Big data-based power operation risk early warning method, system, device and medium

Abstract

The application discloses a power operation risk early warning method, system, device and medium based on big data, belongs to the technical field of operation risk management, and comprises the following steps: collecting multi-source data, preprocessing, extracting data risk features, constructing a risk evolution path model based on the risk features; constructing a risk prediction model to predict future risk changes and dividing risk levels, early warning according to the risk levels; and displaying real-time risk values of operation areas through a visual interface and storing data. The application realizes real-time dynamic management of power operation risks, overcomes the defects of single data dimension analysis, insufficient risk transmission modeling and low real-time performance and prediction accuracy of traditional technologies, effectively improves the ability of power operation risk identification, dynamic evaluation and accurate prediction, provides efficient decision support through hierarchical early warning and operation optimization scheme, and realizes efficient and safe early warning of complex power operation scenes.

Description

Methods, systems, equipment, and media for power operation risk early warning based on big data Technical Field

[0001] This invention relates to the field of operational risk management technology, specifically to a method, system, equipment, and medium for early warning of power operation risks based on big data. Background Technology

[0002] With the rapid development of the power industry and the increasing complexity of its operations, the importance of safety and risk management in power operations has become increasingly prominent. Power operations typically involve high-voltage environments, complex equipment operation, and unpredictable natural conditions. These factors combined make power operation risk a core issue that the industry urgently needs to address. In recent years, the introduction of big data technology has provided new opportunities for power operation risk management. By collecting and analyzing multi-source data such as personnel, equipment operating status, and environmental conditions in real time, big data technology can effectively identify potential safety hazards and provide dynamic risk assessment and early warning functions.

[0003] Specifically, big data technology has been widely applied in scenarios such as equipment fault diagnosis, operational status monitoring, and risk warning. For example, by collecting voltage, current, and temperature data from equipment through IoT sensors and combining this data with historical data analysis, potential equipment failures can be predicted. Simultaneously, using dynamically monitored environmental data (such as wind speed and temperature) to conduct environmental risk assessments of the work area provides auxiliary decision support for power operations. However, these technologies typically focus only on single-dimensional data (such as equipment status or environmental risk), making it difficult to comprehensively and dynamically integrate multi-source data for accurate assessment of the overall risks of power operations. Furthermore, there is still significant room for improvement in the real-time performance and accuracy of existing technologies in areas such as risk transmission, dynamic changes over time series, and spatial visualization. Summary of the Invention

[0004] In view of the above-mentioned problems, the present invention is proposed.

[0005] Therefore, this invention addresses the problem that these technologies typically focus only on single-dimensional data, making it difficult to comprehensively and dynamically integrate multi-source data to accurately assess the overall risks of power operations.

[0006] To address the aforementioned technical problems, this invention provides the following technical solution: a power operation risk early warning method based on big data, comprising,

[0007] Collect and preprocess multi-source data, extract data risk characteristics, and construct a risk evolution path model based on the risk characteristics; construct a risk prediction model to predict future risk changes and classify risk levels, and issue early warnings based on risk levels; based on the risk characteristics, construct the risk evolution path model and risk prediction model, display the real-time risk values ​​of the work area through a visual interface, and store the data.

[0008] As a preferred embodiment of the power operation risk early warning method based on big data described in this invention, the step of collecting multi-source data and preprocessing it to extract data risk characteristics refers to collecting data on operators, equipment, environment, and historical accidents, and standardizing and cleaning abnormal data from the collected data.

[0009] Principal component analysis was performed on the preprocessed data to extract data features, and the risk features were classified into static factors and dynamic factors using support vector machine.

[0010] As a preferred embodiment of the big data-based power operation risk early warning method of the present invention, the risk feature-based risk evolution path model construction includes constructing a dynamic Bayesian network based on the risk feature-based risk evolution path model to describe the conditional probability relationship between risk factors.

[0011] Static factors are used as base nodes, and dynamic factors are used as dependent nodes. Dynamic factors and static factors are associated to form a preliminary relationship between factors.

[0012] Based on domain knowledge and data analysis, the set of parent nodes A for each dynamic factor is determined. The correlation between factors is calculated through mutual information. A threshold B is set, and parent nodes with high correlation greater than the threshold B are selected to form a preliminary factor dependency structure.

[0013] The time step is set according to the data sampling frequency. The time step is added to the Bayesian network, and the state of each dynamic factor is defined to depend not only on the current set of parent nodes, but also on its own state in the previous time step.

[0014] Extend the conditional probability formula to form a temporal Bayesian network;

[0015] A risk evolution path model is constructed based on risk characteristics. The state transition probability between time steps is calculated using historical data. A state transition matrix is ​​generated and a time-series Bayesian network is initialized. Cross-validation is used to check the network prediction accuracy after introducing time series data. Through factor correlation analysis, the gain of time series dependence on dynamic factor state prediction is verified, and a time-series Bayesian network initialized with time series dependence is obtained.

[0016] Time series data is obtained through a temporal Bayesian network. The time series data is segmented into subsamples according to time windows. The state transition relationship and conditional probability table of the time series are retained in each subsample to obtain a subsample dataset with time series dependency characteristics.

[0017] Based on the newly added time series sub-samples, multiple sub-sample datasets are generated. A dynamic Bayesian network is trained independently on each sub-sample, and the network structure and conditional probability table of each sub-sample are output.

[0018] Calculate the posterior probability under subsamples, integrate the network results of subsamples into the final network through model averaging, and integrate the training results of each subsample through weighted averaging to generate the global network;

[0019] Using historical data, the conditional probability of each factor is calculated based on the set of parent nodes, a conditional probability table is constructed, and cross-validation is used to evaluate the prediction accuracy of the dynamic Bayesian network, resulting in the final robust dynamic Bayesian network.

[0020] By using the conditional probability information of each factor in the conditional probability table of the dynamic Bayesian network, the weight of the conditional probability is calculated, a threshold C is set, factors with weights greater than the threshold are screened, and L1 regularization is used to further screen risk factors to obtain key factors.

[0021] A factor set is generated based on the key factors selected through screening, initial weights are assigned to the key factors, and the overall risk value is obtained based on the key factors and initial weights.

[0022] The correlation between the selected key factors and the historical overall risk is calculated, and the preliminary weights are calculated based on the correlation values ​​of individual factors with the overall risk and the sum of the correlation values ​​of all factors with the overall risk.

[0023] After normalizing the obtained correlation, the weight allocation is further optimized and an optimized weight set is generated by using an adversarial sampling method.

[0024] Using the selected set of factors and their corresponding weights, a fractional differential equation is established to describe the risk propagation path, and fractional differential operators are used. Modeling the memory effect of risk propagation;

[0025] The initial values ​​of the fractional-order parameters are set, and the dynamic changes of the overall risk value are calculated iteratively to simulate the risk propagation path.

[0026] As a preferred embodiment of the power operation risk early warning method based on big data described in this invention, the risk feature construction risk evolution path model further includes: collecting real-time operation data, integrating the real-time data into a fractional differential equation model, updating the current risk propagation status, calculating the gradient value of the loss function through real-time data, updating the factor weights, and dynamically adjusting the weights using an optimization formula.

[0027] The fractional-order parameter α is dynamically adjusted according to changes in real-time data. The overall risk value is output according to the risk propagation path model. A threshold O is set. Points with an overall risk value exceeding the threshold O are marked as key risk points in the work area. The area where the key risk points are located is designated as a high-risk area, and the other area is designated as a low-risk area.

[0028] Kriging interpolation is used to calculate the target location based on the distribution of key risk points. Risk value ;

[0029] The calculation results are mapped to geospatial data to generate a risk heat map, which visually displays high-risk and low-risk areas using color gradients.

[0030] As a preferred embodiment of the big data-based power operation risk early warning method described in this invention, the risk prediction model includes: constructing a risk prediction model to predict future risk changes and classifying risk levels; using an LSTM model to predict path risk changes; inputting historical risk features as a training set into the LSTM model for model training; defining a loss function and an Adam optimizer for iterative optimization of model parameters; stopping iteration and outputting model parameters to update the LSTM model when the loss of the LSTM model no longer decreases significantly during continuous iteration; and inputting real-time path risk scores into the LSTM model to obtain future risk changes; wherein the Adam optimizer is an optimizer based on a gradient-based adaptive optimization algorithm, and the LSTM model is a long short-term memory network model.

[0031] Set risk thresholds M and N, where M>N. If the predicted risk change is less than threshold N, it is considered low risk. If the predicted risk change is greater than or equal to threshold N and less than threshold M, it is considered medium risk. If the predicted risk change is less than threshold M, it is considered high risk.

[0032] As a preferred embodiment of the big data-based power operation risk early warning method described in this invention, the risk prediction model further includes: issuing early warnings based on risk levels; if the risk level is high, suspending operations, notifying management personnel, and initiating an emergency response; if the risk level is medium, optimizing operation time and equipment allocation; and if the risk level is low, conducting routine monitoring without adjustment.

[0033] As a preferred embodiment of the power operation risk early warning method based on big data described in this invention, the step of displaying the real-time risk value of the operation area through a visual interface refers to using the Sketch tool to create a visual interface, displaying the risk distribution and risk heat map of each operation area, constructing and displaying a risk trend chart, and highlighting the risk changes at key time nodes.

[0034] The key time nodes are identified by comparing them with historical risk data, identifying abnormal patterns in the current operation, and marking them as key time nodes.

[0035] Another objective of this invention is to provide a power operation risk early warning system based on big data.

[0036] To solve the above-mentioned technical problems, the present invention provides the following technical solution: a power operation risk early warning system based on big data, comprising: a data collection module, a path construction module, a prediction module, an early warning module, and a display and storage module;

[0037] The data collection module is used to collect multi-source data and perform preprocessing.

[0038] The path construction module is used to extract data risk features and build a risk evolution path model based on these features.

[0039] The prediction module is used to build risk prediction models to predict future risk changes and classify risk levels;

[0040] The early warning module is used to issue warnings based on risk levels.

[0041] The display storage module is used to show the real-time risk values ​​of the work area and store the data through a visual interface.

[0042] The present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the power operation risk early warning method based on big data.

[0043] The present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that the computer program, when executed by a processor, implements the steps of the power operation risk early warning method based on big data.

[0044] The beneficial effects of this invention are as follows: This invention realizes real-time dynamic management of power operation risks, overcomes the shortcomings of traditional technologies such as single-data-dimensional analysis, insufficient risk transmission modeling, and low real-time performance and prediction accuracy, effectively improves the ability to identify, dynamically assess, and accurately predict power operation risks, and provides efficient decision support through hierarchical early warning and operation optimization schemes, thus realizing efficient and safe management of complex power operation scenarios. Attached Figure Description

[0045] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 is a flowchart of a power operation risk early warning method based on big data provided in an embodiment of the present invention.

[0047] Figure 2 is a schematic diagram of a power operation risk early warning system based on big data provided in an embodiment of the present invention. Detailed Implementation

[0048] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0049] Example 1, referring to Figure 1, is an embodiment of the present invention. This embodiment provides a power operation risk early warning method based on big data, including:

[0050] S1. Collect multi-source data and preprocess it to extract data risk characteristics and build a risk evolution path model based on the risk characteristics;

[0051] Specifically, collecting and preprocessing multi-source data, extracting data risk characteristics refers to collecting data on workers, equipment, environment, and historical accidents, and standardizing and cleaning outlier data from the collected data.

[0052] The worker data includes information such as location, heart rate, and frequency of behavior collected through smart terminals worn by workers (such as smart bracelets or helmet sensors);

[0053] The device data includes real-time device status (current, voltage, temperature, etc.) and fault history records acquired through IoT sensors;

[0054] The environmental data includes data such as temperature, humidity, wind speed, and rainfall obtained through the meteorological station API;

[0055] The historical accident data includes accident type and cause analysis data extracted from the database;

[0056] Principal component analysis was performed on the preprocessed data to extract data features.

[0057] The features include operator characteristics, equipment characteristics, and environmental characteristics;

[0058] The characteristics of the operators include heart rate changes and displacement frequency; the characteristics of the equipment include operating temperature and current fluctuation amplitude; and the environmental characteristics include wind speed changes and abnormal humidity.

[0059] Risk characteristics are classified into static factors and dynamic factors using support vector machines.

[0060] By comprehensively collecting and standardizing multi-source data, extracting features using principal component analysis, classifying factors using support vector machines, and conducting real-time risk assessment and early warning through dynamic modeling, a complete and efficient method for power operation risk management has been developed. Compared with existing technologies, this invention has significant advantages in terms of dynamism, real-time performance, and accuracy, especially in dynamic risk management of complex power operation scenarios, demonstrating superior performance and broad applicability.

[0061] Furthermore, constructing a risk evolution path model based on risk characteristics refers to building a dynamic Bayesian network to describe the conditional probabilistic relationships between risk factors:

[0062] Static factors (such as equipment status and work plans, which change little over time) are used as basic nodes, and dynamic factors (such as environmental parameters: temperature and humidity, and work behavior, which fluctuate greatly over time) are used as dependent nodes. Dynamic factors and static factors are linked to form a preliminary relationship between factors.

[0063] By distinguishing between static and dynamic factors and modeling their relationships, the ability to express the relationships between factors is enhanced, avoiding the shortcomings of traditional static models in dynamic scenarios.

[0064] Based on domain knowledge and data analysis, the set of parent nodes A for each dynamic factor is determined. The correlation between factors is calculated through mutual information. A threshold B is set through experiments, and parent nodes with high correlation greater than the threshold B are selected to form a preliminary factor dependency structure.

[0065] By calculating mutual information, parent nodes that are highly correlated with dynamic factors are selected, avoiding interference from irrelevant variables on modeling accuracy and improving the rationality of the network structure.

[0066] The time step is set according to the data sampling frequency. The time step is added to the Bayesian network, and the state of each dynamic factor is defined to depend not only on the current set of parent nodes, but also on its own state in the previous time step.

[0067] Extending the conditional probability formula, we can form a temporal Bayesian network:

[0068] ,

[0069] In the formula, It is the joint probability distribution at time t, describing the state of global risk factors at the current time step t. It is the state of the i-th factor at time t. It is the set of parent nodes of the i-th factor at the current time step. It represents the state of the i-th factor at time t−1, used to reflect the dynamics in the time series, where n is the total number of risk factors, i is the variable index, and t is the variable index at the time step.

[0070] By optimizing the network through time series dependencies, dynamic Bayesian networks can more accurately predict the state changes of risk factors, making them suitable for complex time-series scenarios in power operations.

[0071] The introduction of time steps in temporal Bayesian networks enables risk assessment models to capture the dynamic changes of factors in time series, thereby improving the real-time prediction capability of power operation risks.

[0072] A risk evolution path model is constructed based on risk characteristics. The state transition probability between time steps is calculated using historical data. A state transition matrix is ​​generated and a time-series Bayesian network is initialized. Cross-validation is used to check the network prediction accuracy after introducing time series data. Through factor correlation analysis, the gain of time series dependence on dynamic factor state prediction is verified, and a time-series Bayesian network initialized with time series dependence is obtained.

[0073] Time series data is obtained through a temporal Bayesian network. The time series data is segmented into subsamples according to time windows. The state transition relationship and conditional probability table of the time series are retained in each subsample to obtain a subsample dataset with time series dependency characteristics.

[0074] Based on the newly added time series sub-samples, multiple sub-sample datasets are generated. A dynamic Bayesian network is trained independently on each sub-sample, and the network structure and conditional probability table of each sub-sample are output.

[0075] Calculate the posterior probability for a subsample:

[0076] ,

[0077] In the formula, Subsample In network structure The posterior probability is given below, where x is the subsample. A single data point in the data. It is a network structure The conditional probability of the next single data point x;

[0078] The network results of the subsamples are integrated into the final network through model averaging:

[0079] ,

[0080] In the formula, This represents the factor state conditional probability of the global network model at time t, where T is the number of subsamples and K is the number of network structures per subsample. It is the posterior probability of a subsample. It is the factor conditional probability given a subsample and its network structure;

[0081] The training results of each subsample are combined by weighted averaging to generate a global network;

[0082] A risk evolution path model is constructed based on risk characteristics. Using historical data, the conditional probability of each factor is calculated according to the set of parent nodes. A conditional probability table is constructed, and cross-validation is used to evaluate the prediction accuracy of the dynamic Bayesian network, resulting in a final robust dynamic Bayesian network.

[0083] The weights of the conditional probabilities are calculated using the conditional probability information of each factor in the conditional probability table of the dynamic Bayesian network.

[0084] A threshold C is set based on the grid search method, and factors with weights greater than the threshold are screened. L1 regularization is then used to further screen risk factors to obtain key factors.

[0085] A factor set is generated based on the key factors selected through screening, initial weights are assigned to the key factors, and the overall risk value is obtained based on the key factors and initial weights.

[0086] Based on the key factors selected, the correlation between the factors and the historical overall risk is calculated using the Pearson correlation coefficient. Preliminary weights are calculated based on the correlation values ​​of individual factors with the overall risk and the sum of the correlation values ​​of all factors with the overall risk.

[0087] After normalizing the obtained correlation, the weight allocation is further optimized and an optimized weight set is generated by using an adversarial sampling method.

[0088] Using the selected set of factors and their corresponding weights, a fractional differential equation is established to describe the risk propagation path, and fractional differential operators are used. Modeling the memory effect of risk propagation:

[0089] ,

[0090] In the formula, R(t) is the overall risk value at time t. It is the weight of the i-th factor. It is the state value of the i-th factor at time t, where n represents the total number of risk factors and i is the variable index;

[0091] Set initial values ​​for the fractional-order parameters and simulate the risk propagation path by iteratively calculating the dynamic changes in the overall risk value:

[0092] ,

[0093] In the formula, R(t+1) is the total risk value at future time t+1. It is the gamma function, used to normalize the results of fractional derivative calculations. It is the memory effect weight, which controls the influence of historical states on the current risk value. It is a historic moment The rate of change of the risk value, z is the fractional order;

[0094] The fractional-order differential model retains the influence of historical states on the current risk value, making the simulation of risk propagation paths more closely resemble reality. Dynamically adjusting the fractional-order parameters can adapt to different propagation speeds based on changes in real-time data, thus improving the applicability of the model.

[0095] A risk evolution path model is constructed based on risk characteristics. Real-time operational data is collected, including key factor values ​​(equipment status, environmental parameters) and operational behavior data. The real-time data is then integrated into a fractional differential equation model to update the current risk propagation status.

[0096] ,

[0097] In the formula, It is the amount of change in factor state caused by real-time data. It is the state value of the i-th factor at time t;

[0098] The gradient of the loss function is calculated using real-time data, the factor weights are updated, and the weights are dynamically adjusted using an optimization formula.

[0099] ,

[0100] In the formula, It is the factor weight at time t+1. It is the factor weight of time t. It is the learning rate, used to control the magnitude of weight adjustments. The loss function L is the weight. The gradient;

[0101] The fractional-order parameter α is dynamically adjusted based on changes in real-time data. If the real-time data indicates that the propagation speed is increasing, α is increased; if the propagation speed is decreasing, α is decreased.

[0102] Based on the risk characteristics, a risk evolution path model is constructed. The overall risk value is output according to the risk propagation path model. A threshold O is set by cross-validation. Points with an overall risk value exceeding the threshold O are marked as key risk points in the work area. The area where the key risk points are located is designated as a high-risk area, and the rest is designated as a low-risk area.

[0103] Kriging interpolation is used to calculate the target location based on the distribution of key risk points. Risk value :

[0104] ,

[0105] In the formula, It is a known location The risk value, These are the interpolation weights, calculated using the semivariance function, where u is the total number of known data points and l is the index of each known point.

[0106] The calculation results are mapped to geospatial data to generate a risk heat map, which visually displays high-risk and low-risk areas using color gradients.

[0107] A risk evolution path model is constructed based on risk characteristics. By combining dynamic Bayesian networks with fractional differential equations, dynamic modeling of risk factors and simulation of propagation paths are achieved. Mutual information filtering and time-series dependency optimization enhance the model's predictive accuracy and dynamic adaptability. The fractional model captures the memory effect of risk propagation, while Kriging interpolation provides a high-precision spatial risk representation. The overall approach, from data modeling to dynamic prediction and visualization, comprehensively improves the real-time performance, accuracy, and decision support capabilities of power operation risk management.

[0108] S2. Construct a risk prediction model to predict future risk changes and classify risk levels, and issue early warnings based on risk levels;

[0109] Specifically, constructing a risk prediction model to predict future risk changes and classify risk levels involves using an LSTM model to predict path risk changes. Historical risk features are used as the training set and input into the LSTM model for model training. A loss function and Adam optimizer are defined to iteratively optimize the model parameters. When the loss of the LSTM model no longer decreases significantly during continuous iteration, the iteration stops, the model parameters are output, and the LSTM model is updated. Real-time path risk scores are input into the LSTM model to obtain future risk changes.

[0110] By statistically analyzing historical risk data, risk thresholds M and N are set, where M > N. If the predicted risk change is less than threshold N, it is considered low risk; if the predicted risk change is greater than or equal to threshold N and less than threshold M, it is considered medium risk; and if the predicted risk change is less than threshold M, it is considered high risk.

[0111] This invention uses an LSTM model to predict changes in path risk. By defining a loss function and optimizing model parameters using the Adam optimizer, the accuracy and efficiency of risk prediction are significantly improved. Through a risk grading mechanism, dynamic risk classification and management can be achieved, providing a basis for matching adaptive strategies to different risk levels. Finally, this invention combines historical risk characteristics with real-time path scoring to achieve dynamic prediction of future risk changes. It possesses real-time performance, accuracy, and high adaptability, effectively addressing the shortcomings of existing technologies in dynamic prediction and real-time response, and significantly improving the scientific rigor and efficiency of risk management.

[0112] Furthermore, warnings are issued based on risk levels. If the risk level is high, operations are suspended, management personnel are notified, and an emergency response is initiated. If the risk level is medium, operation time and equipment allocation are optimized. If the risk level is low, routine monitoring is conducted without any adjustments.

[0113] Through efficient risk level classification and dynamic early warning mechanisms, combined with differentiated handling strategies for different risk levels (such as emergency response for high-risk, optimized operation for medium-risk, and routine monitoring for low-risk), dynamic, precise, and efficient management of the entire power operation process is achieved. Especially in high-risk scenarios, measures such as quickly suspending operations and activating emergency plans can effectively prevent major safety accidents. In medium-risk scenarios, optimizing work plans and equipment allocation can reduce the probability of risk occurrence while improving operational efficiency. In low-risk scenarios, routine monitoring provides stable data support, ensuring that risks remain within a controllable range.

[0114] S3. Based on risk characteristics, construct risk evolution path models and risk prediction models, display real-time risk values ​​of the work area through a visual interface, and store the data.

[0115] Specifically, displaying real-time risk values ​​of work areas through a visual interface involves using the Sketch tool to create a visual interface that displays the risk distribution and risk heat map of each work area, constructs and displays risk trend charts, and highlights risk changes at key time points.

[0116] The key time points refer to identifying abnormal patterns in the current operation by comparing them with historical risk data and marking them as key time nodes.

[0117] By updating risk distribution maps and trend charts in real time, the system can dynamically reflect changes in risk within the work area, helping managers quickly grasp the on-site risk status. Combining historical and real-time data, key time points are identified and marked on the risk trend chart, providing managers with accurate risk warnings and forward-looking decision-making support. Through a visual interface, users can flexibly filter and query risk data for specific areas or times, understanding complex risk information without requiring specialized background. The comprehensive display of risk distribution maps, trend charts, and key time points allows managers to quickly locate high-risk areas, identify risk fluctuation points, and optimize work plans based on chart information.

[0118] Furthermore, data storage refers to storing the collected operational data and the operational risk results generated from the analysis into a database, backing up the stored data to the cloud, and regularly performing integrity checks and implementing access control on the stored and backed-up data.

[0119] By employing four core steps—database storage, cloud backup, integrity checks, and access control—the reliability, security, and availability of data management are significantly improved. Once data is stored in the database, it enables efficient data querying and classification capabilities, providing stable data support for subsequent risk assessment and modeling. Cloud backup ensures disaster recovery capabilities, enabling rapid recovery and business continuity even in extreme situations. Integrity checks further enhance the reliability of stored data, preventing biased analysis results due to data errors. Finally, the introduction of access control, through strict permission management and monitoring mechanisms, effectively protects the confidentiality and integrity of data.

[0120] Example 2, referring to Figure 2, is an embodiment of the present invention. This embodiment provides a power operation risk early warning system based on big data, including: a data collection module, a path construction module, a prediction module, an early warning module, and a display and storage module;

[0121] The data collection module is used to collect multi-source data and perform preprocessing.

[0122] The path construction module is used to extract data risk features and build a risk evolution path model based on these features.

[0123] The prediction module is used to build risk prediction models to predict future risk changes and classify risk levels;

[0124] The early warning module is used to issue warnings based on risk levels.

[0125] The display storage module is used to show the real-time risk values ​​of the work area and store the data through a visual interface.

[0126] This embodiment also provides an electronic device applicable to the power operation risk warning method based on big data, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to realize the power operation risk warning method based on big data as proposed in the above embodiment.

[0127] This embodiment also provides a storage medium on which a computer program is stored. When the program is executed by a processor, it implements the power operation risk early warning method based on big data as proposed in the above embodiments.

[0128] The storage medium proposed in this embodiment and the method for implementing power operation risk early warning based on big data proposed in the above embodiments belong to the same inventive concept. Technical details not described in detail in this embodiment can be found in the above embodiments, and this embodiment has the same beneficial effects as the above embodiments.

[0129] Based on the above description of the implementation methods, those skilled in the art can clearly understand that the present invention can be implemented using software and necessary general-purpose hardware, and of course, it can also be implemented using hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as a computer floppy disk, read-only memory (ROM), random access memory (RAM), flash memory, hard disk, or optical disk, etc., including several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods of the various embodiments of the present invention.

[0130] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A power operation risk early warning method based on big data, characterized by: This includes collecting and preprocessing multi-source data, extracting data risk characteristics, constructing a risk evolution path model based on the risk characteristics, constructing a risk prediction model to predict future risk changes and classifying risk levels, and issuing early warnings based on the risk levels. The risk evolution path model and risk prediction model display the real-time risk values ​​of the work area through a visual interface and store the data. The risk evolution path model based on risk characteristics includes constructing a dynamic Bayesian network to describe the conditional probability relationship between risk factors. By using the conditional probability information of each factor in the conditional probability table of the dynamic Bayesian network, the weight of the conditional probability is calculated, a threshold C is set, factors with weights greater than the threshold are screened, and L1 regularization is used to further screen risk factors to obtain key factors. A factor set is generated based on the key factors selected through screening, initial weights are assigned to the key factors, and the overall risk value is obtained based on the key factors and initial weights. Based on the selected key factors, the correlation between the factors and historical overall risk is calculated. Preliminary weights are calculated based on the correlation values ​​of individual factors with overall risk and the sum of the correlation values ​​of all factors with overall risk. After normalizing the obtained correlations, the weight allocation is further optimized using an adversarial sampling method to generate an optimized weight set. Using the selected factor set and corresponding weights, a fractional differential equation is established to describe the risk propagation path, and fractional differential operators are used. Modeling the memory effect of risk propagation: In the formula, R(t) is the overall risk value at time t. It is the weight of the i-th factor. Let be the state value of the i-th factor at time t, n represent the total number of risk factors, and i be the variable index; set the initial values ​​of the fractional-order parameters and simulate the dynamic changes of the overall risk value through iterative calculation: In the formula, R(t+1) is the total risk value at future time t+1. It is the gamma function, used to normalize the results of fractional derivative calculations. It is the memory effect weight, which controls the influence of historical states on the current risk value. It is a historic moment The rate of change of the risk value, z is the fractional order; Collect real-time operational data, integrate the real-time data into the fractional differential equation model, update the current risk propagation status, calculate the gradient value of the loss function through real-time data, and update the factor weights; dynamically adjust the fractional parameters according to the changes in real-time data, output the overall risk value according to the risk propagation path model, set a threshold, mark the points whose overall risk value exceeds the threshold 0 as key risk points in the operational area, and designate the area where the key risk points are located as a high-risk area, otherwise as a low-risk area. Kriging interpolation is used to calculate the risk value of the target location based on the distribution of key risk points, and a risk heatmap is generated.

2. The power operation risk early warning method based on big data as described in claim 1, characterized in that: The collection of multi-source data and preprocessing, and the extraction of data risk characteristics, refers to the collection of operator data, equipment data, environmental data and historical accident data, and the standardization and abnormal data cleaning of the collected data. Principal component analysis was performed on the preprocessed data to extract data features, and the risk features were classified into static factors and dynamic factors using support vector machine.

3. The power operation risk early warning method based on big data as described in claim 2, characterized in that: The risk evolution path model constructed based on risk characteristics includes: constructing a dynamic Bayesian network to describe the conditional probability relationships between risk factors; using static factors as base nodes and dynamic factors as dependent nodes, linking dynamic and static factors to form preliminary relationships between factors; determining the set of parent nodes A for each dynamic factor based on domain knowledge and data analysis, calculating the correlation between factors through mutual information, setting a threshold B, and filtering out parent nodes with high correlation greater than the threshold B to form a preliminary factor dependency structure; setting the time step according to the data sampling frequency, adding the time step to the Bayesian network, defining that the state of each dynamic factor depends not only on the current set of parent nodes but also on its own state in the previous time step; and extending the conditional probability formula to form a temporal Bayesian network. In the formula, It is the joint probability distribution at time t, describing the state of global risk factors at the current time step t. It is the state of the i-th factor at time t. It is the set of parent nodes of the i-th factor at the current time step. It is time The state of the i-th factor at time n is used to reflect the dynamics in the time series, n is the total number of risk factors, i is the variable index, and t is the variable index of the time step. The introduction of the time step in the temporal Bayesian network enables the risk assessment model to capture the dynamic changes of factors in the time series. A risk evolution path model is constructed based on risk characteristics. The state transition probabilities between time steps are calculated using historical data, a state transition matrix is ​​generated, and a temporal Bayesian network is initialized. Cross-validation is used to check the network's prediction accuracy after introducing time series data. Factor correlation analysis is used to verify the gain of time series dependence on dynamic factor state prediction, resulting in a temporal Bayesian network initialized with time series dependence. Time series data is obtained through the temporal Bayesian network, and the time series data is segmented according to time windows as subsamples. The state transition relationships and conditional probability tables of the time series are retained in each subsample, resulting in a subsample dataset with time series dependence characteristics. Multiple subsample datasets are generated based on the newly added time series subsamples. A dynamic Bayesian network is trained independently on each subsample, and the network structure and conditional probability table of each subsample are output. Calculate the posterior probability for a subsample: In the formula, Subsample In network structure The posterior probability is given below, where x is the subsample. A single data point in the data. It is a network structure The conditional probability of a single data point x; the network results of the subsamples are integrated into the final network through model averaging: In the formula, This represents the factor state conditional probability of the global network model at time t, where T is the number of subsamples and K is the number of network structures per subsample. It is the posterior probability of a subsample. Given a subsample and its network structure, the conditional probabilities of the factors are calculated. The training results of each subsample are integrated by weighted averaging to generate a global network. Using historical data, the conditional probabilities of each factor are calculated based on the set of parent nodes to construct a conditional probability table. Cross-validation is used to evaluate the prediction accuracy of the dynamic Bayesian network to obtain the final robust dynamic Bayesian network.

4. The power operation risk early warning method based on big data as described in claim 3, characterized in that: The risk evolution path model based on risk characteristics also includes collecting real-time operational data, integrating the real-time data into a fractional differential equation model, and updating the current risk propagation status. In the formula, It is the amount of change in factor state caused by real-time data. This represents the state value of the i-th factor at time t; the gradient of the loss function is calculated using real-time data, the factor weights are updated, and the weights are dynamically adjusted using an optimization formula: In the formula, It is the factor weight at time t+1. It is the factor weight of time t. It is the learning rate, used to control the magnitude of weight adjustments. The loss function L is the weight. The gradient is calculated; the fractional-order parameter α is dynamically adjusted based on changes in real-time data. If real-time data indicates that the propagation speed is accelerating, α is increased; if the propagation speed is slowing down, α is decreased. The overall risk value is output based on the risk propagation path model, a threshold O is set, and points whose overall risk value exceeds the threshold O are marked as key risk points in the work area. The area where the key risk points are located is designated as a high-risk area, and the area where the key risk points are located is designated as a low-risk area. Kriging interpolation is used to calculate the target location based on the distribution of key risk points. Risk value ： In the formula, It is a known location The risk value, The interpolation weights are calculated using the semivariance function, where u is the total number of known data points and l is the index of the known points. The calculation results are mapped to geographic space to generate a risk heat map, which visually displays high-risk and low-risk areas through color gradients.

5. The power operation risk early warning method based on big data as described in claim 4, characterized in that: The risk prediction model includes: constructing a risk prediction model to predict future risk changes and classifying risk levels; using an LSTM model to predict path risk changes; inputting historical risk features as a training set into the LSTM model for model training; defining a loss function and an Adam optimizer for iterative optimization of model parameters; stopping iteration and outputting model parameters to update the LSTM model when the loss of the LSTM model no longer decreases significantly during continuous iteration; and inputting real-time path risk scores into the LSTM model to obtain future risk changes. Set risk thresholds M and N, where M>N. If the predicted risk change is less than threshold N, it is considered low risk. If the predicted risk change is greater than or equal to threshold N and less than threshold M, it is considered medium risk. If the predicted risk change is less than threshold M, it is considered high risk.

6. The power operation risk early warning method based on big data as described in claim 5, characterized in that: The risk prediction model also includes issuing early warnings based on risk levels. If the risk level is high, operations are suspended, management personnel are notified, and an emergency response is initiated. If the risk level is medium, operation time and equipment allocation are optimized. If the risk level is low, routine monitoring is conducted without any adjustments.

7. The power operation risk early warning method based on big data as described in claim 6, characterized in that: The provision of a visual interface to display real-time risk values ​​of work areas refers to using the Sketch tool to create a visual interface that displays the risk distribution and risk heat map of each work area, constructs and displays risk trend charts, and highlights risk changes at key time points. The key time points are identified by comparing with historical risk data, identifying abnormal patterns in the current operation and marking them as key time points.

8. A power operation risk early warning system based on big data, employing the power operation risk early warning method based on big data as described in any one of claims 1 to 7, characterized in that, include: The system includes a data collection module, a path construction module, a prediction module, an early warning module, and a display and storage module. The data collection module is used to collect multi-source data and perform preprocessing. The path construction module is used to extract data risk characteristics and build a risk evolution path model based on the risk characteristics; the prediction module is used to build a risk prediction model, predict future risk changes, and classify risk levels. The early warning module is used to issue warnings based on risk levels. The display storage module is used to show the real-time risk values ​​of the work area and store the data through a visual interface.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the power operation risk early warning method based on big data as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the power operation risk early warning method based on big data as described in any one of claims 1 to 7.

Images