Computing power load balancing method and device

Abstract

The present invention discloses a computing power load balancing method and device, which belongs to the field of digital twins, wherein the method includes: reading a predetermined indicator set; collecting the operation data of the distribution network in the digital twin distribution network model based on the predetermined indicator set to obtain the distribution network operation status information; obtaining the distribution network topology of the distribution network; combining the distribution network operation status information and the distribution network topology to calculate the power flow of the distribution network to obtain the distribution network power flow information; visualizing the distribution network power flow information according to the digital twin technology, and obtaining the target power distribution abnormal node; adjusting the load of the distribution network by starting the target load balancer at the target power distribution abnormal node. The present application solves the technical problem of uneven distribution of computing power load in the existing distribution network, and achieves the technical effect of optimizing the configuration of the distribution network computing power load and realizing energy conservation and emission reduction.

Description

Technical Field

[0001] The present invention relates to the field of digital twins, and in particular to a computing power load balancing method and device. Background Art

[0002] With the rapid development of social economy, the computing load of distribution network has shown a trend of continuous growth. However, due to the influence of various factors such as unbalanced regional development and differences in users' electricity consumption habits, the existing distribution network has serious problems of uneven distribution of computing load in different regions and at different times, which has become one of the important factors restricting the safe and economic operation of distribution network. Summary of the invention

[0003] This application aims to solve the technical problem of uneven distribution of computing power load in the existing power distribution network by providing a computing power load balancing method and device.

[0004] In view of the above problems, the present application provides a computing power load balancing method and device.

[0005] The first aspect disclosed in the present application provides a computing power load balancing method, which includes: reading a predetermined indicator set, wherein the predetermined indicator set refers to a collection of internal factors that affect the computing power load of the distribution network; based on the predetermined indicator set, collecting operating data of the distribution network in a digital twin distribution network model to obtain distribution network operating status information; acquiring the distribution network topology of the distribution network; performing power flow calculation on the distribution network in combination with the distribution network operating status information and the distribution network topology to obtain distribution network power flow information; visually displaying the distribution network power flow information according to the digital twin technology, and obtaining a target distribution abnormal node; and adjusting the load of the distribution network by starting a target load balancer at the target distribution abnormal node.

[0006] Another aspect disclosed in the present application provides a computing power load balancing device, which includes: a predetermined indicator reading unit, used to read a predetermined indicator set, wherein the predetermined indicator set refers to a collection of internal factors that affect the computing power load of the distribution network; an operation data collection unit, used to collect operation data of the distribution network in a digital twin distribution network model based on the predetermined indicator set, and obtain distribution network operation status information; a topology acquisition unit, used to obtain the distribution network topology of the distribution network; a flow calculation unit, used to perform flow calculation on the distribution network in combination with the distribution network operation status information and the distribution network topology, and obtain distribution network flow information; a visualization unit, used to visualize the distribution network flow information according to the digital twin technology, and obtain a target distribution abnormal node; a load adjustment unit, used to adjust the load of the distribution network by starting a target load balancer at the target distribution abnormal node.

[0007] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0008] The method of reading a predetermined indicator set is adopted to provide indicator information for obtaining the operating status information of the distribution network; data collection is carried out in the digital twin distribution network model, and the real-time operating status information of the distribution network is obtained according to the predetermined indicator set, laying a data foundation for power flow calculation; the topological structure of the distribution network is obtained to provide network structure information for power flow calculation; power flow calculation is carried out according to the collected operating status information and topological structure, and the load distribution and power flow parameters of the distribution network are analyzed to find areas with uneven load distribution; digital twin technology is used for visualization to intuitively display the load heat map and key abnormal nodes of the distribution network, providing a basis for load adjustment; targeted load adjustment is carried out through the target load balancer, and the computing power load of the abnormal nodes is optimized and adjusted to achieve a technical solution of balanced load distribution in the entire network, which solves the technical problem of uneven distribution of computing power load in the existing distribution network, and achieves the technical effect of optimizing the configuration of the computing power load of the distribution network and realizing energy conservation and emission reduction.

[0009] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Figure 1 A flow chart of a computing power load balancing method is provided for an embodiment of the present application;

[0011] Figure 2 A schematic diagram of a flow chart of obtaining power distribution network flow information in a computing power load balancing method is provided for an embodiment of the present application;

[0012] Figure 3 A structural schematic diagram of a computing power load balancing device is provided for an embodiment of the present application.

[0013] Explanation of the reference numerals: predetermined index reading unit 11, operation data collecting unit 12, topology structure acquiring unit 13, power flow calculating unit 14, visualization display unit 15, load adjustment unit 16. DETAILED DESCRIPTION

[0014] The overall idea of ​​the technical solution provided by this application is as follows:

[0015] The embodiments of the present application provide a computing power load balancing method and device. First, the relevant characteristic indicators of the distribution network are collected, and a high-precision distribution network model is established to grasp the real-time operating status data of the distribution network. Then, on the basis of obtaining the topological structure of the distribution network, the flow calculation and visual analysis are performed to find out the key areas that cause the uneven distribution of computing power load. Subsequently, through the target load balancer, the refined control of the key load nodes is realized, and the load is reasonably distributed to achieve the optimization effect of the computing power load balancing of the whole network, thereby effectively improving the computing power load balancing of the distribution network and realizing the optimal configuration of the load distribution while ensuring the safe operation of the distribution network.

[0016] After introducing the basic principles of the present application, various non-limiting implementation methods of the present application will be specifically described below in conjunction with the drawings in the specification.

[0017] Embodiment 1

[0018] like Figure 1 As shown, an embodiment of the present application provides a computing power load balancing method, which is applied to a computing power load balancing device that is communicatively connected to a distribution network.

[0019] Computing load balancing methods include:

[0020] Reading a predetermined indicator set, wherein the predetermined indicator set refers to a set of internal factors that affect the computing power load of the distribution network;

[0021] Furthermore, the predetermined indicator set includes voltage, current, power factor and load type.

[0022] In the embodiment of the present application, the predetermined indicator set refers to a set of key parameters that affect the computing load of the distribution network, including voltage, current, power factor and load type. Among them, voltage is the voltage level in the distribution network, and voltage fluctuations will affect the load; current is the current size in the distribution network, and excessive current will cause line overload; power factor is a parameter that reflects the nature of the load, and different types of loads have different power factors; load type refers to the type of load, such as lighting load, power load, etc., and different loads have different effects on the distribution network.

[0023] First, the voltage, current, power factor and load type indicators are pre-set as a predetermined indicator set, and the collection range and frequency of each indicator are determined. Then, the predetermined indicator set is stored in the computing power load balancing device. When performing algorithm load balancing, the predetermined indicator set is read first to provide direction for the collection of distribution network operation data.

[0024] Based on the predetermined indicator set, operating data of the distribution network is collected in the digital twin distribution network model to obtain operating status information of the distribution network;

[0025] In the embodiment of the present application, the digital twin distribution network model is a high-precision virtual distribution network established by digital twin technology. After reading the predetermined indicator set, the operation data of the distribution network is collected on the digital twin distribution network model to obtain the real-time operation status of the distribution network.

[0026] Specifically, the parameters of the predetermined indicators, such as voltage and current, also have their corresponding parameters in the digital twin distribution network model. By setting data collection points in the digital twin distribution network model, real-time data of each predetermined indicator can be obtained at each collection point. The actual operating data of these collected predetermined indicators can accurately reflect the real-time operating status of the distribution network, that is, the distribution network operating status information.

[0027] By using the digital twin model to obtain the real-time operation data of the distribution network, the operation status information of the distribution network can be obtained, providing data support for subsequent abnormal node detection and load balancing.

[0028] Acquire a distribution network topology structure of the distribution network;

[0029] In the embodiment of the present application, the distribution network topology is a structural schematic diagram of each node in the distribution network and their connection relationships, including the power flow path and method of each device in the distribution network.

[0030] Specifically, the topological connection status of the distribution network can be obtained by querying the design drawings of the distribution network. Alternatively, probes can be set in the digital twin distribution network model to obtain data from each line and node to infer the topological structure of the distribution network. This provides network structure support for power flow calculation and abnormal node judgment, making the analysis results more accurate.

[0031] Combining the distribution network operation status information with the distribution network topology structure, a power flow calculation is performed on the distribution network to obtain the distribution network power flow information;

[0032] In the embodiment of the present application, power flow calculation refers to the process of simulating the flow of power in the distribution network, and performing power flow calculation on the distribution network based on the obtained distribution network operation status information and distribution network topology structure to obtain the distribution network power flow information.

[0033] First, the obtained distribution network topology is imported into the power flow calculation software to establish the basic calculation model of the distribution network. Secondly, the voltage, current, load power and other parameters obtained from the distribution network operation status information are set in the basic calculation model. Then, based on the constructed basic calculation model, the power flow calculation program is run, and calculation methods such as load flow and voltage iteration are selected to obtain the active and reactive power flow of each line and the voltage, phase angle and other power flow calculation results of each node. Subsequently, according to factors such as line load rate and node voltage, the power flow calculation results are analyzed to predict the operation status of the distribution network, find abnormal conditions such as voltage overlimit and line overload, and obtain the power flow information of the distribution network to provide support for subsequent balanced control.

[0034] Perform a visual display of the power flow information of the distribution network according to the digital twin technology, and obtain the target power distribution abnormal node;

[0035] In an embodiment of the present application, based on a digital twin distribution network model, the distribution network flow information is imported into the digital twin distribution network model through digital twin technology, so that the power flow parameters and node voltage parameters of each line in the distribution network are mapped to the digital twin distribution network model, and different colors are used to represent the line load rate and the node voltage, so as to realize the visualization of the distribution network flow information.

[0036] By visualizing the power flow information of the distribution network, we can intuitively see the voltage problem nodes and overloaded lines in the distribution network, so as to determine the nodes with abnormalities in the power flow calculation, that is, the target distribution abnormality nodes, which are used for targeted subsequent load balancing to solve the problem of uneven computing load in the distribution network.

[0037] The load of the distribution network is adjusted by starting the target load balancer at the target power distribution abnormality node.

[0038] In the embodiment of the present application, the target load balancer refers to a control device set at the target abnormal node of the distribution network for achieving load balancing. After obtaining the target abnormal node of the distribution network, the target load balancer set at the node can be started to adjust the load of the distribution network.

[0039] Specifically, the target load balancer receives the status of the corresponding distribution node in real time and detects whether the node is a target distribution abnormal node. When the corresponding node is detected as a target distribution abnormal node, the target load balancer schedules the controllable load at the target distribution abnormal node according to the preset balancing control strategy, such as transferring the load to a light load, thereby realizing the load adjustment of the distribution network.

[0040] Through the automated and intelligent adjustment of the target load balancer, rapid response and balanced control of the load at the target power distribution abnormal node can be achieved, solving the load imbalance problem of the distribution network and improving the stability and efficiency of the power grid.

[0041] Further, such as Figure 2 As shown, the embodiment of the present application also includes:

[0042] Extracting first operating status information of a first distribution branch from the operating status information of the distribution network;

[0043] Wherein, the first operating state information includes a first voltage, a first current, a first power factor and a first load type;

[0044] Calling a load prediction fusion model to analyze the first operating state information to obtain a first predicted load of the first distribution branch;

[0045] The first predicted load is added to the power distribution network flow information.

[0046] In a preferred embodiment, in order to accurately calculate the load flow of the distribution network and obtain the flow information of the distribution network, it is necessary to consider the specific operating status of each branch in the distribution network. First, from the collected operating status information of the entire distribution network, determine and extract the operating parameters related to the first distribution branch as the first operating status information of the first distribution branch, including the first voltage, the first current, the first power factor and the first load type. Among them, the first voltage is the voltage magnitude parameter of the node on the first distribution branch; the first current is the current parameter of the line on the first distribution branch; the first power factor is the load property parameter on the first distribution branch; the first load type is the load category on the first distribution branch. By extracting the first operating status information, the actual operating status of the branch is accurately reflected, providing data support for obtaining the first predicted load.

[0047] Then, the first predicted load of the first distribution branch is added to the distribution network flow information. When the distribution network flow information is visualized, the first predicted load will participate in the power balance and voltage calculation process of the entire distribution network together with the predicted loads of other distribution branches, providing support for obtaining the target distribution abnormal node.

[0048] Furthermore, the embodiment of the present application also includes:

[0049] Dividing the data group constructed based on the historical load records into a training data group and a test data group, wherein the training data group includes N training data subgroups, where N is an integer greater than 1;

[0050] Using the neural network principle, supervised learning is performed on the N data groups in sequence to obtain N initial prediction models respectively;

[0051] Performing prediction analysis on the test data group using the N initial prediction models in sequence to obtain N prediction results;

[0052] Forming a target training data set based on the N prediction results and the load value in the test data set;

[0053] Performing supervised learning on the target training data set using the neural network principle to obtain a target meta-prediction model;

[0054] The load prediction fusion model is formed based on the N initial prediction models and the target meta-prediction model.

[0055] In a preferred embodiment, in order to build an accurate and reliable load prediction fusion model, first, a sufficient data set is collected for training. The data set is constructed based on historical load records, including load operation records of the distribution network over a period of time in the past. Secondly, the data set is divided into a training data set and a test data set, wherein the training data set is used for training the prediction model, and the group is divided into N smaller training data groups, where N is an integer greater than 1. Then, N training data groups are extracted respectively, and each training data group is used to train an initial prediction model separately. The training process follows the back propagation algorithm of the neural network, and the network weights are iteratively adjusted to gradually fit the training data, so that the network model can effectively predict the load of the input features, and finally obtain N initial prediction models.

[0056] Subsequently, N initial prediction models are used to predict the test data set, that is, the features of the test data are input, and the predicted load value results are output through model calculation. Repeat this process to obtain N prediction results, reflecting the generalization prediction ability of each initial prediction model for new data. Next, the prediction results of each initial prediction model for the same test data are aligned with the actual load value to form a new data set containing multiple sets of prediction features and labels, and the target training data set is obtained. Furthermore, the target training data set is used as input, the expected output is set, and the network parameters are adjusted through the back propagation algorithm to gradually train the target meta-prediction model that integrates the advantages of the initial model. The target meta-prediction model comprehensively learns the characteristics of multiple initial models and further enhances the adaptability to the actual load situation. Finally, the N initial prediction models obtained are weighted fused and integrated with the trained target meta-prediction model to improve the generalization ability and robustness of the final prediction, and to construct the required load prediction fusion model. The load prediction fusion model comprehensively combines the advantages of each initial model and the target meta-model on different training data, and can achieve more stable and accurate load prediction.

[0057] Furthermore, the embodiment of the present application also includes:

[0058] Extracting a first historical record from the historical load records, where the first historical record is a load record of a first historical distribution network;

[0059] The first operating status information and the first historical load of the first historical distribution branch in the first historical power distribution network are added to the data group as a first training data group.

[0060] In a preferred embodiment, first, a historical load record is extracted from the collected historical load records containing multiple time periods as a first historical record, and the record corresponds to the past load data of a historical distribution network. Then, the operation status information and the corresponding historical load data of the first historical distribution branch are obtained from the first historical record, that is, the first operation status information and the first historical load, which are constructed as a first training data group and added to the data group.

[0061] By extracting the real historical load records of the distribution network as the data group, the training data can be made more targeted and representative, and the accuracy of the established load prediction model can be improved.

[0062] Furthermore, the embodiment of the present application also includes:

[0063] Reading a predetermined environmental indicator set, wherein the predetermined environmental indicator set refers to a set of environmental factors that affect the computing load of the distribution network, and the predetermined environmental indicator set includes temperature, humidity, air pressure, and dust concentration;

[0064] Dynamically monitoring and obtaining distribution network environmental information based on the predetermined environmental indicator set;

[0065] Weighting the first temperature, the first humidity, the first air pressure and the first dust concentration in the distribution network environment information and calculating a first environmental factor;

[0066] The first predicted load is adjusted by using the first environmental factor as a first weight coefficient of the first predicted load.

[0067] In a preferred embodiment, in order to more comprehensively and accurately predict the load of the distribution network, it is necessary to consider the influence of environmental factors. First, a set of environmental indicators that affect the computing load of the distribution network is predetermined. The indicator set refers to a collection of environmental factors that affect the computing load of the distribution network, including temperature, humidity, air pressure and dust concentration, which can help the model to more comprehensively analyze the impact of the environment on the load and improve the accuracy of the prediction results. Then, based on the determined set of environmental indicators, the real-time distribution network environmental information is dynamically monitored and obtained through the sensing equipment set in the distribution network area. For example, the temperature, humidity, air pressure and dust concentration are detected by temperature and humidity sensors, barometers, smoke sensors and other equipment, and the environmental parameter data is transmitted in real time to obtain the distribution network environmental information.

[0068] Subsequently, different weights are set according to the different degrees of influence of parameters such as temperature and humidity on the distribution network, and the first temperature, first humidity, first air pressure and first dust concentration corresponding to the first predicted load of the distribution network environmental information obtained by monitoring are extracted, and weighted calculation is performed to synthesize the first environmental factor. The environmental factor quantifies the influence of the environmental state on the first predicted load, providing an important reference for considering the environmental impact to improve the prediction accuracy. Afterwards, the first environmental factor is used as the correction coefficient of the first predicted load and numerically adjusted. The larger the environmental factor, the greater the impact of the environment on the load, and the larger the correction coefficient, and the predicted load is further increased. By introducing environmental factors for weight adjustment, the load prediction value can better reflect the impact of the environmental state, improve the accuracy and reliability of the prediction results, and thus use the adjusted more accurate load prediction results to calculate the power grid flow, which is helpful for the reasonable planning of the distribution network.

[0069] Furthermore, the embodiment of the present application also includes:

[0070] If the first predicted load does not meet a predetermined load threshold, issuing a first marking instruction;

[0071] Marking the first power distribution branch as abnormal according to the first marking instruction, and adding it to a power distribution abnormality list;

[0072] A branch is randomly extracted from the power distribution abnormality list and recorded as the target power distribution abnormality node.

[0073] In a preferred embodiment, the abnormality of the distribution branch is judged according to the load prediction value. First, a predetermined load threshold is used as a standard for judging whether the distribution branch is abnormal. Then, the first predicted load obtained is compared with the predetermined load threshold. If the first predicted load is too large or too small, that is, it does not meet the predetermined threshold requirement, the device automatically issues a first marking instruction. The marking instruction indicates that the predicted load of the first distribution branch is abnormal and there may be a problem. Secondly, after the device issues the first marking instruction, the first distribution branch is marked abnormally according to the marking instruction. For example, the abnormal condition of the first distribution branch is indicated by a significant mark on the digital twin distribution network model, such as marking the branch in red. At the same time, the information of the first distribution branch marked with an abnormality is added to the distribution abnormality list and recorded to provide the target branch information for abnormal processing. Then, a distribution branch is selected from the abnormal list with equal probability as the target abnormal node by random extraction, so as to avoid always giving priority to certain branches and achieve balanced processing of each abnormal branch.

[0074] In summary, the computing power load balancing method provided in the embodiments of the present application has the following technical effects:

[0075] Read the predetermined indicator set, which refers to the set of internal factors that affect the computing power load of the distribution network, and provide indicator information for obtaining the operating status information of the distribution network. Based on the predetermined indicator set, the operating data of the distribution network is collected in the digital twin distribution network model to obtain the operating status information of the distribution network, laying a data foundation for the flow calculation. Obtain the distribution network topology of the distribution network to provide support for obtaining the distribution network flow information. Combine the distribution network operating status information and the distribution network topology to calculate the flow of the distribution network, obtain the distribution network flow information, find the load imbalance area, and provide support for locating the target distribution abnormal node. Visualize the distribution network flow information according to the digital twin technology, and obtain the target distribution abnormal node to provide support for load adjustment. By starting the target load balancer at the target distribution abnormal node to adjust the load of the distribution network, the optimal configuration of the distribution network computing power load is achieved, and energy conservation and emission reduction are achieved at the same time.

[0076] Embodiment 2

[0077] Based on the same inventive concept as the computing power load balancing method in the above-mentioned embodiment, Figure 3 As shown, an embodiment of the present application provides a computing power load balancing device, which is communicatively connected to a distribution network and includes:

[0078] A predetermined indicator reading unit 11, used to read a predetermined indicator set, wherein the predetermined indicator set refers to a set of internal factors that affect the computing power load of the distribution network;

[0079] An operation data mobile unit 12 is used to collect the operation data of the distribution network in the digital twin distribution network model based on the predetermined indicator set to obtain the operation status information of the distribution network;

[0080] A topology structure acquisition unit 13, used to acquire a distribution network topology structure of the distribution network;

[0081] A power flow calculation unit 14 is used to perform power flow calculation on the distribution network in combination with the distribution network operation status information and the distribution network topology structure to obtain distribution network power flow information;

[0082] A visualization display unit 15, used for visualizing the power distribution network flow information according to the digital twin technology, and obtaining the target power distribution abnormal node;

[0083] The load adjustment unit 16 is used to adjust the load of the distribution network by starting the target load balancer at the target power distribution abnormality node.

[0084] Furthermore, the predetermined indicator reading unit 11 includes the following contents:

[0085] The predetermined indicator set includes voltage, current, power factor and load type.

[0086] Furthermore, the power flow calculation unit 14 includes the following execution steps:

[0087] Extracting first operating status information of a first distribution branch from the operating status information of the distribution network;

[0088] Wherein, the first operating state information includes a first voltage, a first current, a first power factor and a first load type;

[0089] Calling a load prediction fusion model to analyze the first operating state information to obtain a first predicted load of the first distribution branch;

[0090] The first predicted load is added to the power distribution network flow information.

[0091] Furthermore, the power flow calculation unit 14 further includes the following execution steps:

[0092] Dividing the data group constructed based on the historical load records into a training data group and a test data group, wherein the training data group includes N training data subgroups, where N is an integer greater than 1;

[0093] Using the neural network principle, supervised learning is performed on the N data groups in sequence to obtain N initial prediction models respectively;

[0094] Performing prediction analysis on the test data group using the N initial prediction models in sequence to obtain N prediction results;

[0095] Forming a target training data set based on the N prediction results and the load value in the test data set;

[0096] Performing supervised learning on the target training data set using the neural network principle to obtain a target meta-prediction model;

[0097] The load prediction fusion model is formed based on the N initial prediction models and the target meta-prediction model.

[0098] Furthermore, the power flow calculation unit 14 further includes the following execution steps:

[0099] Extracting a first historical record from the historical load records, where the first historical record is a load record of a first historical distribution network;

[0100] The first operating status information and the first historical load of the first historical distribution branch in the first historical power distribution network are added to the data group as a first training data group.

[0101] Furthermore, the power flow calculation unit 14 further includes the following execution steps:

[0102] Reading a predetermined environmental indicator set, wherein the predetermined environmental indicator set refers to a set of environmental factors that affect the computing load of the distribution network, and the predetermined environmental indicator set includes temperature, humidity, air pressure, and dust concentration;

[0103] Dynamically monitoring and obtaining distribution network environmental information based on the predetermined environmental indicator set;

[0104] Weighting the first temperature, the first humidity, the first air pressure and the first dust concentration in the distribution network environment information and calculating a first environmental factor;

[0105] The first predicted load is adjusted by using the first environmental factor as a first weight coefficient of the first predicted load.

[0106] Furthermore, the visual display unit 15 includes the following execution steps:

[0107] If the first predicted load does not meet a predetermined load threshold, issuing a first marking instruction;

[0108] Marking the first power distribution branch as abnormal according to the first marking instruction, and adding it to a power distribution abnormality list;

[0109] A branch is randomly extracted from the power distribution abnormality list and recorded as the target power distribution abnormality node.

[0110] Any step of the method described above can be stored as a computer instruction or program in an unlimited computer memory, and can be called and recognized by an unlimited computer processor to implement any method in the embodiments of the present application, without any unnecessary restrictions.

[0111] Furthermore, the first or second mentioned above may not only represent an order relationship, but may also represent a specific concept, and / or refer to multiple elements that can be selected individually or in full. Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the present application and its equivalents, the present application intends to include these modifications and variations.

Claims

1. A computing power load balancing method, characterized in that: The method is applied to a computing power load balancing device, the device is communicatively connected to a power distribution network, and the method includes: Reading a predetermined indicator set, wherein the predetermined indicator set refers to a set of internal factors that affect the computing power load of the distribution network; Based on the predetermined indicator set, operating data of the distribution network is collected in the digital twin distribution network model to obtain operating status information of the distribution network; Acquire a distribution network topology structure of the distribution network; Combining the distribution network operation status information with the distribution network topology structure, a power flow calculation is performed on the distribution network to obtain the distribution network power flow information; Perform a visual display of the power flow information of the distribution network according to the digital twin technology, and obtain the target power distribution abnormal node; Performing load adjustment on the distribution network by starting a target load balancer at the target power distribution abnormality node; The step of performing power flow calculation on the distribution network in combination with the distribution network operation status information and the distribution network topology structure to obtain power flow information of the distribution network includes: Extracting first operating status information of a first distribution branch from the operating status information of the distribution network; Wherein, the first operating state information includes a first voltage, a first current, a first power factor and a first load type; Calling a load prediction fusion model to analyze the first operating state information to obtain a first predicted load of the first distribution branch; adding the first predicted load to the power distribution network flow information; Before calling the load prediction fusion model to analyze the first operation status information, the method includes: Dividing the data group constructed based on the historical load records into a training data group and a test data group, wherein the training data group includes N training data subgroups, where N is an integer greater than 1; Using the neural network principle, supervised learning is performed on the N data groups in sequence to obtain N initial prediction models respectively; Performing prediction analysis on the test data group using the N initial prediction models in sequence to obtain N prediction results; Forming a target training data set based on the N prediction results and the load value in the test data set; Performing supervised learning on the target training data set using the neural network principle to obtain a target meta-prediction model; The load prediction fusion model is formed according to the N initial prediction models and the target meta-prediction model; Before adding the first predicted load to the power distribution network flow information, the method further includes: Reading a predetermined environmental indicator set, wherein the predetermined environmental indicator set refers to a set of environmental factors that affect the computing load of the distribution network, and the predetermined environmental indicator set includes temperature, humidity, air pressure, and dust concentration; Dynamically monitoring and obtaining distribution network environmental information based on the predetermined environmental indicator set; Weighting the first temperature, the first humidity, the first air pressure and the first dust concentration in the distribution network environment information and calculating a first environmental factor; The first predicted load is adjusted by using the first environmental factor as a first weight coefficient of the first predicted load.

2. The method according to claim 1, characterized in that: The predetermined indicator set includes voltage, current, power factor and load type.

3. The method according to claim 1, characterized in that: Before dividing the data set constructed based on the historical load records into a training data set and a test data set, the method includes: Extracting a first historical record from the historical load records, where the first historical record is a load record of a first historical distribution network; The first operating status information and the first historical load of the first historical distribution branch in the first historical power distribution network are added to the data group as a first training data group.

4. The method according to claim 1, characterized in that: The visual display of the power flow information of the distribution network according to the digital twin technology and obtaining the target power distribution abnormal node include: If the first predicted load does not meet a predetermined load threshold, issuing a first marking instruction; Marking the first power distribution branch as abnormal according to the first marking instruction, and adding it to a power distribution abnormality list; A branch is randomly extracted from the power distribution abnormality list and recorded as the target power distribution abnormality node.

5. A computing power load balancing device, characterized in that: Used to implement the computing power load balancing method according to any one of claims 1 to 4, the device is communicatively connected to a distribution network, and the device comprises: A predetermined indicator reading unit, the predetermined indicator reading unit is used to read a predetermined indicator set, the predetermined indicator set refers to a set of internal factors that affect the computing power load of the distribution network; An operation data mobile phone unit, the operation data mobile phone unit is used to collect the operation data of the distribution network in the digital twin distribution network model based on the predetermined indicator set to obtain the distribution network operation status information; A topology acquisition unit, wherein the topology acquisition unit is used to acquire a distribution network topology of the distribution network; A power flow calculation unit, the power flow calculation unit is used to perform power flow calculation on the distribution network in combination with the distribution network operation status information and the distribution network topology structure to obtain distribution network power flow information; A visualization display unit, wherein the visualization display unit is used to visualize the power flow information of the distribution network according to the digital twin technology and obtain the target power distribution abnormal node; A load adjustment unit, wherein the load adjustment unit is used to adjust the load of the distribution network by starting a target load balancer at the target power distribution abnormality node.

Images