at the historical moment.

[0072] The initial historical active power refers to the historical active power in the initial state without being modified or updated.

[0073] Specifically, before training the neural network, it is first necessary to supplement the missing values ​​of the data set. General methods for supplementing missing values ​​include mean supplementation, mode supplementation, median supplementation, etc., but this supplementation method often deviates greatly from the actual value. Since the data set contains a strong time correlation, the missing values ​​are often highly correlated with the previous and next sampling points, and generally cannot mutate. Therefore, the method of filling the missing values ​​in this embodiment is to take the average of the previous sampling point and the next sampling point. The compensation of the missing values ​​often improves the training effect of the model. That is, the initial historical active power of the distribution network at the historical moment can be obtained. In the case of an abnormality in the initial historical active power, the active power at the historical moment is compensated based on the historical active power at the previous historical moment and the historical active power at the next moment, and the historical active power of the distribution network at the historical moment is obtained. The historical active power. It can be understood that the above method can be used for both historical reactive power and historical voltage amplitude, and it will not be repeated here.

[0074] Secondly, the data set needs to be normalized. Normalization can process all data into numbers between [0, 1], optimize model performance, and make data processing more convenient. This article uses Max-min normalization, the formula is as follows:

[0075]

[0076] The neural network takes the minimum RMSE as the optimization goal, and the calculation formula is as follows:

[0077]

[0078] In this embodiment, based on the historical active power at the previous historical moment and the historical active power at the next historical moment, power compensation is performed on the active power at the historical moment to obtain the historical active power of the distribution network at the historical moment, which can improve the training effect of the model.

[0079] In an exemplary embodiment, when the voltage amplitudes of each grid node in the candidate parameter set satisfy the voltage constraint conditions, the fitness corresponding to the candidate parameter set is determined, including: when the voltage amplitudes of each grid node in the candidate parameter set satisfy the voltage constraint conditions, the active power of each grid node is superimposed to obtain the total power corresponding to the candidate parameter set; based on the total power, the fitness corresponding to the candidate parameter set is determined.

[0080] Specifically, if the voltage amplitudes of each grid node in the candidate parameter set meet the voltage constraint conditions, the active power of each grid node can be superimposed to obtain the total power corresponding to the candidate parameter set, and the fitness corresponding to the candidate parameter set is determined based on the total power.

[0081] For example, the optimization goal is to maximize the dynamic operation domain, even if the sum of the injected powers is allowed to be minimized (the injected power is negative), and the specific objective function (fitness) formula is as follows:

[0082]

[0083] In the formula is the active power of the grid node, and the set Represents a distributed resource collection , is the number of grid nodes, , A collection of moment indexes in a dataset.

[0084] If at least one of the voltage amplitudes of each grid node in the candidate parameter set does not meet the voltage constraint condition, the fitness value Set to infinity .

[0085] In this embodiment, when the voltage amplitudes of each grid node in the candidate parameter set satisfy the voltage constraint conditions, the active power of each grid node is superimposed to obtain the total power corresponding to the candidate parameter set, and based on the total power, the fitness corresponding to the candidate parameter set is determined, which can ensure the normal operation of the distribution network.

[0086] In an exemplary embodiment, multiple candidate parameter sets corresponding to the distribution network at the target moment are obtained, including: obtaining the current active power and the current reactive power corresponding to each of the multiple grid nodes in the distribution network at the target moment; adjusting each current active power and each current reactive power respectively to obtain multiple candidate parameter sets.

[0087] Among them, the current active power and the current reactive power may be the active power and reactive power corresponding to the target moment.

[0088] Specifically, the current active power and reactive power corresponding to each of the multiple grid nodes in the distribution network at the target time can be obtained first, and then each current active power and each current reactive power can be adjusted respectively to obtain multiple candidate parameter sets. The adjustment method can be to adjust toward the optimal solution through the whale optimization algorithm, thereby improving the efficiency of dynamic operation domain determination.

[0089] In an exemplary embodiment, the target moment includes a future moment; the method also includes: obtaining multiple prediction parameter sets corresponding to the distribution network at the future moment; each prediction parameter set includes predicted active power and predicted reactive power corresponding to multiple grid nodes; multiple grid nodes corresponding to each prediction parameter set are the same; for each grid node, based on the predicted active power and predicted reactive power of the grid node, determine the predicted voltage amplitude corresponding to the grid node; for each prediction parameter set, when the voltage amplitudes of each grid node in the prediction parameter set meet the voltage constraint conditions, determine the prediction fitness corresponding to the prediction set; use each predicted active power in the prediction parameter set whose prediction fitness meets the adaptation conditions as the prediction threshold of the dynamic operation domain corresponding to each grid node at the future moment.

[0090] Among them, the future moment is the moment that has not yet occurred. Therefore, the above-mentioned prediction data are all data obtained by advance estimation.

[0091] Specifically, the input of the model includes the active power of the grid nodes , reactive power , and the reactive power of the grid nodes in the dynamic operation domain needs to be calculated Depending on the corresponding tasks, the input can be divided into the following three cases:

[0092] Real-time calculation: If the input data is real-time data of the node, the model can perform real-time calculation to allocate the dynamic operation domain of each node.

[0093] Short-term prediction calculation: If the input data is the prediction data for the next time point, the model can calculate the dynamic operation domain for the next time point in advance through the prediction data, so that the distribution network operator can release the dynamic operation domain in advance.

[0094] Long-term prediction calculation: If the input data is the prediction data for the next time step (30 minutes, 1 hour, 1 day), the model can calculate the dynamic operation domain for the next time step through the prediction data, and the distribution network operator can publish the dynamic operation domain for a period of time in advance.

[0095] In this embodiment, it is a short-time prediction calculation, that is, for each grid node, based on the predicted active power and predicted reactive power of the grid node, the predicted voltage amplitude corresponding to the grid node can be determined; for each prediction parameter set, when the voltage amplitudes of each grid node in the prediction parameter set meet the voltage constraint conditions, the prediction fitness corresponding to the prediction set is determined, and each predicted active power in the prediction parameter set whose prediction fitness meets the adaptation conditions is used as the prediction threshold of the dynamic operation domain corresponding to each grid node at the future moment, thereby improving the flexibility of distribution network management.

[0096] In a specific embodiment, a method for calculating a dynamic operating domain without an electrical model based on an improved CNN-LSTM neural network model is also provided. This method innovatively uses easily accessible smart meters to obtain data to train the improved CNN-LSTM neural network and capture the structure of the distribution network, thereby avoiding the acquisition of distribution network topology and line parameters that are difficult to observe, realizing the calculation of the dynamic operating domain, saving time and resources, and reducing the complexity of calculation, improving the economy and ensuring the safety of the distribution network within the dynamic operating domain. In addition, the whale optimization algorithm is used to automatically optimize the model parameters, reduce manual operations during model training, simplify the complexity of model use, and improve the portability of the invention, so that it can be put into use in different areas, and it can also use new data to update and reuse the model when the distribution network topology or line parameters change.

[0097] 1Online Training

[0098] 1.1 Building a Dataset

[0099] According to the user's smart meter data, calculate the active power of each grid node at time t (Set the injected power to negative), reactive power , and the voltage amplitude ,in , is the index set of all power grid nodes, is the number of grid nodes, The dataset is a collection of moment indices in the dataset. Therefore, the dataset collects active power, reactive power, and voltage amplitude of each grid node at fixed time intervals (5 minutes, 30 minutes, 60 minutes, etc.), and then considers a whole time period (such as a month, a quarter, a year, etc.) for training.

[0100] The dataset format is as follows:

[0101]

[0102]

[0103]

[0104] 1.2 Building a Deep Learning Neural Network

[0105] The WOA-CNN-LSTM-Attention network architecture proposed in this embodiment consists of four main modules: whale optimization algorithm, convolutional neural network, bidirectional long short-term memory network and attention mechanism. Among them, WOA is used to optimize the three hyperparameters of learning rate, hidden layer grid nodes and regularization coefficient; and the powerful local feature extraction capability of CNN is used to capture the topological structure of the distribution network model and provide BiLSTM with more discriminative input representation; LSTM is used to fully explore the correlation of the voltage amplitude dimension data sequence in the time dimension, and Attention highlights the key moment features that have a greater impact on the prediction results by weighted aggregation of hidden states at different times. The neural network input format is as follows:

[0106]

[0107] 1.2.1 Whale Optimization Algorithm WOA

[0108] WOA simulates the unique search method and hunting mechanism of humpback whales, which mainly includes three stages: hunting prey, bubble net hunting, and searching for prey. The position of each humpback whale in WOA represents a potential solution. By continuously updating the position of the whale in the solution space, the global optimal solution is finally obtained. In the stage of surrounding prey, imitating the hunting strategy of the whale group, each individual in the algorithm adjusts its own position and speed to try to gradually surround the target prey, that is, to find the optimal solution or approximate optimal solution for the target. Specifically, in the entire search space, the individual will move toward the target prey according to the current position, and constantly adjust its speed and direction to better surround the target. After the best search is defined, the search of other individuals will also be close to the best search update. This process is expressed by the following formula:

[0109]

[0110]

[0111] is the current iteration number, is the position vector of the current best individual, is the coefficient vector. If there is a better solution, the algorithm will continue to update the position of the optimal individual. . The vector The expression is as follows:

[0112]

[0113]

[0114] Throughout the iteration process Linearly decreases from 2 to 0; and is a random vector in [0, 1].

[0115] There are two types of humpback whale hunting: using bubble nets for hunting and encircling. In the process of using bubble nets for hunting, whales will gradually narrow their encirclement to get closer to their prey. In this case, in order to simulate the behavior of whales using bubble nets to gather prey, individuals need to perform a spiral search near the current optimal solution to update the position between the whale and the prey, as follows:

[0116]

[0117]

[0118] in is the distance between the current search individual and the current optimal solution, is the spiral shape parameter, is a random number with a uniform distribution in the range of [-1,1]. Since there are two predation behaviors in the process of approaching prey, WOA chooses bubble net predation or encirclement predation according to probability p, and the position update is expressed as follows:

[0119]

[0120] is a random number with a value range of [0,1]. and As the number of iterations increases, it gradually decreases. , then the whales gradually surround the current optimal solution and enter the local optimization stage.

[0121] In the prey search phase, the remaining individuals conduct a global search to find target prey that has not yet been captured. This phase is similar to the behavior of whales searching for prey in the vast ocean. Individuals explore potential prey locations in the solution space by adjusting their positions and directions. When , the search individual will swim towards the random whale, as follows:

[0122]

[0123]

[0124] This embodiment uses WOA to optimize the hyperparameters of the network architecture. The specific method of initializing the WOA algorithm is to use a zero vector to initialize the position vector of the global optimal solution, and the fitness score of the global optimal solution is positive infinity. In the iterative loop, it is necessary to continuously update the leader of the whale, that is, for each search agent, by comparing its fitness with the current global optimal solution, determine whether to update the leader. If the fitness of a search agent is better than the current global optimal solution, the global optimal solution and the corresponding position are updated. In addition, the position of each search agent is also constantly updated. According to the random numbers r1, r2 and the position parameter parameters, the variables (A, C, b, l, p) required to update the search agent position are calculated. Among them, A represents the moving step length of the control search agent, and the value range is [−1,1]; C represents the moving direction of the adjustment search agent, and the value range is [0,2]; b represents the moving speed of the adjustment search agent, which is usually a positive number; l is used to adjust the moving distance of the search agent, which is usually randomly selected in the range of [−1,1]. If the random number p is less than 0.5, it means that the leader following or random exploration update rule is executed, and the position of the search agent is adjusted according to the values ​​of A and C. If the random number p is greater than or equal to 0.5, it means that the local search update rule is executed, and the position of the search agent is adjusted according to the distance from the leader and the randomly generated parameters b and l. The entire algorithm will continuously update the position of the search agent in the loop iteration until the optimal hyperparameters are found.

[0125] 1.2.2CNN-LSTM-Attention Neural Network

[0126] The neural network constructed in this embodiment is as follows. In the input layer, the input data set is processed by the sequence folding layer seqfold. The significance of the sequence folding layer is to fold the input sequence data according to the specified window size to form multiple subsequences, and then arrange these subsequences into a new matrix or tensor, which can not only reduce the data dimension, but also enable the convolution layer to effectively perform convolution operations and extract local features. After the data passes through the sequence folding layer, it enters the first convolution layer conv_1, where a convolution kernel of size 3x1 is used to perform a convolution operation on the data to generate a feature map with a channel number of 32 to extract the features of the input sequence.

[0127] After the initial feature extraction, the data will be processed in parallel through two channels. One of the channels continues to perform convolution operations to extract features. The feature map passes through the activation layer relu_1, which uses the ReLU function for nonlinear mapping to alleviate the gradient vanishing problem and accelerate the convergence speed of the model. Next, it enters the second convolution layer conv_2, which will further extract features based on con_1 to increase the expressiveness and complexity of the model. The convolution kernel size of conv_2 is also 3x1, that is, convolution is performed along the height of the feature map, but no convolution operation is performed in the width direction of the feature map. The purpose of this design is to increase the ability to extract spatial features while maintaining the time series information of the input feature map. After the second convolution layer, a feature map with 64 channels will be generated similarly, and activated by relu_2 to help the model capture more advanced and abstract features, making the model more discriminative and generalizable. While performing convolution, the second data channel is the related operations of the Attention mechanism performed in parallel. After the convolution layer, the feature map is globally averaged pooled through the global average pooling layer gapool to obtain a feature vector. The global average pooling layer does not use a fixed-size pooling window to pool the feature map like a regular pooling layer, but uses the average value of the entire feature map as the output. This means that regardless of the size of the feature map, the global average pooling layer will generate a fixed-size output, thereby reducing the number of parameters and computational complexity, and enhancing the robustness of the model. Next, the feature vector passes through the fully connected layer fc_2, flattens the feature map into a one-dimensional vector, and further processes and abstracts the features extracted by the global average pooling layer to better capture the high-level features of the input data and provide a more meaningful representation for subsequent tasks. After being activated by relu_3, the data enters the second fully connected layer fc_3. Compared with fc_2, fc_3 has more output channels, which can be regarded as a further mapping and abstraction of features for higher-level feature representation and extraction. Next, after the feature extraction is completed, the sigmoid layer is activated to limit the output to the range of (0, 1) and generate attention weights for weighting the feature vector. After the weighting is completed, the dot product layer is used to fuse the parallel operation of the convolution operation and the attention mechanism, multiplying one tensor with another tensor as the attention weight, so that the attention weight is applied to the convolution layer output, strengthening the influence of important features and generating a fused feature representation. The attention mechanism calculation formula is as follows:

[0128]

[0129] The fused feature representation is restored from the folded state to the original sequence form through the sequence unfolding layer sequnfold and the data flattening layer flatten, and flattened into one-dimensional data to meet the input requirements of LSTM. Then it is input into the bidirectional LSTM layer bilstm, which takes the feature sequence as input and performs bidirectional LSTM calculations in the hidden layer to learn the long-term dependencies of the sequence data. The output of the BiLSTM layer is connected to the regression layer through the fully connected layer fc for regression prediction, calculates the loss between the predicted value and the true value, and uses the back propagation algorithm to adjust the network parameters to reduce the value of the loss function. The parameter update is completed through the optimizer, and the weights and biases in the network are updated according to the gradient of the loss function.

[0130] 1.2.3 Training the Neural Network

[0131] Before training the neural network, you first need to fill in the missing values ​​of the data set. General methods for filling in missing values ​​include mean filling, mode filling, median filling, etc., but this method of filling often deviates greatly from the actual value. Because the data set contains a strong time correlation, the missing values ​​are often highly correlated with the previous and next sampling points, and generally cannot change suddenly. Therefore, the way to fill in the missing values ​​in this article is to take the average of the previous sampling point and the next sampling point. Filling in missing values ​​often improves the training effect of the model. Secondly, the data set needs to be normalized. Normalization can process all data into numbers between [0, 1], optimize model performance, and make data processing more convenient. This article uses Max-minnormalization, and the formula is as follows:

[0132]

[0133] The neural network takes the minimum RMSE as the optimization goal, and the calculation formula is as follows:

[0134]

[0135] 2 Offline deployment

[0136] 2.1 Model Input

[0137] The input of the model includes the active power of the non-distributed resource user grid node , reactive power , and the reactive power of distributed resource grid nodes in the dynamic operation domain needs to be calculated Depending on the corresponding tasks, the input can be divided into the following three cases:

[0138] 2.1.1 Real-time calculation: If the input data is the real-time data of the power grid nodes, the model can perform real-time calculation to allocate the dynamic operation domain of each power grid node.

[0139] 2.1.2 Short-term prediction calculation: If the input data is the prediction data for the next time point, the model can calculate the dynamic operation domain for the next time point in advance through the prediction data, so that the distribution network operator can release the dynamic operation domain in advance.

[0140] 2.1.3 Long-term prediction calculation: If the input data is the prediction data for the next time step (30 minutes, 1 hour, 1 day), the model can calculate the dynamic operation domain for the next time step through the prediction data, and the distribution network operator can publish the dynamic operation domain for a period of time in advance.

[0141] 2.2 Calculating the dynamic operation domain

[0142] This embodiment uses the whale optimization algorithm and the trained neural network model to calculate the dynamic operation domain. The optimization goal is to maximize the dynamic operation domain, even if the sum of the injected power is allowed to be minimized (the injected power is negative). The specific objective function formula is as follows:

[0143]

[0144] In the formula is the active power of the distributed resource grid node, and the set Represents a distributed resource collection , is the number of distributed resource grid nodes, , A collection of moment indexes in a dataset.

[0145] The main steps to calculate the dynamic operating domain are as follows:

[0146] 2.2.1 Initialize the model:

[0147] The initial number of whales is 30, the maximum number of iterations is 50, the upper bound is set to 0, the lower bound is set to -10 (depending on the task), and the constraints are voltage constraints, as follows:

[0148]

[0149] In the formula is the voltage amplitude of the i grid node at the required calculation time point t, is the root grid node voltage.

[0150] 2.2.2 Initialization input:

[0151] Initialize the injected power of the distributed resource grid node to 0, that is, , calculate the voltage of each grid node , if the constraints are met, it is set as the optimal population and the initial fitness value is calculated , proceed to the next step.

[0152] 2.2.3 Perform iterative calculation:

[0153] Using the whale optimization algorithm to optimize Search, and after each search, determine the constraints. If the constraints are not met, the fitness value Set to infinity Otherwise, the fitness value is calculated normally. Finally, it is compared with the optimal fitness value. If it is less than the optimal fitness value, the group of populations is set as the optimal population, and the fitness value is replaced and the iteration continues; otherwise, the iteration continues directly.

[0154] 2.2.4 End calculation:

[0155] When the maximum number of iterations is reached, the dynamic running domain calculation ends.

[0156] 2.3 Publishing Dynamic Operation Domains

[0157] After the previous step, the optimal injection power of each distributed resource grid node at time t has been obtained. , then the dynamic operation domain of these power grid nodes is:

[0158]

[0159] The injected power of the jth grid node where the distributed resource is located at time t needs to satisfy:

[0160]

[0161] 3 Model Update

[0162] It should be pointed out that the dynamic operation domain calculation method proposed in this embodiment needs to be dynamically updated. When the topology structure and line parameters of the distribution network change, such as adding new grid nodes and adding new electrical equipment, the accuracy of the model will be reduced, so the neural network model needs to be retrained. Distribution operators can use newly acquired smart meter data to retrain the model regularly according to their own needs, and then update it. In addition, since this embodiment uses the whale optimization algorithm for automatic parameter adjustment and optimization, there is no need to manually set or update parameters when the model is updated, which greatly reduces the workload and reliability.

[0163] In a specific embodiment, Figure 3 As shown, a distribution network management method is also provided, including:

[0164] Step S301, obtaining the initial historical active power of the distribution network at a historical moment;

[0165] Step S302, when the initial historical active power is abnormal, based on the historical active power at the previous historical moment and the historical active power at the next moment, power compensation is performed on the active power at the historical moment to obtain the historical active power of the distribution network at the historical moment;

[0166] Step S303, obtaining the historical reactive power and historical voltage amplitude of the distribution network at the historical moment;

[0167] Step S304, obtaining an initial model, using historical active power, historical reactive power, and historical voltage amplitude as training data for the initial model, training the initial model, and obtaining a neural network model;

[0168] The voltage amplitude is determined based on the output of the neural network model; the input of the neural network model is the candidate active power and candidate reactive power of the grid node;

[0169] Step S305, obtaining the current active power and the current reactive power corresponding to each of the multiple power grid nodes in the distribution network at the target time;

[0170] Step S306, adjusting each current active power and each current reactive power respectively to obtain multiple candidate parameter sets;

[0171] Wherein, each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of a plurality of grid nodes included in the distribution network; each candidate parameter set corresponds to the same grid node;

[0172] Step S307, for each grid node, based on the candidate active power and candidate reactive power of the grid node, determining a voltage amplitude corresponding to the grid node;

[0173] Step S308, for each candidate parameter set, when the voltage amplitude of each grid node in the candidate parameter set satisfies the voltage constraint condition, the active power of each grid node is superimposed to obtain the total power corresponding to the candidate parameter set;

[0174] Step S309, based on the total power, determine the fitness corresponding to the candidate parameter set, and use each candidate active power in the candidate parameter set whose fitness meets the adaptation condition as the threshold of the dynamic operation domain corresponding to each grid node at the target time;

[0175] Among them, the target moment includes a future moment;

[0176] Step S310, obtaining a plurality of prediction parameter sets corresponding to the distribution network at a future time;

[0177] Wherein, each prediction parameter set includes predicted active power and predicted reactive power corresponding to each of the plurality of power grid nodes; the plurality of power grid nodes corresponding to each prediction parameter set are the same;

[0178] Step S311, for each grid node, based on the predicted active power and predicted reactive power of the grid node, determining a predicted voltage amplitude corresponding to the grid node;

[0179] Step S312, for each prediction parameter set, when the voltage amplitudes of the respective power grid nodes in the prediction parameter set all satisfy the voltage constraint condition, determining the prediction fitness corresponding to the prediction set;

[0180] Step S313: each predicted active power in the prediction parameter set whose prediction fitness satisfies the adaptation condition is used as a prediction threshold of the dynamic operation domain corresponding to each grid node at a future moment.

[0181] It should be understood that, although the steps in the flowcharts involved in the above embodiments are displayed in sequence according to the indication of the arrows, these steps are not necessarily executed in sequence according to the order indicated by the arrows. Unless there is a clear explanation in this article, the execution of these steps is not strictly limited in order, and these steps can be executed in other orders. Moreover, at least a part of the steps in the flowcharts involved in the above embodiments may include multiple steps or multiple stages, and these steps or stages are not necessarily executed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily carried out in sequence, but can be executed in turn or alternately with other steps or at least a part of the steps or stages in other steps.

[0182] Based on the same inventive concept, the embodiment of the present application also provides a distribution network management device for implementing the distribution network management method involved above. The implementation solution provided by the device to solve the problem is similar to the implementation solution recorded in the above method, so the specific limitations in one or more distribution network management device embodiments provided below can refer to the limitations of the distribution network management method above, and will not be repeated here.

[0183] In an exemplary embodiment, Figure 4 As shown, a distribution network management device 400 is provided, including: a parameter set acquisition module 402, a voltage amplitude determination module 404, a fitness determination module 406 and a dynamic operation domain determination module 408, wherein:

[0184] The parameter set acquisition module 402 is used to acquire multiple candidate parameter sets corresponding to the distribution network at the target time; each candidate parameter set includes candidate active power and candidate reactive power corresponding to each of multiple grid nodes included in the distribution network; each candidate parameter set corresponds to the same grid node;

[0185] A voltage amplitude determination module 404 is used to determine, for each grid node, a voltage amplitude corresponding to the grid node based on the candidate active power and candidate reactive power of the grid node;

[0186] A fitness determination module 406 is used to determine the fitness corresponding to each candidate parameter set for each candidate parameter set when the voltage amplitude of each grid node in the candidate parameter set satisfies the voltage constraint condition;

[0187] The dynamic operation domain determination module 408 is used to use each candidate active power in the candidate parameter set whose fitness satisfies the adaptation condition as the threshold of the dynamic operation domain corresponding to each grid node at the target moment.

[0188] In an exemplary embodiment, the voltage amplitude is determined based on the output of the neural network model; the input of the neural network model is the candidate active power and candidate reactive power of the grid node. In the case of this embodiment, the distribution network management device 400 also includes a model building module, including:

[0189] A data acquisition unit, used to acquire the historical active power, historical reactive power, and historical voltage amplitude of the distribution network at a historical moment;

[0190] The model training unit is used to obtain an initial model, use historical active power, historical reactive power, and historical voltage amplitude as training data for the initial model, train the initial model, and obtain a neural network model.

[0191] In an exemplary embodiment, the data acquisition unit is specifically used for:

[0192] Obtain the initial historical active power of the distribution network at the historical moment;

[0193] When the initial historical active power is abnormal, the active power at the historical moment is compensated based on the historical active power at the previous historical moment and the historical active power at the next historical moment to obtain the historical active power of the distribution network at the historical moment.

[0194] In an exemplary embodiment, the fitness determination module 406 is specifically used to:

[0195] When the voltage amplitude of each grid node in the candidate parameter set satisfies the voltage constraint condition, the active power of each grid node is superimposed to obtain the total power corresponding to the candidate parameter set;

[0196] Based on the total power, the fitness corresponding to the candidate parameter set is determined.

[0197] In an exemplary embodiment, the parameter set acquisition module 402 is specifically used to:

[0198] Obtain the current active power and current reactive power corresponding to each of multiple grid nodes in the distribution network at the target time;

[0199] Each current active power and each current reactive power are adjusted respectively to obtain a plurality of candidate parameter sets.

[0200] In an exemplary embodiment, the target time includes a future time. In the case of this embodiment, the distribution network management device 400 also includes a future dynamic operation domain determination module, which is specifically used to:

[0201] Acquire multiple prediction parameter sets corresponding to the distribution network at a future moment; each prediction parameter set includes predicted active power and predicted reactive power corresponding to each of multiple grid nodes; and multiple grid nodes corresponding to each prediction parameter set are the same;

[0202] For each grid node, based on the predicted active power and predicted reactive power of the grid node, determine a predicted voltage amplitude corresponding to the grid node;

[0203] For each prediction parameter set, when the voltage amplitudes of the respective power grid nodes in the prediction parameter set satisfy the voltage constraint conditions, determining the prediction fitness corresponding to the prediction set;

[0204] Each predicted active power in the prediction parameter set whose prediction fitness meets the adaptation condition is used as the prediction threshold of the dynamic operation domain corresponding to each grid node at the future moment.

[0205] Each module in the above-mentioned distribution network management device can be implemented in whole or in part by software, hardware, or a combination thereof. Each of the above-mentioned modules can be embedded in or independent of a processor in a computer device in the form of hardware, or can be stored in a memory in a computer device in the form of software, so that the processor can call and execute the operations corresponding to each of the above modules.

[0206] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 5As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface, the display unit and the input device are connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and the external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be implemented through WIFI, a mobile cellular network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a distribution network management method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device shell, or an external keyboard, touchpad or mouse.

[0207] Those skilled in the art will understand that Figure 5 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0208] In one embodiment, a computer device is provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps of the above method when executing the computer program.

[0209] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the above method are implemented.

[0210] In one embodiment, a computer program product is provided, comprising a computer program, which implements the steps of the above method when executed by a processor.

[0211] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0212] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. As an illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. Non-relational databases may include distributed databases based on blockchains, etc., but are not limited to this. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, an artificial intelligence (AI) processor, etc., but are not limited to this.

[0213] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0214] The above embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the present application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the present application shall be subject to the attached claims.

Claims

1. A distribution network management method, characterized in that: The method comprises: Acquire multiple candidate parameter sets corresponding to the distribution network at the target time; each of the candidate parameter sets includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes included in the distribution network; each of the candidate parameter sets corresponds to the same grid node; For each of the grid nodes, determining a voltage amplitude corresponding to the grid node based on a candidate active power and a candidate reactive power of the grid node; For each of the candidate parameter sets, when the voltage amplitudes of the respective power grid nodes in the candidate parameter set satisfy the voltage constraint condition, determining the fitness corresponding to the candidate parameter set; Each candidate active power in the candidate parameter set whose fitness satisfies the adaptation condition is used as a threshold value of the dynamic operation domain corresponding to each of the grid nodes at the target moment.

2. The method according to claim 1, characterized in that: The voltage amplitude is determined based on the output of a neural network model; the input of the neural network model is the candidate active power and candidate reactive power of the grid node; The process of establishing the neural network model includes: Obtaining historical active power, historical reactive power, and historical voltage amplitude of the distribution network at a historical moment; An initial model is obtained, and the historical active power, the historical reactive power, and the historical voltage amplitude are used as training data for the initial model. The initial model is trained to obtain a neural network model.

3. The method according to claim 2, characterized in that The obtaining of the historical active power of the distribution network at a historical moment includes: Obtaining the initial historical active power of the distribution network at a historical moment; When an abnormality exists in the initial historical active power, power compensation is performed on the active power at the historical moment based on the historical active power at the previous historical moment and the historical active power at the next historical moment to obtain the historical active power of the distribution network at the historical moment.

4. The method according to claim 1, characterized in that: The determining of the fitness corresponding to the candidate parameter set when the voltage amplitudes of the respective power grid nodes in the candidate parameter set all satisfy the voltage constraint condition comprises: When the voltage amplitudes of the respective grid nodes in the candidate parameter set satisfy the voltage constraint conditions, performing power superposition on the respective active powers of the respective grid nodes to obtain the total power corresponding to the candidate parameter set; Based on the total power, a fitness corresponding to the candidate parameter set is determined.

5. The method according to claim 1, characterized in that The step of obtaining multiple candidate parameter sets corresponding to the distribution network at the target time includes: Obtain the current active power and current reactive power corresponding to each of multiple grid nodes in the distribution network at the target time; Each of the current active powers and each of the current reactive powers are adjusted respectively to obtain a plurality of candidate parameter sets.

6. The method according to claim 1, characterized in that The target time includes a future time; the method further includes: Acquire multiple prediction parameter sets corresponding to the distribution network at a future moment; each of the prediction parameter sets includes predicted active power and predicted reactive power corresponding to multiple grid nodes; and the multiple grid nodes corresponding to each of the prediction parameter sets are the same; For each of the grid nodes, based on the predicted active power and predicted reactive power of the grid node, determining a predicted voltage amplitude corresponding to the grid node; For each of the prediction parameter sets, when the voltage amplitudes of the respective power grid nodes in the prediction parameter set satisfy the voltage constraint condition, determining the prediction fitness corresponding to the prediction set; Each predicted active power in the prediction parameter set whose prediction fitness satisfies the adaptation condition is used as the prediction threshold of the dynamic operation domain corresponding to each of the power grid nodes at the future moment.

7. A distribution network management device, characterized in that: The device comprises: A parameter set acquisition module, used to acquire multiple candidate parameter sets corresponding to the distribution network at the target time; each of the candidate parameter sets includes candidate active power and candidate reactive power corresponding to each of the multiple grid nodes included in the distribution network; each of the candidate parameter sets corresponds to the same grid node; A voltage amplitude determination module, configured to determine, for each of the grid nodes, a voltage amplitude corresponding to the grid node based on the candidate active power and candidate reactive power of the grid node; A fitness determination module, configured to determine, for each of the candidate parameter sets, the fitness corresponding to the candidate parameter set when the voltage amplitudes of the respective power grid nodes in the candidate parameter set all satisfy the voltage constraint condition; The dynamic operation domain determination module is used to use each candidate active power in the candidate parameter set whose fitness meets the adaptation condition as the threshold of the dynamic operation domain corresponding to each of the grid nodes at the target moment.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 6 are implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 6 are implemented.

Images