A security risk assessment system and method based on artificial intelligence algorithm model

Abstract

The present application provides a safety risk assessment system and method based on an artificial intelligence algorithm model, which belongs to the field of risk management technology, wherein the system includes: an acquisition module for constructing a first-stage risk assessment model; a first assessment module for obtaining a first safety risk assessment result based on real-time resource data of a first operating cycle and the first-stage risk assessment model; a second assessment module for obtaining real-time noise impact data, and dynamically correcting the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model; a comparison module for determining a target risk assessment model based on a comparison result of the first safety risk assessment result and the second safety risk assessment result; a third assessment module for determining a target safety risk assessment result based on the real-time resource data of the second operating cycle and the target risk assessment model. The present application aims to improve the problem of being unable to perform accurate risk assessment on virtual power plants and the systems connected to them.

Description

Technical Field

[0001] The present application relates to the field of risk management technology, and in particular to a security risk assessment system and method based on an artificial intelligence algorithm model. Background Art

[0002] A virtual power plant is a technology and management platform that integrates distributed energy resources. Through advanced information and communication technologies, a virtual power plant can monitor and dispatch distributed energy resources in real time, achieving optimal resource allocation and collaborative operation, thereby improving the flexibility and reliability of the power system and functionally operating like a traditional power plant.

[0003] Virtual power plants are usually used in conjunction with power demand systems. Since virtual power plants do not have the convenient unified management characteristics of traditional power generation units, the risk management methods of traditional power generation units cannot be applied to virtual power plants. Summary of the Invention

[0004] The present invention provides a security risk assessment system and method based on an artificial intelligence algorithm model. The present invention adopts the following technical solutions:

[0005] In a first aspect, a security risk assessment system based on an artificial intelligence algorithm model is provided. The system includes:

[0006] The acquisition module is used to obtain the historical resource data of the virtual power plant and the multi-dimensional operating parameters of the system to be monitored, and to build the first-stage risk assessment model based on the real-time resource data and multi-dimensional operating parameters;

[0007] a first assessment module, configured to obtain real-time resource data of a first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and a first-stage risk assessment model;

[0008] The second assessment module is used to obtain real-time noise impact data and dynamically modify the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model;

[0009] a comparison module, configured to obtain a second security risk assessment result based on the real-time resource data of the first operation cycle and the second-stage risk assessment model, and determine a target risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0010] The third evaluation module is used to obtain real-time resource data of the second operating cycle of the virtual power plant, and determine the target safety risk assessment result based on the real-time resource data of the second operating cycle and the target risk assessment model.

[0011] In a feasible implementation, it includes:

[0012] A screening submodule is used to screen out multiple target risk assessment parameters for risk assessment from the multi-dimensional operating parameters of the system to be monitored;

[0013] The model building submodule is used to build multiple alternative risk assessment models based on target risk assessment parameters and real-time resource data;

[0014] The model fusion submodule is used to fuse multiple alternative risk assessment models to construct the first-stage risk assessment model.

[0015] In a feasible implementation, the model building submodule includes:

[0016] a combining unit, configured to arbitrarily combine at least two target evaluation parameters to obtain a plurality of combination results, wherein each combination result corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set;

[0017] A training unit, configured to train a preset model based on the first data set to obtain an initial risk assessment model;

[0018] The verification unit is configured to verify the accuracy of the initial risk assessment model based on the second data set to obtain an alternative risk assessment model.

[0019] In a feasible implementation, the screening submodule includes:

[0020] A calculation unit, configured to perform dimensionality reduction processing on the operating parameters of each dimension and calculate the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing;

[0021] The screening unit is used to screen the multi-dimensional operating parameters according to the amount of information to obtain the target operating parameters.

[0022] In a feasible implementation, the comparison module includes:

[0023] An optimization direction determination submodule, configured to determine an optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0024] A first optimization submodule is configured to execute a first optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization;

[0025] The second optimization submodule is used to execute a second optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is negative optimization.

[0026] In a feasible implementation, the first optimization submodule includes:

[0027] A model determination unit, used to determine the second-stage risk assessment model as the main model and the first-stage risk assessment model as the auxiliary model;

[0028] An optimization direction determination unit, used to perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0029] The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

[0030] In a feasible implementation, the second optimization submodule includes:

[0031] Model determination unit, used to use the first stage risk assessment model as the main model and the second stage risk assessment model as the auxiliary model

[0032] An optimization direction determination unit, used to perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0033] The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

[0034] In a second aspect, a security risk assessment method based on an artificial intelligence algorithm model is provided, which is applied to the security risk assessment system based on an artificial intelligence algorithm model in the first aspect, and the method includes:

[0035] Obtain historical resource data of the virtual power plant and multi-dimensional operating parameters of the system to be monitored, and build a first-stage risk assessment model based on real-time resource data and multi-dimensional operating parameters;

[0036] Acquire real-time resource data of a first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and a first-stage risk assessment model;

[0037] Obtain real-time noise impact data, and dynamically modify the first-stage risk assessment model based on the real-time noise impact data to obtain the second-stage risk assessment model;

[0038] Obtaining a second security risk assessment result based on the real-time resource data of the first operation cycle and the second-stage risk assessment model, and determining a target risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0039] Acquire real-time resource data of the second operation cycle of the virtual power plant, and determine the target safety risk assessment result based on the real-time resource data of the second operation cycle and the target risk assessment model.

[0040] In one feasible implementation, a first-stage risk assessment model is constructed based on real-time resource data and multi-dimensional operating parameters, including:

[0041] Arbitrarily combining at least two target evaluation parameters to obtain a plurality of combination results, wherein each combination result corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set;

[0042] Based on the first data set, a preset model is trained to obtain an initial risk assessment model;

[0043] Based on the second data set, the initial risk assessment model is verified for accuracy to obtain an alternative risk assessment model.

[0044] In a feasible implementation, multiple target risk assessment parameters for risk assessment are screened from the multi-dimensional operating parameters of the system to be monitored, including:

[0045] Perform dimensionality reduction processing on the operating parameters of each dimension, and calculate the information amount corresponding to the operating parameters of each dimension after dimensionality reduction processing;

[0046] According to the amount of information, the multidimensional operating parameters are screened to obtain the target operating parameters.

[0047] In a feasible implementation, multiple target risk assessment parameters for risk assessment are screened from the multi-dimensional operating parameters of the system to be monitored, including:

[0048] A calculation unit, configured to perform dimensionality reduction processing on the operating parameters of each dimension and calculate the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing;

[0049] The screening unit is used to screen the multi-dimensional operating parameters according to the amount of information to obtain the target operating parameters.

[0050] In a feasible implementation, determining the target risk assessment model according to a comparison result of the first security risk assessment result and the second security risk assessment result includes:

[0051] An optimization direction determination submodule, configured to determine an optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0052] The first optimization submodule is used to execute the first optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization.

[0053] The second optimization submodule is used to execute the second optimization strategy when the optimization direction of the second-stage risk assessment model is negative optimization relative to the first-stage risk assessment model.

[0054] In a feasible implementation, when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization, the first optimization strategy is executed:

[0055] Determine the second-stage risk assessment model as the main model and the first-stage risk assessment model as the auxiliary model;

[0056] Perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0057] Model fusion is performed based on the optimized auxiliary model and main model to obtain the target risk assessment model.

[0058] In a feasible implementation, when the optimization direction of the risk assessment model in the second stage is negative optimization relative to the risk assessment model in the first stage, executing the second optimization strategy includes:

[0059] The first-stage risk assessment model is the main model, and the second-stage risk assessment model is the auxiliary model.

[0060] Perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0061] Model fusion is performed based on the optimized auxiliary model and main model to obtain the target risk assessment model.

[0062] In summary, the above-mentioned security risk assessment system based on artificial intelligence algorithm model has the following technical effects:

[0063] The embodiment of the present application provides a safety risk assessment system based on an artificial intelligence algorithm model. By acquiring the historical resource data of the virtual power plant and the multi-dimensional operating parameters of the system to be monitored, an initial risk assessment model is constructed, and the model is evaluated using real-time resource data in the first operating cycle of the virtual power plant to obtain a preliminary safety risk assessment result. Subsequently, by acquiring real-time noise impact data, the initial model is dynamically corrected to form a more accurate second-stage risk assessment model. Based on the real-time resource data of the first operating cycle and the corrected model, a risk assessment is performed again to obtain the safety risk assessment result of the second stage. By comparing the assessment results of the first and second stages, the final target risk assessment model is determined. Finally, the real-time resource data of the second operating cycle of the virtual power plant is analyzed using the target model to obtain the final target safety risk assessment result. In this way, by gradually optimizing and dynamically adjusting the assessment model, the present application can effectively respond to the management challenges of the dispersion and diversity of virtual power plants, and achieve accurate risk management and real-time regulation. BRIEF DESCRIPTION OF THE DRAWINGS

[0064] Figure 1 A schematic diagram of the structure of a security risk assessment system based on an artificial intelligence algorithm model provided in an embodiment of the present application;

[0065] Figure 2 A schematic diagram of the steps of an evaluation method provided in an embodiment of the present application. DETAILED DESCRIPTION

[0066] The technical solution in this application will be described below with reference to the accompanying drawings.

[0067] In order to make the purpose, technical solutions and advantages of this application more clear, the following Figure 1 、 Figure 2 It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0068] The terms "first", "second", etc. involved in the embodiments of the present application are only used to distinguish features of the same type and cannot be understood as indicating relative importance, quantity, order, etc.

[0069] The terms "exemplary" or "for example" in the embodiments of this application are used to indicate examples, illustrations, or descriptions. Any embodiment or design described in this application as "exemplary" or "for example" should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0070] The terms "coupling" and "connection" involved in the embodiments of this application should be understood in a broad sense. For example, they may refer to a physical direct connection, or an indirect connection achieved through electronic devices, such as a connection achieved through resistors, inductors, capacitors or other electronic devices.

[0071] A virtual power plant (VPP) is a new type of power management and distributed energy integration system. It integrates geographically dispersed small distributed energy resources (such as solar panels, wind turbines, battery storage systems, and controllable loads) into a centralized management and scheduling platform. Its functionality is similar to that of a traditional power plant, but without the need for a large-scale power infrastructure.

[0072] Virtual power plants are often used in conjunction with power demand systems, such as air conditioning management systems, smart lighting systems, and electric vehicle charging management systems, to achieve more efficient energy management and dispatch. Traditional power generation units typically employ centralized control and management, with risk management focused on the operational safety, maintenance, and troubleshooting of large-scale power generation facilities. This management approach relies on centralized monitoring and control systems and is suitable for large-scale, single-type power generation equipment. Furthermore, the equipment and operating procedures of traditional power generation units are relatively standardized, allowing risk management approaches to rely on unified standards and procedures. However, due to the diverse resources and widespread geographical distribution of virtual power plants, traditional centralized control and unified standards are difficult to apply. Consequently, effective and accurate risk assessment and evaluation of the overall virtual power plant and power demand system combination is impossible.

[0073] Based on this, the inventors proposed the inventive concept of this application. Through a comprehensive analysis of historical data and multi-dimensional operating parameters, they established a risk assessment model for the virtual power plant and power demand system as a whole. This model is then dynamically revised and optimized using real-time data during actual operation. This model not only takes into account the diversity and dispersion of different resources and systems, but also makes the risk assessment model more accurate and adaptable through continuous feedback from real-time data and adjustments to noise effects, thereby achieving effective and accurate risk assessment and evaluation of the entire system.

[0074] Reference Figure 1 , this application provides a security risk assessment system based on artificial intelligence algorithm model, including

[0075] The acquisition module 101 is used to obtain the historical resource data of the virtual power plant and the multi-dimensional operating parameters of the system to be monitored, and to construct a first-stage risk assessment model based on the real-time resource data and the multi-dimensional operating parameters.

[0076] In this embodiment, in order to construct the first-stage risk assessment model, it is necessary to obtain the historical resource data of the virtual power plant and the multi-dimensional operating parameters of the system to be monitored. The historical resource data of the virtual power plant includes its past power generation, energy consumption, fault records and other information, while the multi-dimensional operating parameters of the system to be monitored cover all kinds of data involved in its operation process, such as voltage, current, temperature, load conditions, etc. By comprehensively analyzing these historical data and operating parameters, a model that reflects the mutual influence and interaction between the virtual power plant and the system to be monitored can be established. This model can not only reflect the performance of each distributed energy source and equipment under different conditions, but also reveal potential risk factors and operating laws, thereby providing a reliable basis for subsequent real-time monitoring and risk assessment.

[0077] In a feasible implementation, the acquisition module includes:

[0078] A screening submodule is used to screen out multiple target risk assessment parameters for risk assessment from the multi-dimensional operating parameters of the system to be monitored;

[0079] The model building submodule is used to build multiple alternative risk assessment models based on target risk assessment parameters and real-time resource data;

[0080] The model fusion submodule is used to fuse multiple alternative risk assessment models to construct the first-stage risk assessment model.

[0081] In this embodiment, for the system to be monitored, in actual operation, the system may collect a large amount of operating data, and not all data are equally important for risk assessment. The task of the screening submodule is to screen out the most relevant and critical parameters based on the requirements and standards of risk assessment, which serve as the basis for subsequent modeling. Once the target risk assessment parameters are determined, multiple alternative risk assessment models can be constructed based on the combination of these target risk assessment parameters, and each risk assessment model can be constructed. Different risk assessment models may each have their own advantages and limitations and may be based on different assumptions, data distributions or modeling algorithms. Model fusion can integrate the prediction results of each model, balance the strengths and weaknesses of each model, and thus improve the accuracy of the overall prediction. By integrating multiple models, the errors that may be caused by specific data deviations in a single model can be reduced, and the stability and reliability of the final prediction can be improved. In practical applications, a single complex model may overfit the training data, resulting in poor performance on new data. Through model fusion, the risk of overfitting can be reduced, making the final risk assessment model more generalizable.

[0082] In a feasible implementation, the screening submodule includes:

[0083] A calculation unit, configured to perform dimensionality reduction processing on the operating parameters of each dimension and calculate the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing;

[0084] The screening unit is used to screen the multi-dimensional operating parameters according to the amount of information to obtain the target operating parameters.

[0085] In this embodiment, the dimensionality reduction process may include statistical methods or feature selection techniques, which aim to reduce the complexity and redundancy of the data while retaining key information. In this process, the operating parameters of each dimension will be converted or compressed for more efficient processing and analysis. After the dimensionality reduction process, the calculation unit will evaluate the amount of information corresponding to the operating parameters of each dimension. The amount of information can be understood as the importance of the data or the degree of contribution to the system state. It is usually calculated using concepts in information theory, such as information entropy or other information measurement methods, to quantify the relative importance of each parameter in system monitoring and evaluation. The screening unit screens the multidimensional operating parameters based on the amount of information calculated. That is, only the operating parameters with a large amount of information are retained, while the parameters with a small amount of information or no importance are discarded. The parameters that have the greatest impact on the system state are selected to improve the efficiency and accuracy of monitoring and evaluation. These parameters are considered to be the most critical indicators of the system operating state and are the basis for subsequent analysis, modeling or decision-making processes.

[0086] In a feasible implementation, the model building submodule includes:

[0087] a combining unit, configured to arbitrarily combine at least two target evaluation parameters to obtain a plurality of combination results, wherein each combination result corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set;

[0088] A training unit, configured to train a preset model based on the first data set to obtain an initial risk assessment model;

[0089] The verification unit is configured to verify the accuracy of the initial risk assessment model based on the second data set to obtain an alternative risk assessment model.

[0090] In this embodiment, when constructing an alternative risk assessment model, at least two parameters are selected from a plurality of screened target assessment parameters and are combined. These parameters can be key data points or indicators in the risk assessment process, usually for a more comprehensive analysis of specific aspects of the system or the combined effects. Each selected parameter combination can produce an independent combined result. These results reflect the performance or impact of each combined parameter on the historical resource data set, that is, how they appear on the actual data set, and can be statistical data, model prediction values ​​or other analysis results. The historical resource data set may include a first data set and a second data set, and the first data set and the second data set may represent different time periods, different conditions or different data sources, and are used to compare or comprehensively analyze the effects of different parameter combinations.

[0091] The training unit trains a preset risk assessment model based on a first dataset. This first dataset is typically known historical data or collected data samples, used for model initialization and training. During the training process, the model learns patterns and relationships within the data and adjusts model parameters to best align with the data characteristics and expected output. A preset model is a pre-selected or designed basic model used for risk assessment. These models can be statistical models, machine learning models (such as regression models, decision trees, neural networks, etc.), or other mathematical models, and are selected based on the characteristics and requirements of the specific problem.

[0092] After model training is complete, the verification unit performs an accuracy check on the initial risk assessment model based on a second dataset. By comparing the model's predictions on the second dataset with the actual observations, the model's performance on unknown data can be evaluated, thereby determining its feasibility and effectiveness in practical applications.

[0093] The result of the validation unit is the generation of candidate risk assessment models. These models have been verified to perform well on the second dataset. These candidate models can include different parameter settings for a single model or different types of models, which are used for subsequent model fusion or selection of the final risk assessment model.

[0094] The first assessment module 102 is used to obtain real-time resource data of the first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and the first stage risk assessment model.

[0095] In this embodiment, once the construction of the first-stage risk assessment model is completed, the real-time resource data of the first operating cycle and the previously established model can be used to perform the first security risk assessment. The specific duration of the first operating cycle is not clearly defined in this application, so the appropriate cycle length can be selected based on actual conditions. These real-time resource data are used as input features, and after preprocessing and feature engineering, they are input into the first-stage risk assessment model for analysis. The purpose of this process is to extract key features from multidimensional data to identify potential security risk factors and output corresponding security risk assessment results through the model.

[0096] The second assessment module 103 is configured to obtain real-time noise impact data and dynamically modify the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model.

[0097] In this embodiment, noise impact data refers to interference data that affects the normal operation of the virtual power plant. Noise impact data can cause fluctuations in the virtual power plant's operating status, leading to inaccurate predictions in the first-stage risk assessment model, which characterizes the relationship between the virtual power plant and the monitored system. Therefore, it is necessary to dynamically correct the first-stage risk assessment model based on the noise impact data to obtain the second-stage risk assessment model. For example, at a first moment, the virtual power plant displays an energy reserve of A. However, due to the noise impact of network transmission delay, the actual energy reserve available and usable by the virtual power plant at that moment is far less than A. In this case, if the first-stage risk assessment model only makes predictions based on static data (energy reserve A), it may lead to a misinterpretation of the power plant's operating status and risk assessment. By introducing dynamic corrections, this noise impact data can be captured and responded to in a timely manner, thereby more accurately reflecting actual operating conditions and potential safety risks. Therefore, continuously monitoring and analyzing noise impact data and feeding it back into the risk assessment model ensures that the model can adapt and respond to changes and interference in the real-time environment. This dynamic correction and optimization method can effectively improve the safety management level of the virtual power plant system and ensure its stable operation and reliability under complex environmental conditions.

[0098] The comparison module 104 is used to obtain a second security risk assessment result based on the real-time resource data of the first operation cycle and the second stage risk assessment model, and determine the target risk assessment model based on the comparison result of the first security risk assessment result and the second security risk assessment result.

[0099] In this embodiment, the impact of noise data is bidirectional, meaning it can be either negative or positive. Negative noise data impacts may cause instability or abnormal fluctuations in the virtual power plant's operating status. For example, negative effects from environmental noise or electromagnetic interference may interfere with the accuracy of sensor data or lead to instability in equipment operating conditions. In this case, the first-stage risk assessment model may be misleading and unable to accurately predict the power plant's actual operating conditions and potential safety risks. On the other hand, positive noise data impacts may help improve the operation or safety assessment of the virtual power plant. For example, some noise data may contain sensitive information about environmental changes or equipment status changes, enabling early detection of potential problems or abnormalities. Through effective data analysis and model modification, this information can be utilized to optimize the risk assessment model and enhance the ability to accurately understand and predict the power plant's operating status. Therefore, the bidirectional impact of noise data requires comprehensive consideration and dynamic response during risk assessment. The second safety risk assessment model, which acts as a modification of the first safety risk assessment model, can also be bidirectional. Therefore, judgments must be made based on a comparison of the first and second safety risk assessment results.

[0100] In a feasible implementation, the comparison module includes:

[0101] An optimization direction determination submodule, configured to determine an optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0102] A first optimization submodule is configured to execute a first optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization;

[0103] The second optimization submodule is used to execute a second optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is negative optimization.

[0104] In the present embodiment, the optimization direction determination submodule determines the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on the comparison of the first safety risk assessment result and the second safety risk assessment result. The comparison result may be that the second-stage risk assessment model is superior to the first-stage risk assessment model in some aspects, or that the first-stage risk assessment model is superior to the second-stage risk assessment model in some aspects. When it is determined that the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is positive, that is, the second-stage risk assessment model performs better or there is room for improvement, the first optimization strategy is executed. This may include adjustment of model parameters, further optimization of data features, or improvement of model algorithms to enhance the predictive ability and accuracy of the model. When it is determined that the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is negative, that is, the first-stage risk assessment model performs better or there are defects in the second-stage risk assessment model, the second optimization strategy is executed.

[0105] In a feasible implementation, the first optimization submodule includes:

[0106] A model determination unit, used to determine the second-stage risk assessment model as the main model and the first-stage risk assessment model as the auxiliary model;

[0107] An optimization direction determination unit, used to perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0108] The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

[0109] In this embodiment, when the second-stage risk assessment model is determined as the primary model and the first-stage risk assessment model as the secondary model, the impact of noise data on the operating status of the virtual power plant is degrading. Therefore, after determining the roles of the primary and secondary models, knowledge distillation is used to determine the optimization direction of the secondary model. Knowledge distillation is a model compression technique that transfers knowledge from the primary model to the secondary model, guiding the secondary model towards a preset optimization direction. Then, model fusion is performed based on the secondary model optimized through knowledge distillation and the primary model. The model fusion process may include ensemble learning methods (such as voting and weighted averaging), hierarchical fusion (such as stacking or concatenating multiple models), or other techniques to comprehensively leverage the strengths of each model to obtain the final target risk assessment model. By leveraging the knowledge of the primary model to improve the secondary model, and then through effective model fusion techniques, the overall risk assessment model is optimized and improved.

[0110] In a feasible implementation, the second optimization submodule includes:

[0111] Model determination unit, used to use the first stage risk assessment model as the main model and the second stage risk assessment model as the auxiliary model

[0112] An optimization direction determination unit, used to perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0113] The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

[0114] In this embodiment, when the first-stage risk assessment model is determined as the primary model and the second-stage risk assessment model as the secondary model, the impact of noise data on the operating status of the virtual power plant is an optimization influence. Therefore, after determining the roles of the primary and secondary models, knowledge distillation is used to determine the optimization direction of the secondary model. Knowledge distillation is a model compression technique that transfers knowledge from the primary model to the secondary model, guiding the secondary model towards a preset optimization direction. Then, model fusion is performed based on the secondary model optimized through knowledge distillation and the primary model. The model fusion process may include ensemble learning methods (such as voting, weighted averaging), hierarchical fusion (such as stacking or concatenating multiple models), or other techniques to comprehensively leverage the strengths of each model to obtain the final target risk assessment model. By leveraging the knowledge of the primary model to improve the secondary model, and then through effective model fusion techniques, the overall risk assessment model is optimized and improved.

[0115] The third evaluation module 105 is used to obtain real-time resource data of the second operation cycle of the virtual power plant, and determine the target safety risk assessment result based on the real-time resource data of the second operation cycle and the target risk assessment model.

[0116] In this embodiment, the third assessment module obtains real-time resource data for the second operating cycle of the virtual power plant. The second operating cycle and the first operating cycle constitute a complete test and verification cycle, and uses this data to analyze with a predetermined target risk assessment model to determine the target safety risk assessment result. By collecting and analyzing the real-time resource data of the second operating cycle and inputting it into the target risk assessment model, the final safety risk assessment result and the corresponding processing strategy are obtained. Therefore, the output result of the third assessment module not only provides the safety risk status of the virtual power plant in the current operating cycle, but also provides appropriate measures to optimize and manage the operation of the power plant to ensure that it operates in a safe and efficient state.

[0117] The embodiment of the present application provides a safety risk assessment system based on an artificial intelligence algorithm model. By acquiring the historical resource data of the virtual power plant and the multi-dimensional operating parameters of the system to be monitored, an initial risk assessment model is constructed, and the model is evaluated using real-time resource data in the first operating cycle of the virtual power plant to obtain a preliminary safety risk assessment result. Subsequently, by acquiring real-time noise impact data, the initial model is dynamically corrected to form a more accurate second-stage risk assessment model. Based on the real-time resource data of the first operating cycle and the corrected model, a risk assessment is performed again to obtain the safety risk assessment result of the second stage. By comparing the assessment results of the first and second stages, the final target risk assessment model is determined. Finally, the real-time resource data of the second operating cycle of the virtual power plant is analyzed using the target model to obtain the final target safety risk assessment result. In this way, by gradually optimizing and dynamically adjusting the assessment model, the present application can effectively respond to the management challenges of the dispersion and diversity of virtual power plants, and achieve accurate risk management and real-time regulation.

[0118] Based on the same inventive concept, Figure 2 , shows an evaluation method provided by an embodiment of the present application, which should be configured as the security risk evaluation system based on the artificial intelligence algorithm model proposed in the first aspect of the present application. The evaluation method may include the following steps:

[0119] S201: Obtain historical resource data of the virtual power plant and multi-dimensional operating parameters of the system to be monitored, and build a first-stage risk assessment model based on the real-time resource data and multi-dimensional operating parameters;

[0120] S202: Acquire real-time resource data of a first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and a first-stage risk assessment model;

[0121] S203: Acquire real-time noise impact data, and dynamically modify the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model;

[0122] S204: Obtain a second security risk assessment result based on the real-time resource data of the first operation cycle and the second-stage risk assessment model, and determine a target risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0123] S205: Acquire real-time resource data of the second operation cycle of the virtual power plant, and determine a target safety risk assessment result based on the real-time resource data of the second operation cycle and a target risk assessment model.

[0124] In one feasible implementation, a first-stage risk assessment model is constructed based on real-time resource data and multi-dimensional operating parameters, including:

[0125] Arbitrarily combining at least two target evaluation parameters to obtain a plurality of combination results, wherein each combination result corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set;

[0126] Based on the first data set, a preset model is trained to obtain an initial risk assessment model;

[0127] Based on the second data set, the initial risk assessment model is verified for accuracy to obtain an alternative risk assessment model.

[0128] In a feasible implementation, multiple target risk assessment parameters for risk assessment are screened from the multi-dimensional operating parameters of the system to be monitored, including:

[0129] Perform dimensionality reduction processing on the operating parameters of each dimension, and calculate the information amount corresponding to the operating parameters of each dimension after dimensionality reduction processing;

[0130] According to the amount of information, the multidimensional operating parameters are screened to obtain the target operating parameters.

[0131] In a feasible implementation, multiple target risk assessment parameters for risk assessment are screened from the multi-dimensional operating parameters of the system to be monitored, including:

[0132] A calculation unit, configured to perform dimensionality reduction processing on the operating parameters of each dimension and calculate the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing;

[0133] The screening unit is used to screen the multi-dimensional operating parameters according to the amount of information to obtain the target operating parameters.

[0134] In a feasible implementation, determining the target risk assessment model according to a comparison result of the first security risk assessment result and the second security risk assessment result includes:

[0135] An optimization direction determination submodule, configured to determine an optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result;

[0136] The first optimization submodule is used to execute the first optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization.

[0137] The second optimization submodule is used to execute the second optimization strategy when the optimization direction of the second-stage risk assessment model is negative optimization relative to the first-stage risk assessment model.

[0138] In a feasible implementation, when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization, the first optimization strategy is executed:

[0139] Determine the second-stage risk assessment model as the main model and the first-stage risk assessment model as the auxiliary model;

[0140] Perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0141] Model fusion is performed based on the optimized auxiliary model and main model to obtain the target risk assessment model.

[0142] In a feasible implementation, when the optimization direction of the risk assessment model in the second stage is negative optimization relative to the risk assessment model in the first stage, executing the second optimization strategy includes:

[0143] The first-stage risk assessment model is the main model, and the second-stage risk assessment model is the auxiliary model.

[0144] Perform knowledge distillation on the auxiliary model based on the main model to determine the optimization direction of the auxiliary model;

[0145] Model fusion is performed based on the optimized auxiliary model and main model to obtain the target risk assessment model.

[0146] It should be noted that the specific implementation of the evaluation method in the embodiment of the present application refers to the specific implementation of the security risk assessment system based on the artificial intelligence algorithm model proposed in the first aspect of the embodiment of the present application, and will not be repeated here.

[0147] The above embodiments can be implemented in whole or in part via software, hardware (e.g., circuits), firmware, or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. A computer program product comprises one or more computer instructions or computer programs. When the computer instructions or computer program are loaded or executed on a computer, the processes or functions according to the embodiments of the present application are fully or partially generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable system. The computer program or instructions can be stored in a computer-readable storage medium or transferred from one computer-readable storage medium to another. For example, the computer program or instructions can be transferred from one website, computer, server, or data center to another website, computer, server, or data center via wired means (e.g., infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium accessible by a computer or a data storage device such as a server or data center that contains a collection of one or more available media. Available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media. The semiconductor media can be a solid-state drive.

[0148] It should be understood that the term "and / or" as used herein simply describes a relationship between associated objects, indicating that three possible relationships exist. For example, "A and / or B" can represent: A alone, A and B together, or B alone. A and B can be singular or plural. Furthermore, the character " / " as used herein generally indicates an "or" relationship between the associated objects, but it may also indicate an "and / or" relationship. For specific understanding, please refer to the context.

[0149] In this application, "at least one" means one or more, and "plurality" means two or more. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, "at least one of a, b, or c" can mean: a, b, c, ab, ac, bc, or abc, where a, b, and c can be single or plural.

[0150] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0151] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of 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 beyond the scope of this application.

[0152] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0153] In the several embodiments provided in this application, it should be understood that the disclosed systems, systems and methods can be implemented in other ways. For example, the system embodiments described above are merely illustrative. For example, the division of units is only a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the system or unit can be electrical, mechanical or other forms.

[0154] Units described as separate components may or may not be physically separate, and components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0155] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0156] If the 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 existing technology, or the part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for enabling a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, and other media that can store program code.

[0157] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

Claims

1. A security risk assessment system based on an artificial intelligence algorithm model, characterized in that: Applied to virtual power plants and power demand systems, the security risk assessment system includes: An acquisition module is used to acquire historical resource data of the virtual power plant and multi-dimensional operating parameters of the system to be monitored, and to construct a first-stage risk assessment model based on the historical resource data and the multi-dimensional operating parameters; a first assessment module, configured to obtain real-time resource data of a first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and the first-stage risk assessment model; a second assessment module, configured to obtain real-time noise impact data and dynamically modify the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model; a comparison module, configured to obtain a second security risk assessment result based on the real-time resource data of the first operation cycle and the second-stage risk assessment model, and determine a target risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result; a third assessment module, configured to obtain real-time resource data of a second operation cycle of the virtual power plant, and determine a target safety risk assessment result based on the real-time resource data of the second operation cycle and the target risk assessment model; The comparison module includes: an optimization direction determination submodule, configured to determine an optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result; A first optimization submodule is configured to execute a first optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is forward optimization; A second optimization submodule is configured to execute a second optimization strategy when the optimization direction of the second-stage risk assessment model relative to the first-stage risk assessment model is negative optimization; The first optimization submodule includes: A model determination unit, configured to determine the second-stage risk assessment model as a primary model and the first-stage risk assessment model as a secondary model; an optimization direction determining unit, configured to perform knowledge distillation on the auxiliary model according to the main model to determine an optimization direction of the auxiliary model; The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

2. The evaluation system according to claim 1, wherein: The acquisition module includes: A screening submodule, configured to screen out a plurality of target risk assessment parameters for risk assessment from the multi-dimensional operating parameters of the system to be monitored; A model building submodule, configured to build a plurality of alternative risk assessment models according to the target risk assessment parameters and the real-time resource data; The model fusion submodule is used to perform model fusion on the multiple alternative risk assessment models to construct a first-stage risk assessment model.

3. The evaluation system according to claim 2, wherein: The model building submodule includes: a combining unit, configured to arbitrarily combine at least two of the target risk assessment parameters to obtain a plurality of combination results, wherein each of the combination results corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set; A training unit, configured to train a preset model based on the first data set to obtain an initial risk assessment model; A verification unit is configured to perform accuracy verification on the initial risk assessment model based on the second data set to obtain the alternative risk assessment model.

4. The evaluation system according to claim 2, wherein: The screening submodule includes: a calculation unit, configured to perform dimensionality reduction processing on the operating parameters of each dimension, and calculate the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing; A screening unit is used to screen the multidimensional operating parameters according to the amount of information to obtain the target risk assessment parameters.

5. The evaluation system according to claim 1, wherein: The second optimization submodule includes: Model determination unit, used to use the first stage risk assessment model as the main model and the second stage risk assessment model as the auxiliary model an optimization direction determining unit, configured to perform knowledge distillation on the auxiliary model according to the main model to determine an optimization direction of the auxiliary model; The optimization unit is used to perform model fusion based on the optimized auxiliary model and the main model to obtain the target risk assessment model.

6. A security risk assessment method based on an artificial intelligence algorithm model, characterized in that: The security risk assessment system based on the artificial intelligence algorithm model according to any one of claims 1 to 5, wherein the method comprises: Acquire historical resource data of the virtual power plant and multi-dimensional operating parameters of the system to be monitored, and construct a first-stage risk assessment model based on the historical resource data and the multi-dimensional operating parameters; Acquire real-time resource data of a first operation cycle of the virtual power plant, and obtain a first safety risk assessment result based on the real-time resource data of the first operation cycle and the first-stage risk assessment model; Acquiring real-time noise impact data, and dynamically revising the first-stage risk assessment model based on the real-time noise impact data to obtain a second-stage risk assessment model; Obtaining a second security risk assessment result based on the real-time resource data of the first operation cycle and the second-stage risk assessment model, and determining a target risk assessment model based on a comparison result of the first security risk assessment result and the second security risk assessment result; Acquire real-time resource data of the second operation cycle of the virtual power plant, and determine the target safety risk assessment result based on the real-time resource data of the second operation cycle and the target risk assessment model.

7. The evaluation method according to claim 6, characterized in that The first-stage risk assessment model is constructed based on the historical resource data and the multi-dimensional operating parameters, including: Arbitrarily combining at least two target risk assessment parameters to obtain a plurality of combination results, wherein each combination result corresponds to a historical resource data set, and the historical resource data set includes a first data set and a second data set; Based on the first data set, a preset model is trained to obtain an initial risk assessment model; Based on the second data set, the initial risk assessment model is verified for accuracy to obtain an alternative risk assessment model.

8. The evaluation method according to claim 6, wherein: A plurality of target risk assessment parameters for risk assessment are screened out from the multi-dimensional operating parameters of the system to be monitored, including: Performing dimensionality reduction processing on the operating parameters of each dimension, and calculating the amount of information corresponding to the operating parameters of each dimension after the dimensionality reduction processing; According to the amount of information, the multidimensional operating parameters are screened to obtain the target risk assessment parameters.

Images