Full-flow through system of power distribution network

By building machine learning models in the distribution network to predict node faults and electricity demand, and generating distribution plans, the problems of lagging fault identification and insufficient perception of electricity demand are solved, thereby improving the automation level and operating efficiency of the distribution network.

CN118944287APending Publication Date: 2024-11-12STATE GRID FUJIAN POWER ELECTRIC CO ECONOMIC RESEARCH INSTITUTE +1
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202410984916.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-22
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

In existing technologies, fault identification in distribution networks is lagging behind, and there is a lack of integration between end-user electricity demand perception and dispatch, resulting in low stability and efficiency of the power system.

Method used

By employing machine learning models combined with a perception module, a model building module, and a state analysis module, and by acquiring operational information from distribution network nodes and end users, node fault prediction models and electricity demand prediction models are constructed to generate distribution plans for power equipment scheduling.

Benefits of technology

It enables real-time fault prediction and electricity demand prediction in the distribution network, ensuring supply and demand balance and improving the automation level and operating efficiency of the distribution network.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118944287A_ABST
    Figure CN118944287A_ABST
Patent Text Reader

Abstract

The invention relates to a power distribution network full-flow through system, and the system comprises a sensing module which is used for obtaining the operation information of each node in a target power distribution network and the power utilization information of each terminal user; the model construction module is used for training a first machine learning model based on the historical operation information of each node to obtain a node fault prediction model; training a second machine learning model based on the historical power consumption data of each terminal user to obtain a terminal power consumption demand prediction model; the state analysis module is used for acquiring real-time operation parameter data of each current node and inputting the real-time operation parameter data into the node fault prediction model to obtain a fault prediction result of each node; obtaining a power consumption demand prediction result of each user at a future moment through the terminal power consumption demand prediction model; and the scheduling module generates a target power distribution network power distribution plan based on the fault prediction result of each node and the power demand prediction result of each user at the future moment, and performs power equipment scheduling based on the target power distribution network power distribution plan.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to a full-process through-system for a distribution network, and belongs to the technical field of distribution network automation. Background Art

[0002] As a vital component of the power system, the distribution network undertakes the important task of transmitting electric energy from the transmission network to the user's terminal. However, in the actual operation process, there are often many defects in the entire process of the distribution network, which not only affects the stable operation of the power system, but also may lead to a decline in power quality and an increase in safety hazards. Therefore, in-depth analysis of the defects of the entire process of the distribution network and taking corresponding measures to improve them are of great significance to improving the overall performance and reliability of the power system.

[0003] The Chinese invention patent with patent number CN110474334A discloses a method for controlling the whole process of distribution network dispatching and operation. In the distribution automation system, a closed-loop control process for fault handling covering all types of faults in the distribution network and running through the whole process of fault handling is set; faults are automatically analyzed and identified, and trip points and fault intervals are located; fault isolation and power transfer schemes are generated and corresponding operation sequences are provided; information interaction channels with OMS are connected, and power outage information is released; a one-button sequential control function is provided for the scheme; the fault point is manually reset, and the fault isolation and power transfer scheme is regenerated; the secondary fault is merged with the primary fault handling process; the fault and emergency repair content are recorded and synchronized to the OMS dispatch log, and the selection of security measures information is supported; remote control failures, missed reports, and false reports are identified, and the list of primary and secondary equipment involved is pushed to the defect control process; a post-analysis report is generated; and a historical data query tool is provided. This method is conducive to improving the efficiency of distribution network fault handling and the level of distribution network automation.

[0004] However, the problem with the above patent is that the fault identification is delayed and there is a lack of a process that connects the power demand perception part of the terminal users of the distribution network and the dispatching part of the distribution network. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention proposes a distribution network full-process through-system.

[0006] The technical solution of the present invention is as follows:

[0007] On the one hand, the present invention proposes a distribution network full-process through-system, comprising:

[0008] A sensing module is used to obtain the operating information of each node in the target distribution network and the power consumption information of each end user, wherein the operating information of each node includes operating parameter data and operating fault data;

[0009] A model building module, which builds a first training set based on the historical operating parameter data and historical operating fault information of each node, trains a first machine learning model through the first training set, and obtains a node fault prediction model; builds a second data set based on the historical power consumption data of each terminal user, and trains a second machine learning model through the second data set to obtain a terminal power demand prediction model;

[0010] The status analysis module is used to obtain the real-time operating parameter data of each node at present, input it into the node fault prediction model, and obtain the fault prediction results of each node; and obtain the power demand prediction results of each user at the future time through the terminal power demand prediction model;

[0011] The scheduling module generates a target distribution network distribution plan based on the fault prediction results of each node and the power demand prediction results of each user in the future, and performs power equipment scheduling based on the target distribution network distribution plan.

[0012] As a preferred implementation, the model building module builds a first training set based on the historical operating parameter data and historical operating fault information of each node, and trains the first machine learning model through the first training set to obtain the node fault prediction model in the following steps:

[0013] Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results;

[0014] Perform sample multiplication on the fault operation parameter data set so that the sample size of the fault operation parameter data set is equal to that of the normal operation parameter data set;

[0015] Add labels to sample data in the normal operating parameter data set and the faulty operating parameter data set after sample proliferation and put them into a first training set, wherein the labels are used to indicate the fault type of the sample data;

[0016] The sample data in the first training set is used as input and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The parameters in the first machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained node fault prediction model.

[0017] As a preferred implementation, the model building module builds a second data set based on the historical electricity consumption data of each terminal user, and trains a second machine learning model through the second data set to obtain the terminal electricity demand prediction model in the following steps:

[0018] Based on the historical electricity consumption data of each user, a time series data graph of electricity consumption of the corresponding user is constructed;

[0019] The electricity consumption time series data graph of each user is divided into equally spaced time windows, and the interval time and length of each time window are the same; and each time window is divided into two small time windows, including a current time window with a time sequence of L1 and a future time window with a time sequence of L2;

[0020] The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained;

[0021] A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set;

[0022] The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

[0023] As a preferred implementation, the first machine learning model adopts an RNN neural network model, and the second machine learning model adopts an LSTM neural network model.

[0024] On the other hand, the present invention also proposes a method for connecting the entire distribution network, which is characterized by comprising the following steps:

[0025] Obtaining the operating information of each node in the target distribution network and the power consumption information of each end user, wherein the operating information of each node includes operating parameter data and operating fault data;

[0026] A first training set is constructed based on the historical operating parameter data and historical operating fault information of each node, and a first machine learning model is trained by the first training set to obtain a node fault prediction model; a second data set is constructed based on the historical power consumption data of each terminal user, and a second machine learning model is trained by the second data set to obtain a terminal power demand prediction model;

[0027] Obtain the real-time operating parameter data of each node at present, input it into the node fault prediction model, and obtain the fault prediction results of each node; and obtain the power demand prediction results of each user at the future time through the terminal power demand prediction model;

[0028] The target distribution network distribution plan is generated based on the fault prediction results of each node and the power demand prediction results of each user in the future, and the power equipment is dispatched based on the target distribution network distribution plan.

[0029] As a preferred implementation, the steps of constructing a first training set based on the historical operating parameter data and historical operating fault information of each node, training a first machine learning model through the first training set, and obtaining a node fault prediction model are specifically as follows:

[0030] Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results;

[0031] Perform sample multiplication on the fault operation parameter data set so that the sample size of the fault operation parameter data set is equal to that of the normal operation parameter data set;

[0032] Add labels to sample data in the normal operating parameter data set and the faulty operating parameter data set after sample proliferation and put them into a first training set, wherein the labels are used to indicate the fault type of the sample data;

[0033] The sample data in the first training set is used as input and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The parameters in the first machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained node fault prediction model.

[0034] As a preferred implementation, the steps of constructing a second data set based on the historical electricity consumption data of each terminal user, training a second machine learning model through the second data set, and obtaining a terminal electricity demand prediction model are specifically as follows:

[0035] Based on the historical electricity consumption data of each user, a time series data graph of electricity consumption of the corresponding user is constructed;

[0036] The electricity consumption time series data graph of each user is divided into equally spaced time windows, and the interval time and length of each time window are the same; and each time window is divided into two small time windows, including a current time window with a time sequence of L1 and a future time window with a time sequence of L2;

[0037] The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained;

[0038] A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set;

[0039] The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

[0040] As a preferred implementation, the first machine learning model adopts an RNN neural network model, and the second machine learning model adopts an LSTM neural network model.

[0041] On the other hand, the present invention further proposes an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein when the processor executes the program, the method for connecting the entire distribution network as described in any embodiment of the present invention is implemented.

[0042] On the other hand, the present invention further proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the distribution network full-process connection method as described in any embodiment of the present invention.

[0043] The present invention has the following beneficial effects:

[0044] The present invention obtains the operation information of distribution network nodes and the electricity consumption data of end users, combines artificial intelligence technology, analyzes and mines a large amount of historical data and real-time data, and gives the prediction results of node failure and user electricity demand. Therefore, a scheduling plan is generated according to the prediction results to ensure the supply and demand balance and efficient operation of the distribution network. It runs through the various processes of distribution network status perception, operation monitoring, distribution network scheduling, and terminal demand matching, and improves the automation level of the distribution network. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] Figure 1 This is a schematic diagram of the system structure of Embodiment 1 of the present invention;

[0046] Figure 2 This is a schematic diagram of the method flow of Embodiment 2 of the present invention. DETAILED DESCRIPTION

[0047] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0048] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0049] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0050] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0051] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0052] Embodiment 1:

[0053] See also Figure 1 This embodiment provides a distribution network full-process through-connection system, including:

[0054] The perception module is used to obtain the operation information of each node in the target distribution network and the power consumption information of each terminal user. The operation information of each node includes operation parameter data and operation fault data; the operation parameter data covers multiple dimensions of data such as node equipment voltage, current, power factor, operation time, load, temperature, pressure, etc., which can reflect the real-time operation status of each node in the distribution network. Operation fault data records detailed information such as the type of fault, occurrence time, and impact range of each node in the distribution network. By obtaining operation information, support is provided for subsequent fault prediction. The power consumption information of terminal users includes key data such as the user's power consumption, power consumption time, and power load. By collecting and analyzing these data, we can understand the user's power consumption habits and power demand, and provide support for subsequent prediction of the user's power demand.

[0055] The model building module builds a first training set based on the historical operating parameter data and historical operating fault information of each node, trains the first machine learning model through the first training set, and obtains a node fault prediction model; builds a second data set based on the historical electricity consumption data of each terminal user, trains the second machine learning model through the second data set, and obtains a terminal electricity demand prediction model; through the constructed node fault prediction model and terminal electricity demand prediction model, the distribution network node faults and the user's electricity demand can be predicted. In the process of building and training these machine learning models, some methods such as reinforcement learning can also be used to further improve the prediction accuracy and generalization ability of the model.

[0056] The state analysis module is used to obtain the real-time operating parameter data of each node at present, input it into the node fault prediction model, and obtain the fault prediction results of each node; and obtain the power demand prediction results of each user at the future time through the terminal power demand prediction model;

[0057] The scheduling module generates the target distribution network distribution plan based on the fault prediction results of each node and the power demand prediction results of each user in the future. In the process of generating the target distribution network distribution plan, firstly, according to the fault prediction results of each node, a corresponding fault response plan is formulated to ensure rapid response when a fault occurs and reduce the impact of the fault on the operation of the distribution network. At the same time, combined with the power demand prediction results of each user in the future, the power supply capacity of the distribution network is evaluated and the power supply strategy is determined. At the same time, the economy, safety and reliability of the distribution network are considered. Reasonable resource allocation and scheduling are carried out to improve the power supply efficiency of the distribution network, reduce energy loss, and ensure the stability of power supply quality.

[0058] After generating the target distribution network distribution plan, the power equipment dispatching work is further carried out. The power equipment dispatching work is specifically as follows:

[0059] According to the power demand forecast results, the power equipment is preliminarily classified and sorted. For high-demand areas, power supply is prioritized to ensure that the power demand of users is met. At the same time, according to the predicted results of possible failures at each node, a plan is formulated for possible failures.

[0060] In the process of equipment dispatching, we consider the power demand and fault plan to carry out reasonable dispatching to achieve load balancing of power equipment and reduce the risk of equipment failure caused by overload. At the same time, we fully consider the energy efficiency and environmental performance of power equipment, give priority to high-efficiency and low-emission equipment for dispatching, so as to achieve the goal of energy conservation and emission reduction.

[0061] Through the above-mentioned dispatching wages, the stable operation of the power system can be ensured, the electricity needs of users can be met, and the rational use of resources and cost optimization can be achieved.

[0062] As a preferred implementation of this embodiment, the model building module builds a first training set based on the historical operating parameter data and historical operating fault information of each node, and trains the first machine learning model through the first training set to obtain the node fault prediction model in the following steps:

[0063] Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results;

[0064] In order to ensure the accuracy and reliability of the analysis results, this embodiment performs sample proliferation on the fault operation parameter data set so that the sample size thereof is equal to the sample size of the normal operation parameter data set.

[0065] Since faults occur relatively rarely, the sample size of the fault operation parameter data set is often much smaller than that of the normal operation parameter data set. This will result in the failure operation parameter data set not having enough information to support comprehensive and accurate analysis during analysis. Therefore, it is necessary to multiply the samples of the fault operation parameter data set to increase its sample size, thereby improving the accuracy and reliability of the analysis. There are many methods for sample multiplication, such as statistical resampling methods, machine learning-based generative adversarial network (GAN) methods, etc. These methods can generate more similar samples based on the existing fault operation parameter data set, thereby increasing the size of the data set.

[0066] In this embodiment, a suitable method can be selected according to specific needs and conditions. For example, if the fault operation parameter data set has a relatively stable distribution characteristic, a statistically based resampling method can be used for sample proliferation. This method can generate new sample data that conforms to the distribution according to the distribution characteristics of the data set.

[0067] If the fault operation parameter data set has a more complex distribution feature, or if you want to generate more realistic and diverse sample data, you can use the generative adversarial network (GAN) method based on machine learning. GAN consists of two neural networks, the generator and the discriminator. Through mutual competition and collaboration, it can generate new samples with a high degree of similarity to real data.

[0068] In the process of sample proliferation, it is necessary to ensure that the generated sample data has similar characteristic distribution and statistical characteristics as the real data to ensure the accuracy of the analysis results. Secondly, attention should be paid to avoid the occurrence of overfitting, that is, the generated sample data should not be too close to the existing sample data, but should have a certain diversity and generalization ability.

[0069] The sample data in the normal operating parameter data set and the fault operating parameter data set after sample proliferation are labeled and put into the first training set. The label is used to indicate the fault type of the sample data. For example, for the fault operating parameter data set, a corresponding label can be assigned to each sample based on information such as the type, degree or location of the fault, to help the machine learning model learn and identify different fault types.

[0070] The sample data in the first training set is used as input, and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The loss function is used to measure the difference between the model prediction result and the true label. Common loss functions include cross entropy loss and mean square error. By calculating the loss, the prediction ability of the model at the current stage can be evaluated. The back propagation algorithm is used to adjust the parameters in the model. In the process of adjusting the parameters, we usually use optimization algorithms such as gradient descent and Adam optimizer to speed up the convergence speed and improve the performance of the model. By continuously iterating the training process, the model parameters are gradually optimized, so that the performance of the model in fault prediction is improved. When the model converges, that is, the loss reaches a relatively stable value, it is considered that the first machine learning model has been trained. At this point, we can apply the trained node fault prediction model to the actual scenario to predict node faults.

[0071] As a preferred implementation of this embodiment, the model building module builds a second data set based on the historical electricity consumption data of each terminal user, and trains a second machine learning model through the second data set to obtain the terminal electricity demand prediction model in the following steps:

[0072] Based on the historical electricity consumption data of each user, a time series data graph of the corresponding user's electricity consumption is constructed; a time series data graph is a graphical representation of the trend of data changes over time. For electricity use, a time series data graph can clearly reflect the user's electricity consumption in different time periods, including daily, weekly, monthly and even annual electricity consumption fluctuations. The electricity consumption time series data graph can intuitively show the user's electricity demand.

[0073] In the process of building a time series data graph of electricity consumption, we first need to collect the user's historical electricity consumption data. This data usually comes from the power company's recording system, which contains information such as the user's electricity consumption and electricity charges in different time periods. By sorting and analyzing this data, we can get detailed information about the user's electricity consumption, such as peak hours and low hours.

[0074] In order to deeply analyze the electricity consumption characteristics of each user, this embodiment divides the electricity consumption time series data graph into equally spaced time windows, and ensures that the interval time and length of each time window are the same, thereby ensuring the continuity and comparability of the data.

[0075] Specifically, the length of each time window is first determined. Then the entire electricity consumption time series data graph is evenly divided into several time windows according to this time length. Each time window contains electricity consumption data within a certain period of time, which reflects the user's electricity consumption behavior within this period of time.

[0076] Next, each time window is further divided into two small time windows. These two small time windows are adjacent in time series, but they differ in data content. Among them, the first small time window is the current time window, which is located in the first half of the entire time window and has a length of L1. This time window mainly focuses on the real-time electricity consumption of users in a specific time period and provides instant feedback on the user's electricity consumption behavior. The second small time window is the future time window, which is located in the second half of the entire time window and has a length of L2. The second small time window is used to predict the user's electricity demand in the next period of time. By comparing the data changes in the current time window and the future time window, we can more accurately predict the user's electricity consumption trend.

[0077] The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained;

[0078] A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set;

[0079] The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

[0080] As a preferred implementation, the first machine learning model adopts a recurrent neural network model (RNN), and the second machine learning model adopts a long short-term memory neural network model (LSTM).

[0081] Embodiment 2:

[0082] See also Figure 2 This embodiment proposes a method for connecting the entire distribution network, which is characterized by comprising the following steps:

[0083] S100, obtaining the operation information of each node in the target distribution network and the power consumption information of each terminal user, wherein the operation information of each node includes operation parameter data and operation fault data; the function of step S100 is used to realize the function of the perception module in the first embodiment, which will not be repeated here;

[0084] S200, constructing a first training set based on the historical operating parameter data and historical operating fault information of each node, training a first machine learning model through the first training set, and obtaining a node fault prediction model; constructing a second data set based on the historical power consumption data of each terminal user, and training a second machine learning model through the second data set to obtain a terminal power demand prediction model; the function of step S200 is used to implement the function of the perception module in embodiment 1, and will not be repeated here;

[0085] S300, obtaining the real-time operating parameter data of each node at present, inputting it into the node fault prediction model, and obtaining the fault prediction result of each node; and obtaining the power demand prediction result of each user at a future time through the terminal power demand prediction model; the function of step S300 is used to realize the function of the perception module in the first embodiment, and will not be repeated here;

[0086] S400, generate a target distribution network distribution plan based on the fault prediction results of each node and the power demand prediction results of each user in the future, and dispatch power equipment based on the target distribution network distribution plan; the function of step S400 is used to implement the function of the perception module in Example 1, which will not be repeated here.

[0087] As a preferred implementation of this embodiment, the steps of constructing a first training set based on the historical operating parameter data and historical operating fault information of each node, training a first machine learning model through the first training set, and obtaining a node fault prediction model are specifically as follows:

[0088] Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results;

[0089] Perform sample multiplication on the fault operation parameter data set so that the sample size of the fault operation parameter data set is equal to that of the normal operation parameter data set;

[0090] Add labels to sample data in the normal operating parameter data set and the faulty operating parameter data set after sample proliferation and put them into a first training set, wherein the labels are used to indicate the fault type of the sample data;

[0091] The sample data in the first training set is used as input and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The parameters in the first machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained node fault prediction model.

[0092] As a preferred implementation of this embodiment, the steps of constructing a second data set based on the historical electricity consumption data of each terminal user, training a second machine learning model through the second data set, and obtaining a terminal electricity demand prediction model are specifically as follows:

[0093] Based on the historical electricity consumption data of each user, a time series data graph of electricity consumption of the corresponding user is constructed;

[0094] The electricity consumption time series data graph of each user is divided into equally spaced time windows, and the interval time and length of each time window are the same; and each time window is divided into two small time windows, including a current time window with a time sequence of L1 and a future time window with a time sequence of L2;

[0095] The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained;

[0096] A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set;

[0097] The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

[0098] As a preferred implementation of this embodiment, the first machine learning model adopts an RNN neural network model, and the second machine learning model adopts an LSTM neural network model.

[0099] Embodiment three:

[0100] This embodiment proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the distribution network full-process connection method as described in any embodiment of the present invention is implemented.

[0101] Embodiment 4:

[0102] This embodiment provides a computer-readable storage medium on which a computer program is stored. When the program is executed by a processor, the method for connecting the entire distribution network as described in any embodiment of the present invention is implemented.

[0103] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0104] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0105] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0106] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0107] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A full-process access system for distribution networks, It is characterized in that include: A sensing module is used to obtain the operating information of each node in the target distribution network and the power consumption information of each end user, wherein the operating information of each node includes operating parameter data and operating fault data; A model building module, which builds a first training set based on the historical operating parameter data and historical operating fault information of each node, trains a first machine learning model through the first training set, and obtains a node fault prediction model; builds a second data set based on the historical power consumption data of each terminal user, and trains a second machine learning model through the second data set to obtain a terminal power demand prediction model; The status analysis module is used to obtain the real-time operating parameter data of each node and input it into the node fault prediction model to obtain the fault prediction results of each node; And obtain the power demand forecast results of each user at a future moment through the terminal power demand forecast model; The scheduling module generates a target distribution network distribution plan based on the fault prediction results of each node and the power demand prediction results of each user in the future, and performs power equipment scheduling based on the target distribution network distribution plan.

2. A distribution network full-process through-system according to claim 1, It is characterized in that The model building module builds a first training set based on the historical operating parameter data and historical operating fault information of each node, and trains a first machine learning model through the first training set to obtain a node fault prediction model in the following steps: Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results; Perform sample multiplication on the fault operation parameter data set so that the sample size of the fault operation parameter data set is equal to that of the normal operation parameter data set; Add labels to sample data in the normal operating parameter data set and the faulty operating parameter data set after sample proliferation and put them into a first training set, wherein the labels are used to indicate the fault type of the sample data; The sample data in the first training set is used as input and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The parameters in the first machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained node fault prediction model.

3. A distribution network full-process through-system according to claim 1, It is characterized in that The model building module builds a second data set based on the historical electricity consumption data of each terminal user, and trains a second machine learning model through the second data set to obtain a terminal electricity demand prediction model. Specifically, the steps are: Based on the historical electricity consumption data of each user, a time series data graph of electricity consumption of the corresponding user is constructed; The electricity consumption time series data graph of each user is divided into equally spaced time windows, and the interval time and length of each time window are the same; and each time window is divided into two small time windows, including a current time window with a time sequence of L1 and a future time window with a time sequence of L2; The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained; A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set; The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

4. A distribution network full-process through-system according to claim 1, Features: The first machine learning model adopts an RNN neural network model, and the second machine learning model adopts an LSTM neural network model.

5. A method for connecting the entire distribution network process. It is characterized in that The following steps are involved: Obtaining the operating information of each node in the target distribution network and the power consumption information of each end user, wherein the operating information of each node includes operating parameter data and operating fault data; A first training set is constructed based on the historical operating parameter data and historical operating fault information of each node, and a first machine learning model is trained by the first training set to obtain a node fault prediction model; a second data set is constructed based on the historical power consumption data of each terminal user, and a second machine learning model is trained by the second data set to obtain a terminal power demand prediction model; Obtain the real-time operating parameter data of each node, input it into the node fault prediction model, and obtain the fault prediction results of each node; And obtain the power demand forecast results of each user at a future moment through the terminal power demand forecast model; The target distribution network distribution plan is generated based on the fault prediction results of each node and the power demand prediction results of each user in the future, and the power equipment is dispatched based on the target distribution network distribution plan.

6. A method for connecting the entire distribution network according to claim 5, It is characterized in that The steps of constructing a first training set based on the historical operating parameter data and historical operating fault information of each node, training a first machine learning model through the first training set, and obtaining a node fault prediction model are specifically as follows: Based on the historical operation fault information of each node, the historical operation parameter data is classified into normal / faulty categories, and the historical operation fault information of each node is put into the normal operation parameter data set and the fault operation parameter data set according to the classification results; Perform sample multiplication on the fault operation parameter data set so that the sample size of the fault operation parameter data set is equal to that of the normal operation parameter data set; Add labels to sample data in the normal operating parameter data set and the faulty operating parameter data set after sample proliferation and put them into a first training set, wherein the labels are used to indicate the fault type of the sample data; The sample data in the first training set is used as input and the label is used as output to train the first machine learning model. The loss is calculated based on the fault type prediction vector output by the first machine learning model and the corresponding label. The parameters in the first machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained node fault prediction model.

7. A method for connecting the entire distribution network according to claim 5, It is characterized in that The steps of constructing a second data set based on the historical electricity consumption data of each terminal user, training a second machine learning model through the second data set, and obtaining a terminal electricity demand prediction model are specifically as follows: Based on the historical electricity consumption data of each user, a time series data graph of electricity consumption of the corresponding user is constructed; The electricity consumption time series data graph of each user is divided into equally spaced time windows, and the interval time and length of each time window are the same; and each time window is divided into two small time windows, including a current time window with a time sequence of L1 and a future time window with a time sequence of L2; The pre-trained CNN convolutional neural network model is used to extract features from the electricity consumption time series data graph of each user, and the feature vectors of the current time window and the future time window in each time window are obtained; A pair of feature vectors of the current time window and feature vectors of the future time window are taken as a training sample and put into the second training set; The feature vector of the current time window in the training sample is used as input, and the feature vector of the future time window is used as output to train the second machine learning model. The loss is calculated based on the feature vector of the future moment output by the second machine learning model and the feature vector of the corresponding future time window. The parameters in the second machine learning model are reversely adjusted according to the calculated loss until the model converges to obtain a trained terminal electricity demand prediction model.

8. A method for connecting the entire distribution network according to claim 5, Features: The first machine learning model adopts an RNN neural network model, and the second machine learning model adopts an LSTM neural network model.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, It is characterized in that When the processor executes the program, the full-process connection method of the distribution network is implemented as described in any one of claims 5 to 8.

10. A computer-readable storage medium having a computer program stored thereon, It is characterized in that When the program is executed by a processor, the distribution network full-process connection method as described in any one of claims 5 to 8 is implemented.

Citation Information

Patent Citations

  • Electric power distribution system based on machine learning

    CN115333113A

  • Hybrid deep learning model-based electric power prediction method, system and equipment

    CN115630286A

  • Method and system for commanding network interaction based on power grid situation

    CN116365513A

  • Fault detection method and device for complex computing power system, equipment and storage medium

    CN117873839A

  • Electric vehicle short-term charging load prediction method based on LSTM-GNN combination model

    CN118070841A