Risk assessment method, device and equipment for operation state of power grid and medium

Abstract

The invention discloses a power grid operation state risk assessment method, device and equipment and a medium, and relates to the technical field of power grid risk assessment, and the method comprises the steps: carrying out the structure optimization of a first deep neural network model, and constructing a second deep neural network model; aiming at the characteristic data corresponding to different dimensions of the power grid operation state, constructing a total information element data model; based on the database, training the second deep neural network model to obtain a target deep neural network model; and based on the target deep neural network model, predicting a target risk assessment level of the current power grid operation state. The method is suitable for risk assessment of the operation state of the power grid in scenes such as extreme weather and new energy generation.

Description

Technical Field

[0001] The present application relates to the technical field of power grid risk assessment, and particularly to a risk assessment method, device, equipment and medium for the operating state of a power grid. Background Art

[0002] With the large-scale access of existing new energy, the demand for the power system's balance regulation ability needs to be enhanced urgently. At the same time, the large-scale grid connection of intermittent power generation resources such as wind power and solar power generation makes the power balance show obvious spatial and temporal imbalance, exacerbating the burden of system regulation. On the one hand, a large number of conventional units are in deep peak shaving or even standby state, but the frequent adjustment of output brings problems such as cost increase and shortened life. The large-scale replacement of conventional units leads to insufficient system regulation ability, prominent contradictions in peak shaving, frequency modulation and voltage regulation, and significantly increases the difficulty of power grid operation control. On the other hand, the power grid form has changed, and the safe operation presents new characteristics, that is, with the large-scale production of new energy units at the sending end and the large-scale feeding of HVDC at the receiving end, the coupling relationship between future AC / DC, sending / receiving ends, and source-network-load is closer, which further leads to an increasing risk of cascading failures and affects the overall safety of the power grid.

[0003] Especially in scenarios such as extreme weather and large-scale new energy generation, the operating characteristics of the power grid are intertwined and complex. The power grid operating state risk system cannot comprehensively and comprehensively perceive the risk of the power grid operating state, and thus cannot achieve early perception of the power grid's tight balance state. Summary of the Invention

[0004] In view of this, the present application provides a risk assessment method, device, equipment and medium for the operating state of a power grid, mainly aiming to solve the technical problems in the prior art that a large number of conventional units are in deep peak shaving or even standby state, and the cascading failure risk brought by the coupling relationship in the new power grid form, resulting in low power grid security and inability to predict the risk state of power grid operation in advance.

[0005] According to one aspect of the present application, a risk assessment method for the operating state of a power grid is provided. The method includes:

[0006] Construct a second deep neural network model by optimizing the structure of the first deep neural network model;

[0007] Construct a full information element data model for the characteristic data corresponding to different dimensions of the power grid operating state;

[0008] Train the second deep neural network model based on the full information element data model to obtain a target deep neural network model;

[0009] Predict the target risk assessment level of the current power grid operating state based on the target deep neural network model.

[0010] According to another aspect of the present application, there is provided a risk assessment device for the operating state of a power grid, the device comprising:

[0011] A first construction module for constructing a second deep neural network model by optimizing the structure of a first deep neural network model;

[0012] A second module for constructing a full information element data model for the feature data corresponding to different dimensions of the power grid operating state;

[0013] A training module for training the second deep neural network model based on the full information element data model to obtain a target deep neural network model;

[0014] A prediction module for predicting the target risk assessment level of the current power grid operating state based on the target deep neural network model.

[0015] According to another aspect of the present application, there is provided a computer storage medium having a computer program stored thereon, and when the program is executed by a processor, the above-mentioned risk assessment method for the operating state of the power grid is implemented.

[0016] According to still another aspect of the present application, there is provided a computer device comprising a storage medium, a processor, and a computer program stored on the storage medium and executable on the processor, and when the processor executes the program, the above-mentioned risk assessment method for the operating state of the power grid is implemented.

[0017] By means of the above technical solutions, compared with the existing technical solutions that cannot predict the risk state of the power grid operation in advance, the risk assessment method, device, equipment and medium for the operating state of the power grid provided by the present application optimize the structure of the first deep neural network model to construct a second deep neural network model; construct a full information element data model for the feature data corresponding to different dimensions of the power grid operating state; train the second deep neural network model based on the full information element data model to obtain a target deep neural network model; predict the target risk assessment level of the current power grid operating state based on the target deep neural network model. It can be seen that by training the improved deep neural network model based on the feature data corresponding to different dimensions of the power grid operating state, the online identification of the risk state of the power grid operation is realized, the purpose of advanced perception of the tight balance state of the power grid is achieved, and thus a warning tool is provided for the risk state of the power grid operation that may be caused by extreme weather or rapid load climb.

[0018] The above description is only an overview of the technical solutions of the present application. In order to be able to understand the technical means of the present application more clearly, it can be implemented according to the content of the specification. And in order to make the above and other purposes, features and advantages of the present application more obvious and understandable, the following specific embodiments of the present application are specifically exemplified. Description of the Drawings

[0019] The drawings described herein are provided to further understand the present application and form a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation to the present application. In the drawings:

[0020] Figure 1 A schematic flowchart of a risk assessment method for the grid operation state provided by an embodiment of the present application is shown;

[0021] Figure 2 A schematic flowchart of another risk assessment method for the grid operation state provided by an embodiment of the present application is shown;

[0022] Figure 3 A schematic diagram showing the structural improvement of a second deep neural network model provided by an embodiment of the present application is shown;

[0023] Figure 4 A schematic flowchart of a target deep neural network model provided by an embodiment of the present application is shown;

[0024] Figure 5 A schematic diagram of the structure of a risk assessment device for the grid operation state provided by an embodiment of the present application is shown;

[0025] Figure 6 A schematic diagram of the structure of another risk assessment device for the grid operation state provided by an embodiment of the present application is shown. Detailed Description of the Embodiments

[0026] The present application will be described in detail below with reference to the drawings and in combination with embodiments. It should be noted that, without conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0027] In view of the technical problems that a large number of conventional units in the prior art are in the state of deep peak shaving or even shutdown for standby, and the risk of cascading failures brought by the coupling relationship under the new form of the power grid, resulting in low power grid security and inability to predict the risk state of power grid operation in advance. This embodiment provides a risk assessment method for the power grid operation state, which trains an improved deep neural network model based on the characteristic data corresponding to different dimensions of the power grid operation state, so as to realize the online identification of the risk state of the power grid operation, achieve the purpose of advanced perception of the tight balance state of the power grid, and further provide a warning tool for the risk state of power grid operation that may be caused by extreme weather or rapid load climb. As Figure 1 shown, the above method includes the following steps:

[0028] Step 101: Construct a second deep neural network model by optimizing the structure of the first deep neural network model.

[0029] In this embodiment, the first deep neural network model is a Transformer model. Based on the multi-head self-attention mechanism of the Transformer model, it can consider the correlation between features at different times simultaneously, achieve multi-level and multi-subspace attention to the input sequence data, and thus capture rich feature information in the input sequence data. To further improve the accuracy of model prediction, combined with the data features related to the power grid operation state, the structure of the Transformer model is optimized to construct a second deep neural network model. The input sequence data is the feature data corresponding to different dimensions of the power grid operation state, mainly including power grid feature data and environmental feature data.

[0030] Step 102: Construct a full-information element data model for the feature data corresponding to different dimensions of the power grid operation state.

[0031] In this embodiment, constructing a full-information element data model means constructing corresponding databases for the feature data corresponding to different dimensions of the power grid operation state. The feature data corresponding to different dimensions of the power grid operation state includes power grid feature data and environmental feature data. The power grid feature data includes the positive and negative reserve status data of the whole network and the operation status data of key sections. The first database is a database constructed based on the positive and negative reserve status data of the whole network, the second database is a database constructed based on the operation status data of key sections, and the third database is a database constructed based on the environmental feature data.

[0032] Step 103: Based on the full-information element data model, train the second deep neural network model to obtain a target deep neural network model.

[0033] In this embodiment, based on the feature data in the full-information element data model, simulation data for a preset time period is generated through the simulation verification platform of the dispatching automation system and used as sample data for training the second deep neural network model to achieve the training of the second deep neural network model. Further, the cross-entropy loss function and the Adam optimizer are used to iterate the second deep neural network model, thereby obtaining a target deep neural network model that can effectively improve the accuracy of early warning of the power grid operation risk state.

[0034] Step 104: Based on the target deep neural network model, predict the target risk assessment level of the current power grid operation state.

[0035] In this embodiment, by selecting the power grid operation status data of a preset time period, which is from the full information element data model and used as input sequence data to input into the target deep neural network model, so as to extract effective features from the multi-dimensional power grid operation status data and accurately predict the current tight balance state of the power grid (i.e., the target risk assessment level). The risk assessment level can be divided into 4 levels: extremely tight, severely tight, tight, and tense. According to the requirements of the actual application scenario, it can also be further determined as a non-risk state and a risk state. Here, the risk assessment level is not specifically limited.

[0036] For this embodiment, the above solution can be adopted. By optimizing the structure of the first deep neural network model, the second deep neural network model is constructed; for the characteristic data corresponding to different dimensions of the power grid operation status, a full information element data model is constructed; based on the full information element data model, the second deep neural network model is trained to obtain the target deep neural network model; based on the target deep neural network model, the target risk assessment level of the current power grid operation status is predicted. Compared with the technical solution that cannot predict the power grid operation risk status in advance, in this embodiment, by comprehensively considering the complex interweaving of multi-dimensional power grid operation status data in scenarios such as large-scale new energy generation, using the improved deep neural network model, the risk of the power grid operation status is intelligently evaluated from different dimensions such as time and space, and the online identification of the power grid operation risk status is carried out through deep neural network learning, so as to achieve the purpose of perceiving the tight balance state of the power grid in advance, and further provide a warning tool for the power grid operation risk status that may be caused by extreme weather or rapid load climb.

[0037] Further, as a refinement and extension of the specific implementation manner of the above embodiment, in order to fully illustrate the specific implementation process of this embodiment, another risk assessment method for the power grid operation status is provided. This embodiment can solve the technical problem that the power grid operation status risk system in the existing technical solution cannot comprehensively and synthetically perceive the power grid operation status risk, and thus cannot achieve the advanced perception of the tight balance state of the power grid. This embodiment can comprehensively consider uncertain factors such as new energy power generation fluctuations and external environments, and train the improved deep neural network model based on the characteristic data corresponding to different dimensions of the power grid operation status to realize the online identification of the power grid operation risk status, that is, the early warning of the power grid operation risk status, so as to provide dispatching and operation personnel with the opportunity to adjust the operation mode, reduce the investment in power grid construction for low-risk fault prevention and the corresponding margin control in the operation adjustment process, thereby effectively reducing the probability of cascading failures, enhancing the ability of the power grid to resist possible risks, improving the economy of system operation, and ensuring the safe operation of the power grid.

[0038] As Figure 2 shown, the risk assessment method for the power grid operation status based on the TransFormer model includes:

[0039] Step 201: Build a second deep neural network model by optimizing the structure of the first deep neural network model.

[0040] To illustrate the specific implementation of step 201, as a preferred embodiment, step 201 includes: replacing the convolutional layer in the multi-head self-attention mechanism of the first deep neural network model with a multi-scale dilated convolutional layer; and replacing the multi-layer perceptron in the first deep neural network model with a convolutional layer to obtain the second deep neural network model. It should be noted that there is no order limit for the above steps.

[0041] In implementation, the first deep neural network model is a Transformer model. The second deep neural network model fuses multiple encoded feature vectors obtained by encoding different stages of the multi-head self-attention mechanism as supplementary feature vectors with the initial input vector in the decoding stage to obtain the target input vector in the decoding stage.

[0042] In implementation, build a basic Transformer model applicable to the risk assessment of power grid operation status. The Transformer model can not only capture complex patterns and dynamic changes in time series, but also has high sensitivity and recognition ability for subtle changes in the actual power grid operation status, and has strong classification ability.

[0043] Considering that the risk assessment of power grid operation status is an important part of power system dispatching operation, to improve the accuracy of model prediction, optimize the structure of the basic Transformer model (the first deep neural network model). Specifically, as Figure 3 shown, first, replace the convolutional layer in the multi-head self-attention mechanism with a multi-scale dilated convolutional layer, so as to associate more factor information by parallel processing of sequence data, reduce the number of parameters, improve the computational efficiency of the network, and reduce the deployment requirements for hardware; second, replace the traditional multi-layer perceptron with 1×1 convolutional layers and 3×3 convolutional layers with fewer parameters (for example, 1 1×1 convolutional layer and 3 3×3 convolutional layers), so as to fully consider the requirement of retaining some redundant information for information recovery while having less computational volume; finally, use the encoded features at different stages as supplementary feature vectors in the decoding stage to achieve data feature completion and improve the comprehensiveness and accuracy of model prediction. It can be seen that the optimized Transformer model (the second deep neural network model) can handle long-distance dependence problems and greatly improve the learning ability and efficiency of the Transformer model.

[0044] Step 202: Build a full-information element data model for the feature data corresponding to different dimensions of the power grid operation status.

[0045] In implementation, the characteristic data corresponding to different dimensions of the power grid operation state include power grid characteristic data and environmental characteristic data; the power grid characteristic data includes the positive and negative reserve state data of the whole network and the operation state data of key sections; correspondingly, the full information element data model includes the positive and negative reserve state data model of the whole network, the operation state data model of key sections, and the external environmental characteristic data model. The positive and negative reserve state data of the whole network in the positive and negative reserve state data model of the whole network includes one or several of the power generation data of the whole network, the load data of the whole network, the network loss data of the whole network, the positive and negative reserve capacity of the whole network, the positive and negative reserve margin of the whole network, the maximum transmission power of external power, and the proportion of external power in the load; the operation state data of key sections in the operation state data model of key sections includes one or several of the load rate of the key section for power off in the target area, the load rate of the first and second level section currents in the DC near area, the total power of the key transmission channel, the maximum transmission power of the key transmission channel, and the margin of the key transmission channel; the environmental characteristic data (meteorological factors) in the external environmental characteristic data model includes one or several of temperature, humidity, rainfall, wind speed, and solar irradiance intensity. Among them, the maximum transmission power of external power is the maximum transmission power determined by the cross-region and cross-province plan.

[0046] (1) The calculation formulas for the positive and negative reserve capacity and the positive and negative reserve margin in the positive and negative reserve state data of the whole network are as follows:

[0047] ① Positive and negative reserve capacity of the whole network

[0048] The positive and negative reserve capacity of the whole network refers to the capacity that the system's active operation reserve capacity can be adjusted upward (downward) under the condition of safe operation of the power grid. The positive reserve capacity P Tup of the whole network and the negative reserve capacity P Tdown of the whole network are calculated as follows:

[0049]

[0050] Among them, Pi,max is the maximum output of all synchronous units in the system, Pi is the output of the i-th synchronous unit in the system, and Pi,min is the minimum output of all synchronous units in the system.

[0051] ② Positive and negative reserve margin of the whole network

[0052] The positive and negative reserve margin of the whole network refers to the range that the system's active operation reserve capacity can be adjusted upward (downward) under the condition of safe operation of the power grid. The positive reserve margin P Tup-magrin of the whole network and the negative reserve margin P Tdown-magrin of the whole network are calculated as follows:

[0053]

[0054] (2) The calculation formulas for the load rate of the key section for power flow from the 500kV partition to the lower network, the load rate of the power flow in the first and second-level sections in the DC near area, the maximum transmission power of the key transmission channels, and the margin of the key transmission channels in the operation status data of the key sections are as follows:

[0055] ① Load rate of the key section for power flow from the 500kV partition to the lower network

[0056] The key section for power flow from the 500kV partition to the lower network is the focus of monitoring by the regulation system. The key section is restricted by the thermal stability transmission limit of the primary power transmission equipment. Therefore, the load rate P Srw of the key section for power flow from the 500kV partition to the lower network is calculated as follows:

[0057]

[0058] Among them, P Slimit is the thermal stability transmission limit capacity of the section considering N-1 outage, and P Sf is the transmission power of the key section for power flow from the 500kV partition to the lower network.

[0059] ② Load rate of the power flow in the first and second-level sections in the DC near area

[0060] The power of the external transmission channel is arranged according to the superior dispatching plan. The power flow in the first and second-level sections in the DC near area is limited by the section stability limit. Therefore, the load rate P S-DC of the power flow in the first and second-level sections in the DC near area is calculated as follows:

[0061]

[0062] Among them, P Sdcjq is the power flow in the first and second-level sections in the DC near area, and P Sdcjqlimit is the section stability limit of the first and second-level sections in the DC near area.

[0063] ③ Maximum transmission capacity of the key transmission channels

[0064] The maximum transmission capacity P Ch-ability of the key transmission channels is calculated as follows:

[0065]

[0066] Among them, K is a coefficient, P Ch,i is the transmission power of the key transmission channel i, is the stability limit of the key transmission channel i.

[0067] ④ Margin of the key transmission channels

[0068] The margin P Ch-magrin of the key transmission channels is calculated as follows:

[0069]

[0070] Step 203: Generate simulation data for a preset time period based on the characteristic data in the full information element data model, and use it as sample data for training the second deep neural network model.

[0071] Step 204: Train the second deep neural network model based on the sample data to obtain a target deep neural network model.

[0072] To illustrate the specific implementation of step 204, as a preferred embodiment, step 204 includes: presetting multiple output results of the second deep neural network model, each output result corresponding to a different risk assessment level; and iteratively training the second deep neural network model based on the sample data until the output result of the second deep neural network model is consistent with the risk assessment label of the preset power grid operation state, so as to obtain a target deep neural network model.

[0073] In implementation, the characteristic data in the full information element data model is historical data. Based on historical data (taking a provincial power grid as an example, selecting historical data of the power grid operation state for 7 days in a week), one year of simulation data is generated through the simulation verification platform of the dispatching automation system and used as sample data to construct a sample data set for training the improved TransFormer model. The power grid operation risk state labels in the sample data set are determined by expert rules. Further, the second deep neural network model is trained using the cross-entropy loss function and the Adam optimizer to obtain a target deep neural network model, that is, an agent for predicting the power grid operation risk state. It can be seen that during the agent training process, by improving the robustness of the model under different time periods and different load conditions, the error of the loss function decreases, and the accuracy of the agent increases, which not only verifies the effectiveness of the TransFormer model but also deepens the Transformer model's understanding of the characteristics of the power grid operation state.

[0074] In implementation, such as Figure 4As shown, the improved Transformer model can extract data features from multiple feature heads more meticulously based on multi-head attention, and fuse and decode the data features of different feature heads through multiple linear layers and two 1×1 convolutional layers respectively, outputting the risk assessment results of the power grid operation state. The risk assessment results include a predicted probability vector between 0 and 1 for each set risk assessment level (extremely tight, severely tight, tight, tense). Select the maximum value from the obtained 4 predicted probability vectors, and use the risk assessment level corresponding to the maximum value as the target risk assessment level. For example, if the predicted probability vector obtained by softmax is (0.8, 0.5, 0.4, 0.2) and the maximum value is 0.8, it is determined that the power grid reserve is in an extremely tight state. It can be seen that by predicting the target risk assessment level of the current power grid operation state through the target deep neural network model, the hierarchical determination and early warning of the power grid tight balance state can be realized.

[0075] Step 205: Obtain the current feature data corresponding to different dimensions of the current power grid operation state.

[0076] Step 206: According to the current feature data, use the target deep neural network model to output multiple predicted probability values for different risk assessment levels.

[0077] Step 207: Use the risk assessment level corresponding to the maximum predicted probability value among the multiple predicted probability values as the target risk assessment level.

[0078] By applying the technical solution of this embodiment, taking a provincial power grid as an example, the target deep neural network model can extract effective features from multi-dimensional power grid feature data, accurately predict the tight balance state of the power grid, and then accurately distinguish the risk state and non-risk state in the power grid operation state, and the accuracy rate reaches the expected target. It can be seen that the improved target deep neural network model not only performs well in sample construction and feature selection, but also achieves a satisfactory effect in prediction accuracy.

[0079] It can be seen that by optimizing the structure of the first deep neural network model, the second deep neural network model is constructed; for the feature data corresponding to different dimensions of the power grid operation state, a full-information element data model is constructed; based on the full-information element data model, the second deep neural network model is trained to obtain the target deep neural network model; based on the target deep neural network model, the target risk assessment level of the current power grid operation state is predicted. Compared with the existing technical solutions that cannot predict the power grid operation risk state in advance, in this embodiment, a TransForm model for predicting the power grid operation risk state is constructed through structure optimization, and a data model containing full information elements is constructed to fully consider the influence of power grid feature data and environmental feature data on the power grid operation state, and provide a data basis for subsequent sample data construction and model training. It can be seen that accurate prediction of the power grid operation state can help power grid operators adjust the allocation of power resources in a timely manner to optimize the power grid operation strategy.

[0080] Further, as Figure 1 a specific implementation of the method, an embodiment of the present application provides a risk assessment device for the power grid operation state, as Figure 5 shown. The device includes: a first construction module 51, a second construction module 52, a training module 53, and a prediction module 54.

[0081] The first construction module 51 is used to construct the second deep neural network model by optimizing the structure of the first deep neural network model.

[0082] The second construction module 52 is used to construct a full-information element data model for the feature data corresponding to different dimensions of the power grid operation state.

[0083] The training module 53 is used to train the second deep neural network model based on the constructed full-information element data model to obtain the target deep neural network model.

[0084] The prediction module 54 is used to predict the target risk assessment level of the current power grid operation state based on the target deep neural network model.

[0085] In a specific application scenario, as Figure 6 shown, the first construction module 51 includes: a first replacement sub-module 511 and a second replacement sub-module 512.

[0086] The first replacement sub-module 511 is used to replace the convolutional layer in the multi-head self-attention mechanism of the first deep neural network model with a multi-scale dilated convolutional layer.

[0087] The second replacement sub-module 512 is used to replace the multi-layer perceptron in the first deep neural network model with a convolutional layer to obtain the second deep neural network model.

[0088] In a specific application scenario, the second deep neural network model fuses multiple encoded feature vectors obtained by encoding different stages of the multi-head self-attention mechanism with the initial input vector in the decoding stage as supplementary feature vectors to obtain the target input vector in the decoding stage.

[0089] In a specific application scenario, the characteristic data corresponding to different dimensions of the power grid operation state includes power grid characteristic data and environmental characteristic data; the power grid characteristic data includes the positive and negative reserve state data of the whole network and the operation state data of key sections; the positive and negative reserve state data of the whole network includes one or several of the power generation data of the whole network, the load data of the whole network, the network loss data of the whole network, the positive and negative reserve capacity of the whole network, the positive and negative reserve margins of the whole network, the maximum transmission power of external power, and the proportion of external power in the load; the operation state data of key sections includes one or several of the load rate of the key section for power grid connection in the target area, the load rate of the first and second level section power flows in the DC near area, the total power of the key transmission channel, the maximum transmission power of the key transmission channel, and the margin of the key transmission channel; the environmental characteristic data includes one or several of temperature, humidity, rainfall, wind speed, and solar irradiance intensity.

[0090] In a specific application scenario, the training module 53 includes a generation sub-module 531 and a training sub-module 532.

[0091] The generation sub-module 531 is used to generate simulation data for a preset time period according to the characteristic data in the full information element data model and use it as sample data for training the second deep neural network model.

[0092] The training sub-module 532 is used to train the second deep neural network model according to the sample data to obtain the target deep neural network model.

[0093] In a specific application scenario, the training sub-module 532 is specifically used to preset multiple output results of the second deep neural network model, and each output result corresponds to a different risk assessment level; and, iteratively train the second deep neural network model based on the sample data until the output result of the second deep neural network model is consistent with the risk assessment label of the preset power grid operation state to obtain the target deep neural network model.

[0094] In a specific application scenario, the prediction module 54 includes an acquisition sub-module 541, an output sub-module 542, and a selection sub-module 543.

[0095] The acquisition sub-module 541 is used to acquire the current characteristic data corresponding to different dimensions of the current power grid operation state.

[0096] The output sub-module 542 is configured to output multiple prediction probability values for different risk assessment levels by using a target deep neural network model according to the current feature data.

[0097] The selection sub-module 543 is configured to use the risk assessment level corresponding to the maximum prediction probability value among the multiple prediction probability values as the target risk assessment level.

[0098] It should be noted that for other corresponding descriptions of each functional unit involved in the risk assessment device for the power grid operation state provided in the embodiments of the present application, reference can be made to Figure 1 and Figure 2 the corresponding descriptions therein, which will not be elaborated herein.

[0099] Based on the above methods as Figure 1 and Figure 2 shown, correspondingly, the embodiments of the present application further provide a computer storage medium, on which a computer program is stored, and when the program is executed by a processor, it implements the risk assessment method for the power grid operation state as Figure 1 and Figure 2 shown.

[0100] Based on such an understanding, the technical solution of the present application can be embodied in the form of a software product, which can be stored in a storage medium (such as a CD-ROM, a USB flash drive, a mobile hard disk, etc.), and includes several instructions for causing a computer device (such as a personal computer, a server, or a network device, etc.) to execute the methods described in various implementation scenarios of the present application.

[0101] Based on the above methods as Figure 1 , Figure 2 shown, and Figure 5 , Figure 6 shown in the virtual device embodiments, in order to achieve the above object, the embodiments of the present application further provide a computer device, specifically a personal computer, a server, a network device, etc., and the entity device includes a storage medium and a processor; the storage medium is used to store a computer program; the processor is used to execute the computer program to implement the risk assessment method for the power grid operation state as Figure 1 and Figure 2 shown.

[0102] Optionally, the computer device may further include a user interface, a network interface, a camera, a radio frequency (RF) circuit, sensors, an audio circuit, a WI-FI module, etc. The user interface may include a display and an input unit such as a keyboard, etc. Optionally, the user interface may further include a USB interface, a card reader interface, etc. The network interface may optionally include a standard wired interface, a wireless interface (such as a Bluetooth interface, a WI-FI interface), etc.

[0103] Those skilled in the art can understand that the structure of a computer device provided in this embodiment does not constitute a limitation on the physical device, and may include more or fewer components, or combine certain components, or have different component arrangements.

[0104] The storage medium may further include an operating system and a network communication module. The operating system is a program for managing the hardware and software resources of the computer device, and supports the operation of information processing programs and other software and / or programs. The network communication module is used to implement communication between components inside the storage medium, as well as communication between other hardware and software in the physical device.

[0105] Through the description of the above embodiments, those skilled in the art can clearly understand that the present application can be implemented by means of software plus a necessary general hardware platform, or can also be implemented by hardware. By applying the technical solution of the present application, compared with the technical solution in the prior art where the power grid operation status risk system cannot comprehensively and comprehensively perceive the power grid operation status risk, in this embodiment, a TransForm model for predicting the power grid operation risk status is constructed through structural optimization, and a data model containing all information elements is constructed, so as to fully consider the influence of power grid characteristic data and environmental characteristic data on the power grid operation status, and provide a data basis for subsequent construction of sample data and model training. It can be seen that accurate prediction of the power grid operation status can help the power grid operator timely adjust the allocation of power resources to optimize the power grid operation strategy.

[0106] Those skilled in the art can understand that the drawings are only schematic diagrams of a preferred implementation scenario, and the modules or processes in the drawings are not necessarily essential for implementing the present application. Those skilled in the art can understand that the modules in the device in the implementation scenario can be distributed in the device in the implementation scenario according to the description of the implementation scenario, or can be correspondingly changed and located in one or more devices different from the present implementation scenario. The modules in the above implementation scenario can be combined into one module, or further split into multiple sub-modules.

[0107] The above serial numbers of this application are only for description and do not represent the advantages or disadvantages of the implementation scenarios. The above disclosure is only several specific implementation scenarios of this application. However, this application is not limited thereto, and any changes that can be conceived by those skilled in the art should fall within the protection scope of this application.

Claims

1. A method for risk assessment of power grid operation status, characterized in that: include: By optimizing the structure of the first deep neural network model, a second deep neural network model is constructed; Construct a full information element data model based on the characteristic data corresponding to different dimensions of the power grid operation status; Based on the full information element data model, training the second deep neural network model to obtain a target deep neural network model; Based on the target deep neural network model, a target risk assessment level of the current power grid operation status is predicted.

2. The method according to claim 1, characterized in that The first deep neural network model is a TransFormer model, and the step of constructing a second deep neural network model by structurally optimizing the first deep neural network model includes: Replace the convolutional layers in the multi-head self-attention mechanism of the first deep neural network model with multi-scale dilated convolutional layers; The multi-layer perceptron in the first deep neural network model is replaced with a convolutional layer to obtain a second deep neural network model.

3. The method according to claim 1 or 2, characterized in that: The second deep neural network model uses multiple encoded feature vectors obtained by encoding at different stages of the multi-head self-attention mechanism as supplementary feature vectors and fuses them with the initial input vector of the decoding stage to obtain the target input vector of the decoding stage.

4. The method according to claim 1, characterized in that: The characteristic data corresponding to different dimensions of the power grid operation status include power grid characteristic data and environmental characteristic data; the power grid characteristic data include positive and negative standby status data of the whole network and operation status data of key sections; the positive and negative standby status data of the whole network include one or more of the power generation data of the whole network, the load data of the whole network, the network loss data of the whole network, the positive and negative standby capacity of the whole network, the positive and negative standby margin of the whole network, the maximum transmission power of external electricity, and the proportion of external electricity in the load; the operation status data of the key section include one or more of the load rate of the key section of the target partition, the load rate of the first and second level sections of the DC near area, the total power of the key transmission channel, the maximum transmission power of the key transmission channel, and the margin of the key transmission channel; the environmental characteristic data include one or more of temperature, humidity, rainfall, wind speed, and solar radiation intensity.

5. The method according to claim 1 or 4, characterized in that: The step of training the second deep neural network model to obtain a target deep neural network model based on the full information element data model includes: Generate simulation data of a preset time period according to the feature data in the full information element data model, and use it as sample data for training the second deep neural network model; Based on the sample data, the second deep neural network model is trained to obtain a target deep neural network model.

6. The method according to claim 5, characterized in that The step of training the second deep neural network model to obtain a target deep neural network model according to the sample data includes: Preset multiple output results of the second deep neural network model, each output result corresponds to a different risk assessment level; The second deep neural network model is iteratively trained based on the sample data until an output result of the second deep neural network model is consistent with a preset risk assessment label of the power grid operation status, thereby obtaining a target deep neural network model.

7. The method according to claim 1, characterized in that The step of predicting a target risk assessment level of a current power grid operation state based on the target deep neural network model comprises: Obtain current feature data corresponding to different dimensions of the current power grid operation status; Outputting a plurality of predicted probability values ​​for different risk assessment levels using a target deep neural network model according to the current feature data; The risk assessment level corresponding to the maximum predicted probability value among the multiple predicted probability values ​​is used as the target risk assessment level.

8. A risk assessment device for power grid operation status, characterized in that: include: A first construction module is used to construct a second deep neural network model by optimizing the structure of the first deep neural network model; The second construction module is used to construct a full information element data model according to the characteristic data corresponding to different dimensions of the power grid operation status; A training module, used for training the second deep neural network model to obtain a target deep neural network model based on the constructed full information element data model; The prediction module is used to predict the target risk assessment level of the current power grid operation status based on the target deep neural network model.

9. A computer storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the risk assessment method for the operation status of a power grid as described in any one of claims 1 to 7 is implemented.

10. A computer device comprising a storage medium, a processor, and a computer program stored in the storage medium and executable on the processor, characterized in that: When the processor executes the program, the risk assessment method for the power grid operation status described in any one of claims 1 to 7 is implemented.

Images