Power transmission line earthquake risk prediction method and system driven by multi-source data intelligence

Abstract

The present application relates to the technical field of power system and structure anti-seismic safety, and provides a multi-source data intelligent driving power transmission line earthquake risk prediction method and system to solve the problems of single applicable object, inability to dynamically update and lack of risk closed-loop management in the prior art. The multi-source data intelligent driving power transmission line earthquake risk prediction method comprises: using a continuously updated hybrid deep learning model to process actual seismic wave time history, predicting the top displacement time history of the power transmission line as the dynamic response index of the power transmission line; comparing the predicted top displacement time history of the power transmission line with the known structural performance limit value of the power transmission line, judging the structural performance state of the power transmission line and the corresponding risk level, and simultaneously using the vulnerability curve method to convert the predicted top displacement time history into the occurrence probability of different structural performance states; the method has the characteristics of high prediction accuracy and strong adaptability, and can realize real-time risk early warning.

Description

Technical Field

[0001] This invention relates to the field of seismic safety technology for power systems and structures, and in particular to a multi-source data-driven intelligent method and system for predicting the seismic risk of transmission lines. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Transmission lines are a crucial component of power systems. Their towers and foundations are susceptible to damage and even collapse under seismic loads, potentially triggering large-scale power outages and secondary disasters. Current seismic assessment methods primarily rely on finite element numerical simulations or empirical formulas, often neglecting the integration of experimental and field monitoring data, resulting in limited prediction accuracy. Furthermore, traditional methods are mostly offline static models, lacking dynamic update mechanisms, making them ill-suited to complex, time-varying seismic inputs and actual environmental conditions, thus limiting the real-time nature and reliability of risk warnings.

[0004] Transmission lines are typical long-line, distributed structural systems. Their seismic response not only exhibits significant temporal correlation but also strong spatial correlation and regional variability. Existing Long Short-Term Memory (LSTM) or Transformer models are mostly designed for single measurement points, single components, or single structures, making it difficult to simultaneously characterize the coupling relationship between "local nonlinear response and overall risk evolution" in multi-span tower-line systems. Moreover, earthquake risk prediction is not a one-time offline response prediction problem but requires a continuously updated dynamic process. Before, during, and after an earthquake, the state parameters, environmental conditions, and operating conditions of transmission lines continuously change. In existing research, LSTM and Transformer models are usually based on fixed training sets for offline modeling, lacking online updates and model adaptation mechanisms for new earthquake and monitoring data. This makes it difficult to meet the needs of engineering systems for real-time risk assessment and early warning, as well as to form a closed-loop management mechanism of "prediction-assessment-early warning-feedback update." Summary of the Invention

[0005] To address the aforementioned technical problems, this invention provides a multi-source data-driven intelligent method and system for predicting earthquake risks in transmission lines. This method features high prediction accuracy, strong adaptability, and the ability to provide real-time risk warnings.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] The first aspect of the present invention provides a method for predicting seismic risk of transmission lines based on multi-source data intelligent driving.

[0008] In one or more embodiments, a multi-source data-driven intelligent method for predicting seismic risk of transmission lines is provided, including:

[0009] A continuously updated hybrid deep learning model is used to process the actual seismic wave time history to predict the top displacement time history of the transmission line, which serves as an indicator of the transmission line's dynamic response.

[0010] The predicted top displacement time history of the transmission line is compared with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, the vulnerability curve method is used to convert the predicted top displacement time history into the probability of occurrence of different structural performance states.

[0011] In the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data.

[0012] As one implementation method, the hybrid deep learning model consists of an LSTM module, a Transformer module, and a feature fusion module. The LSTM module is used to extract the nonlinear evolution features of the power transmission line dynamic response in the local time scale. The Transformer module is used to extract the global dependency features of the training samples in the long time scale. The feature fusion module is used to concatenate the features extracted by the LSTM module and the Transformer module and output the predicted top displacement time history.

[0013] As one implementation method, an adaptive learning rate adjustment mechanism is introduced during the parameter update process of the hybrid deep learning model. When the prediction error of the hybrid deep learning model on new samples is lower than a set threshold, the learning rate is automatically reduced; when the prediction error is higher than the set threshold, the learning rate is increased to speed up the correction.

[0014] As one implementation method, in the process of pre-training the hybrid deep learning model, the hybrid deep learning model is pre-trained by jointly using simulation supervision loss and real supervision loss; wherein, the simulation supervision loss introduces a simulation supervision confidence coefficient of a positive number less than or equal to 1, and uses it as a weight coefficient for the squared difference between the simulated displacement time history and the predicted top displacement time history; the real supervision loss is the squared difference between the real displacement time history and the predicted top displacement time history.

[0015] As one implementation method, during long-term operation, when the hybrid deep learning model performs incremental learning to update parameters, its corresponding incremental update loss consists of the true supervision loss and the distillation constraint loss; wherein, the distillation constraint loss is the squared difference between the output of the hybrid deep learning model before the update and the output of the hybrid deep learning model after the update; the true supervision loss is the squared difference between the true displacement time history and the predicted top displacement time history.

[0016] As one implementation method, a partial parameter freezing mechanism is adopted during the parameter update process of the hybrid deep learning model, updating only the parameters related to site effects and structural nonlinearity.

[0017] As one implementation method, the performance evaluation metrics for the continuously updated hybrid deep learning model include the peak displacement prediction error, response time history similarity, and component strain prediction accuracy.

[0018] A second aspect of the present invention provides a multi-source data-driven intelligent earthquake risk prediction system for power transmission lines.

[0019] In one or more embodiments, a multi-source data-driven intelligent earthquake risk prediction system for transmission lines includes:

[0020] The prediction output module is used to process the actual seismic wave time history using a continuously updated hybrid deep learning model to predict the top displacement time history of the transmission line as an indicator of the transmission line's dynamic response.

[0021] The graded early warning module is used to compare the predicted top displacement time history of the transmission line with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, it uses the vulnerability curve method to convert the predicted top displacement time history into the probability of occurrence of different structural performance states.

[0022] In the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data.

[0023] A third aspect of the present invention provides a computer-readable storage medium.

[0024] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps in the multi-source data intelligent-driven method for predicting seismic risks of transmission lines as described above.

[0025] A fourth aspect of the present invention provides an electronic device.

[0026] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, it implements the steps in the multi-source data intelligent-driven method for predicting seismic risks of transmission lines as described above.

[0027] Compared with the prior art, the beneficial effects of the present invention are:

[0028] (1) The present invention adopts a hybrid deep learning modeling technology scheme based on the fusion of real seismic motion input and numerical simulation supervision label. Its principle is to use real seismic wave time history as model input and numerical simulation structural response as initial supervision, and introduce a simulation credibility weighting mechanism to suppress the influence of simulation simplification error, thereby establishing a nonlinear mapping relationship between seismic motion and transmission line structural response. This solves the problem in the prior art that the model relies on single simulation training and is difficult to generalize to real structural conditions, and improves the reliability and engineering applicability of seismic response prediction.

[0029] (2) This invention adopts a seismic response prediction technology based on hierarchical feature extraction and LSTM-Transformer hybrid network. The principle is to characterize the long-term global effect of ground motion through event-level features, characterize the short-term abrupt change features through time-series features, and model the global dependence and local evolution law through the hybrid network. This technology solves the problem that a single model in the existing technology is difficult to simultaneously characterize the long-term cumulative effect of ground motion and the nonlinear abrupt change response of the structure, and improves the prediction accuracy of the critical response peak and abrupt change process of transmission lines.

[0030] (3) The present invention adopts an incremental update and risk warning technology based on experimental and online monitoring data. Its principle is to continuously introduce real structural response data for controlled updates on the basis of model pre-training, and combine the predicted response with vulnerability assessment to trigger graded warnings. This solves the problems of fixed model parameters, inability to adapt to long-term structural evolution and lack of closed-loop risk management in the existing technology, and realizes the long-term adaptive optimization and proactive risk warning capability of the seismic response prediction model of transmission lines. Attached Figure Description

[0031] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0032] Figure 1 This is a flowchart of the multi-source data intelligent-driven method for predicting seismic risks of transmission lines according to an embodiment of the present invention;

[0033] Figure 2This is a schematic diagram of a hybrid deep learning network structure based on incremental learning according to an embodiment of the present invention;

[0034] Figure 3 This is a flowchart of risk prediction and early warning according to an embodiment of the present invention. Detailed Implementation

[0035] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0036] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0037] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0038] To address the problems in the background technology, and the fact that existing models mostly remain at the "response prediction" level without solving the mapping problem between prediction results and engineering risk and early warning decisions, even if structural displacement, strain, or acceleration responses can be obtained, there is still a lack of systematic technical solutions for how to transform them into interpretable and quantifiable seismic risk indicators for transmission lines and form a closed-loop management mechanism of "prediction-assessment-early warning-feedback update." Simple combinations of deep learning models cannot complete this risk closed loop.

[0039] This invention addresses the specific engineering scenario of seismic risk prediction for power transmission lines. It constructs a prediction and early warning method and system that integrates multi-source data and allows for dynamic updates. This system achieves a closed-loop process from multi-source monitoring data input, joint modeling of spatiotemporal characteristics, simultaneous assessment of seismic response and risk indicators, to early warning decision-making and model self-updating. The technical solution of this invention solves the problems of existing technologies, such as limited model applicability, inability to dynamically update models, and lack of closed-loop risk management, thus achieving substantial technical advantages that distinguish it from existing technologies.

[0040] Combination Figure 1 and Figure 3 The multi-source data-driven intelligent method for predicting seismic risks of transmission lines in this embodiment may include the following steps:

[0041] Step 1: Process the actual seismic wave time history using a continuously updated hybrid deep learning model to predict the top displacement time history of the transmission line, which can be used as an indicator of the dynamic response of the transmission line.

[0042] During the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data.

[0043] In this embodiment, as Figure 2 As shown, the hybrid deep learning model consists of an LSTM (Long Short-Term Memory) module, a Transformer module, and a feature fusion module. The hybrid deep learning model uses numerical simulation data and experimental bench data as joint training samples driven by multiple data sources. The LSTM module is used to extract the nonlinear evolution features of the power transmission line dynamic response in local time scales. The Transformer module is used to extract the global dependency features of the joint training samples in long time scales. The feature fusion module is used to concatenate the features extracted by the LSTM module and the Transformer module and output the predicted top displacement time history.

[0044] Specifically, the multi-source dataset composed of joint training samples includes: actual seismic wave acceleration time histories. The input features include peak ground acceleration (PGA), peak ground velocity (PGV), response spectrum features, and energy distribution. To fully characterize the intensity characteristics, spectral structure, and duration of ground motion, and to match the long-term and short-term division of labor modeling requirements of the LSTM and Transformer hybrid model, this embodiment performs hierarchical feature extraction on the ground motion input, forming two types of input: event-level features and time-series features.

[0045] Let the actual seismic wave acceleration time history be Discretized into Peak ground acceleration and peak ground velocity are defined as follows:

[0046] ;

[0047] in, The seismic velocity time history is obtained from acceleration integration. The response spectrum characteristics are obtained using the standard response spectrum calculation method, and values ​​are taken at representative periods of the transmission line structure to form the response spectrum feature vector. Simultaneously, time-frequency analysis was performed on the seismic input to obtain the energy trajectories of each frequency band. , , Define the number of frequency bands and the energy percentage of each band:

[0048] ;

[0049] This forms a feature vector at the level of a seismic event. , as the global conditional input of the Transformer module, where .

[0050] Meanwhile, in order to characterize the short-term abrupt changes and local peak evolution patterns in the seismic input, this embodiment constructs seismic time-series features and inputs them into the LSTM module.

[0051] The actual seismic wave acceleration is discretized as Then the local peak sequence is defined as:

[0052] ;

[0053] in This represents a local time window centered on the current moment.

[0054] Simultaneously define the local energy sequence:

[0055] ;

[0056] In addition, time-frequency analysis of earthquake motion yields energy trajectories for each frequency band. This is retained as a frequency band energy temporal characteristic. This constitutes a seismic motion temporal level characteristic sequence:

[0057] ;

[0058] Input the LSTM module to learn the local temporal evolution and short-time nonlinear abrupt change characteristics of the seismic input.

[0059] The numerical simulation data consists of the time histories of nodal displacements, accelerations, and internal forces obtained through finite element dynamic analysis. A dynamic model of the transmission tower was established using finite element software (such as OpenSees), and different seismic wave inputs were applied to obtain the displacement, velocity, and acceleration responses. The dynamic equation is as follows:

[0060] ;

[0061] in, For the quality matrix, Here is the damping matrix. Here is the stiffness matrix. These are the displacement, velocity, and acceleration response vectors, respectively. This is the earthquake influence vector. That is, inputting seismic acceleration. .

[0062] The experimental data are the actual response time histories obtained through shaking table and bench tests. The displacement time histories at the top of the tower were obtained through shaking table and bench tests. The strain time history of the components is used for model correction and incremental learning. During the data preprocessing stage, the above data is denoised, normalized, and time-history aligned. The standardization method is as follows: wind-ice load is applied to the established finite element model of the transmission line, multiple discrete failure probability points are calculated, and the failure probability points are selected for fitting to obtain the failure probability curve;

[0063] ;

[0064] in, For the original sample, The mean, The standard deviation is denoted as .

[0065] The nonlinear evolution features of the power transmission line dynamic response extracted by the LSTM module within a local time scale are time-series features, and their input is a time-series feature sequence of ground motion:

[0066] .

[0067] The LSTM module is used to extract short-term dependency features, and its state update formula is:

[0068]

[0069] in, This is a forget gate, used to control the cell state at the previous moment. The degree of preservation; This is an input gate used to control the proportion of current input information written into the cell state; The candidate cell state is used to characterize the updatable state information at the current moment; As a cellular state, it is responsible for transmitting and accumulating key information over time; This is the hidden state, representing the output feature at the current moment. For a moment The input feature vector, This is the hidden state from the previous moment; and These are the weight matrices corresponding to the forget gate, input gate, candidate cell state, and output gate, respectively. and These are the bias terms for the corresponding gate structures. Function This represents the Sigmoid activation function. This represents the hyperbolic tangent function.

[0070] Complete from to After iteration, the short-term dependency feature vector output by the LSTM module is defined as:

[0071] ;

[0072] in This represents the pooling operator.

[0073] The global dependency features of the joint training samples extracted by the Transformer module over long time scales are event-level features, and their input is the seismic event-level feature vector. event-level feature vectors The input representation of the Transformer is obtained through embedding mapping:

[0074]

[0075] in These are the embedding layer weight matrix and the bias term, respectively. Subsequently, in the... In the self-attention layer, according to the first The input representation of a self-attention layer Generate the corresponding query matrix Key matrix Sum matrix :

[0076] ;

[0077] in This generates weight matrices corresponding to the query, key, and value, respectively. The attention output of this layer is then obtained through a self-attention mechanism. :

[0078]

[0079] in For the key vector dimension. Query matrix. This is used to characterize the structural response of transmission lines under the current seismic event or moment being analyzed. It reflects the seismic motion characteristics that the model needs to focus on under the current operating conditions and the direction of their influence on the structural response. Key matrix Value matrix is ​​a feature identifier used to characterize different seismic ground motion events and operating conditions over a long time scale. It describes the physical properties and correlation characteristics of each event feature within the overall response process, thus providing a reference for similarity matching. Used to characterize the structural response information or its featured representation corresponding to the key matrix, it carries effective response information formed by the long-term coupling effect of seismic motion and transmission lines, and is used to output the final global dependency feature after attention weighting. This is achieved by analyzing the query matrix. AND key matrix Perform similarity calculation and pair the value matrix based on the similarity. By performing weighted summation, the Transformer module can highlight the structural response features most relevant to the current seismic event and operating conditions over a long time scale, thereby achieving adaptive modeling of the global dependency of seismic response. This is the transpose of the matrix.

[0080] Furthermore, the layer output is obtained through residual connections and a feedforward network. :

[0081] ;

[0082] in This represents a feedforward network. When it passes through... After the layer, the final output is taken as the long-term global dependency feature vector output by the Transformer module. :

[0083] ;

[0084] The feature fusion module concatenates LSTM and Transformer features:

[0085] ;

[0086] Output the predicted top displacement time history:

[0087] ;

[0088] in, This is the predicted top displacement time history.

[0089] This embodiment establishes a complex nonlinear mapping relationship between seismic motion input, structural response and operating condition information and the overall dynamic response of transmission lines by adaptively fusing the output features of two types of networks based on gating weights, thereby forming a pre-trained model for predicting the seismic response of transmission lines.

[0090] In the process of pre-training the hybrid deep learning model, simulation supervision loss and real supervision loss are used together to pre-train the hybrid deep learning model. Among them, the simulation supervision loss introduces a positive simulation supervision confidence coefficient of less than or equal to 1, and uses it as the weight coefficient of the square difference between the simulated displacement time history and the predicted top displacement time history; the real supervision loss is the square difference between the real displacement time history and the predicted top displacement time history.

[0091] The model is trained on a large-scale numerical simulation dataset. To address potential systematic biases in simulation labels, this embodiment introduces a simulation supervision credibility coefficient. And construct the simulation weighted loss, i.e., the simulation supervised loss. :

[0092] ;

[0093] in, The values ​​are adaptively determined based on the distribution characteristics of the input ground motion intensity and energy within the structure's sensitive frequency band. A larger value is taken when the reliability of the simulation condition is high, and a smaller value is taken when there is strong nonlinearity or simulation error risk, thereby suppressing the adverse effects of simulation simplification errors on model training. For the first During this training process The displacement time history simulated at every moment. For the first During this training process The time history of the top displacement predicted at each moment; T is the total training time.

[0094] When experimental bench tests or online monitoring data are obtained, a realistic monitoring loss is constructed. :

[0095] ;

[0096] in, The actual structural response obtained from experiments or online monitoring, i.e., the actual displacement time history; The predicted structural response is the predicted top displacement time history; this loss is used to guide the model to learn the true structural dynamic characteristics step by step and correct the differences between the simulation domain and the real domain.

[0097] During the pre-training phase, a joint loss method combining simulation supervision and real supervision is employed. :

[0098] ;

[0099] in This is a balancing coefficient used to adjust the weight ratio of real supervision in joint training. It is usually preset according to the specific scenario.

[0100] During long-term operation, the parameters of the hybrid deep learning model are updated using incremental learning based on experimental data from test benches, online monitoring data, and post-earthquake inspection data. This allows the deep learning model to maintain its original seismic response mapping capabilities while gradually absorbing new information reflecting changes in structural state and operating conditions. This approach avoids performance degradation or catastrophic forgetting during continuous learning, enabling the transmission line seismic response prediction model to continuously optimize and adaptively evolve under long-term operating conditions.

[0101] Incremental learning is employed, gradually introducing experimental and monitoring data, using a parameter partial freeze and mini-batch update mechanism. The parameter update formula is:

[0102] ;

[0103] in, For the first Parameters of the rotation generation; For the first Parameters of the rotation generation; For learning rate, For the new dataset; This represents the gradient of the joint loss. During the parameter update process of the hybrid deep learning model, an adaptive learning rate adjustment mechanism is introduced. When the prediction error of the hybrid deep learning model on new samples is lower than a set threshold, the learning rate is automatically reduced; when the prediction error is higher than the set threshold, the learning rate is increased to accelerate the correction speed.

[0104] This embodiment uses seismic motion input as model input and structural response (especially the top displacement time history) as supervision signal. It employs the squared difference as the loss function for training, resulting in a pre-trained model capable of characterizing the complex mapping relationship between seismic motion and structural response. Based on the pre-trained model, experimental data from test benches, online monitoring data, and post-earthquake inspection data are introduced, and a small-batch data update strategy is used for model optimization. In the specific implementation, a partial parameter freezing mechanism is used, updating only parameters related to site effects and structural nonlinearity to prevent catastrophic forgetting. A sliding time window mechanism is used to gradually introduce new data, allowing the model to continuously absorb new information while maintaining historical learning results, achieving adaptation to new environments and seismic conditions. The incremental learning process employs adaptive learning rate adjustment and regularization methods.

[0105] During long-term operation, when the hybrid deep learning model performs incremental learning to update parameters, its corresponding incremental update loss consists of the real supervision loss and the distillation constraint loss; where the distillation constraint loss is the squared difference between the output of the hybrid deep learning model before the update and the output of the hybrid deep learning model after the update.

[0106] During long-term operation, when the model is incrementally updated based on new experimental or monitoring data, this embodiment introduces an anti-forgetting constraint to prevent the model from losing its original seismic response mapping ability due to learning from new data. Let the model before the update be... During this training process The output at time is Then construct distillation constraint loss :

[0107] ;

[0108] Incremental update of loss is adopted. :

[0109] ;

[0110] in To prevent forgetting the constraint weight coefficients, It is usually preset according to the specific scenario.

[0111] This embodiment utilizes an independent validation dataset (including some numerical simulation data and experimental data not used in training) to validate the model. Evaluation metrics include the peak displacement prediction error at the top, response time history similarity, and component strain prediction accuracy. During an earthquake, the actual seismic wave time history is acquired in real time and input into the trained model. The model outputs the dynamic response of the transmission line structure, focusing on the displacement time history at the top of the towers. This time history not only reflects the dynamic response characteristics of the structure under seismic loading but also serves as a core input indicator for subsequent risk assessment and early warning.

[0112] Step 2: Compare the predicted top displacement time history of the transmission line with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, use the vulnerability curve method to convert the predicted top displacement time history into the probability of occurrence of different structural performance states.

[0113] See Figure 3 When an earthquake occurs, the actual seismic wave time history will be... Input the model to obtain the top displacement prediction. The predicted top displacement time history is compared with the structural performance limits (including yield displacement, instability displacement, and critical failure displacement) to determine the structural performance status. Furthermore, the vulnerability curve method is used to convert the predicted top displacement time history into the probability of occurrence of different structural performance states, thereby obtaining the system-level risk value.

[0114] The predicted top displacement time history is converted into the probability of occurrence of different structural performance states using the fragility curve method. The expression is:

[0115] ;

[0116] in, It is an index of seismic intensity (such as PGA). The median intensity The standard deviation is the logarithm. It is a normal distribution function; For transmission tower response limits, It can be divided into and ; For unstable displacement, These are the yield displacements, all of which are known data and can be obtained experimentally.

[0117] The early warning criterion is based on the predicted peak value of the top displacement time history. Comparison with known structural performance limits for transmission lines:

[0118] like A yellow warning will be issued, along with the probability of the structural performance condition occurring; on-site inspections will be strengthened, and no action is required if there is no obvious bending of the members visible to the naked eye.

[0119] like The system will issue an orange alert and provide the probability of the structural performance condition occurring; dispatch personnel to conduct on-site inspections, and stockpile the necessary accessories for transmission tower components and reinforcement measures, so that any damage can be repaired immediately.

[0120] like It outputs a red alert and provides the probability of the occurrence of structural performance status, assesses the cost of repairing and rebuilding the transmission tower, and carries out the reconstruction work of the transmission tower.

[0121] In other embodiments, such as Figure 3 As shown, based on structural seismic resistance requirements, risk levels are categorized (minor, moderate, severe, and collapse). When the predicted result exceeds a certain level threshold, the corresponding early warning signal is automatically triggered. For example, a yellow warning corresponds to minor damage, an orange warning to moderate or severe damage, and a red warning to the risk of structural instability or collapse. Finally, the risk level, affected area, and handling recommendations are pushed to the power grid dispatch and operation center via a communication interface, achieving a closed loop for rapid earthquake risk prediction and early warning.

[0122] In one or more embodiments, a multi-source data intelligent-driven transmission line earthquake risk prediction system is also provided, which can be implemented in software. The multi-source data intelligent-driven transmission line earthquake risk prediction system includes the following software modules: a prediction output module and a hierarchical early warning module.

[0123] The functions of each software module in the multi-source data-driven intelligent earthquake risk prediction system for transmission lines are described below:

[0124] The prediction output module is used to process the actual seismic wave time history using a continuously updated hybrid deep learning model to predict the top displacement time history of the transmission line as an indicator of the transmission line's dynamic response.

[0125] The graded early warning module is used to compare the predicted top displacement time history of the transmission line with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, it uses the vulnerability curve method to convert the predicted top displacement time history into the probability of occurrence of different structural performance states.

[0126] In the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data.

[0127] It should be noted that each module in the multi-source data intelligent-driven transmission line earthquake risk prediction system of this invention corresponds one-to-one with each step in the multi-source data intelligent-driven transmission line earthquake risk prediction method in the above embodiments, and their specific implementation processes are the same, so they will not be repeated here.

[0128] The structure of the electronic device according to embodiments of the present invention will be described in detail below. The electronic device provided in embodiments of the present invention includes: at least one processor, a memory, a user interface, and at least one network interface. The various components in the multi-source data intelligent-driven transmission line earthquake risk prediction system are coupled together through a bus system. It can be understood that the bus system is used to realize the connection and communication between these components. In addition to a data bus, the bus system also includes a power bus, a control bus, and a status signal bus. The user interface may include a display, keyboard, mouse, trackball, click wheel, buttons, a touchpad, or a touch screen, etc.

[0129] It is understood that the memory can be volatile memory or non-volatile memory, or both. The memory in this embodiment of the invention is capable of storing data to support the operation of the terminal. Examples of this data include any computer programs used to operate on the terminal, such as operating systems and applications. The operating system includes various system programs, such as the framework layer, core library layer, driver layer, etc., used to implement various basic services and handle hardware-based tasks. Applications can include various applications.

[0130] In some embodiments, the multi-source data intelligent-driven transmission line seismic risk prediction system provided in this invention can be implemented using a combination of hardware and software. As an example, the multi-source data intelligent-driven transmission line seismic risk prediction system provided in this invention can be a processor in the form of a hardware decoding processor, programmed to execute the multi-source data intelligent-driven transmission line seismic risk prediction method provided in this invention. For example, the hardware decoding processor can employ one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components.

[0131] As an example, a processor can be an integrated circuit chip with signal processing capabilities, such as a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., where a general-purpose processor can be a microprocessor or any conventional processor, etc.

[0132] As an example of the hardware implementation of the multi-source data intelligent-driven transmission line seismic risk prediction system provided in this embodiment of the invention, the device provided in this embodiment of the invention can be directly executed by a processor in the form of a hardware decoding processor. For example, it can be executed by one or more application-specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field-programmable gate arrays (FPGAs), or other electronic components to implement the multi-source data intelligent-driven transmission line seismic risk prediction method provided in this embodiment of the invention.

[0133] The memory in this embodiment of the invention is used to store various types of data to support the operation of a multi-source data-driven intelligent transmission line seismic risk prediction system, or to store data for execution. Figure 1The program code for the method shown. Examples of this data include: any executable instructions for operation on a multi-source data intelligent-driven transmission line seismic risk prediction system, such as executable instructions that can be included in the executable instructions to implement the multi-source data intelligent-driven transmission line seismic risk prediction method of the embodiments of the present invention.

[0134] Specifically, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising a computer program carried on a computer-readable medium, the computer program including functions for executing... Figure 1 The program code for the method shown. In such an embodiment, the computer program can be downloaded and installed from a network via a communication component, and / or installed from a removable medium. When the computer program is executed by the central processing unit, it performs the various functions defined in the apparatus of this application.

[0135] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0136] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-source data-driven intelligent method for predicting seismic risk of transmission lines, characterized in that, include: A continuously updated hybrid deep learning model is used to process the actual seismic wave time history to predict the top displacement time history of the transmission line, which serves as an indicator of the transmission line's dynamic response. The predicted top displacement time history of the transmission line is compared with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, the vulnerability curve method is used to convert the predicted top displacement time history into the probability of occurrence of different structural performance states. In the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data. The hybrid deep learning model consists of an LSTM module, a Transformer module, and a feature fusion module. The LSTM module is used to extract the nonlinear evolution features of the power transmission line dynamic response in local time scales. The Transformer module is used to extract the global dependency features of the training samples in long time scales. The feature fusion module is used to concatenate the features extracted by the LSTM module and the Transformer module and output the predicted top displacement time history. The global dependency features of the joint training samples extracted by the Transformer module over long time scales are event-level features, and their input is the seismic event-level feature vector. ;in, ; Let the actual seismic wave acceleration time history be Discretized into ; Peak ground acceleration and peak ground velocity are defined as follows: ； in, The seismic velocity time history is obtained by acceleration integration; the response spectrum characteristics are obtained by standard response spectrum calculation methods, and values ​​are taken at representative periods of the transmission line structure to form the response spectrum feature vector. ； Simultaneously, time-frequency analysis was performed on the seismic input to obtain the energy trajectories of each frequency band. , , Define the number of frequency bands and the energy percentage of each band: ； The nonlinear evolution features of the power transmission line dynamic response extracted by the LSTM module within a local time scale are time-series features, and their input is a time-series feature sequence of ground motion: ； The actual seismic wave acceleration is discretized as Then the local peak sequence is defined as: ； in This represents a local time window centered on the current moment. Simultaneously define the local energy sequence: ； In addition, time-frequency analysis of earthquake motion yields energy trajectories for each frequency band. , , The number of frequency bands is used, and these are retained as frequency band energy timing characteristics.

2. The multi-source data intelligent-driven method for predicting seismic risk of transmission lines as described in claim 1, characterized in that, During the parameter update process of the hybrid deep learning model, an adaptive learning rate adjustment mechanism is introduced. When the prediction error of the hybrid deep learning model on new samples is lower than a set threshold, the learning rate is automatically reduced; when the prediction error is higher than the set threshold, the learning rate is increased to speed up the correction.

3. The multi-source data intelligent-driven method for predicting seismic risk of transmission lines as described in claim 1, characterized in that, In the process of pre-training the hybrid deep learning model, simulation supervision loss and real supervision loss are used together to pre-train the hybrid deep learning model. Among them, the simulation supervision loss introduces a positive simulation supervision confidence coefficient of less than or equal to 1, and uses it as the weight coefficient of the square difference between the simulated displacement time history and the predicted top displacement time history; the real supervision loss is the square difference between the real displacement time history and the predicted top displacement time history.

4. The multi-source data intelligent-driven method for predicting seismic risk of transmission lines as described in claim 1, characterized in that, During long-term operation, when the hybrid deep learning model performs incremental learning to update parameters, its corresponding incremental update loss consists of the true supervision loss and the distillation constraint loss; where the distillation constraint loss is the squared difference between the output of the hybrid deep learning model before the update and the output of the hybrid deep learning model after the update; the true supervision loss is the squared difference between the true displacement time history and the predicted top displacement time history.

5. The multi-source data intelligent-driven method for predicting seismic risk of transmission lines as described in claim 1, characterized in that, During the parameter update process of the hybrid deep learning model, a partial parameter freezing mechanism is adopted, and only parameters related to site effects and structural nonlinearity are updated.

6. The multi-source data intelligent-driven method for predicting seismic risk of transmission lines as described in claim 1, characterized in that, The performance evaluation metrics for continuously updated hybrid deep learning models include peak displacement prediction error, response time history similarity, and component strain prediction accuracy.

7. A multi-source data-driven intelligent earthquake risk prediction system for transmission lines, characterized in that, The multi-source data-driven method for predicting seismic risk of transmission lines, based on any one of claims 1-6, includes: The prediction output module is used to process the actual seismic wave time history using a continuously updated hybrid deep learning model to predict the top displacement time history of the transmission line as an indicator of the transmission line's dynamic response. The graded early warning module is used to compare the predicted top displacement time history of the transmission line with the known structural performance limits of the transmission line to determine the structural performance status of the transmission line and its corresponding risk level. At the same time, it uses the vulnerability curve method to convert the predicted top displacement time history into the probability of occurrence of different structural performance states. In the pre-training process, the hybrid deep learning model uses numerical simulation data and test bench data as joint training samples driven by multiple data sources. During long-term operation, the parameters of the hybrid deep learning model are updated based on incremental learning methods using test bench data, online monitoring data, and post-earthquake inspection data.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps in the multi-source data intelligent-driven method for predicting seismic risks of transmission lines as described in any one of claims 1-6.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps in the multi-source data intelligent-driven method for predicting seismic risks of transmission lines as described in any one of claims 1-6.

Images