Method for judging meteorological conditions influencing cableway operation safety
By combining multi-source meteorological data acquisition with a deep time-series model, the problems of high false alarm rate and delayed response in cableway meteorological monitoring have been solved, enabling real-time risk assessment and safety decision-making for cableway operation.
Patent Information
- Application Number
- CN202510988896.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-07-17
- Publication Date
- 2025-10-17
AI Technical Summary
Existing cableway meteorological monitoring technology relies on single-factor static threshold judgment, with a high false alarm rate and a delayed response to small-scale sudden meteorological events in mountain scenic areas, resulting in low operational efficiency.
By employing multi-source meteorological data acquisition, deep time-series models, and dynamic threshold correction methods, combined with the Kafka platform and edge gateway, we can achieve multi-source data cleaning, normalization processing, and real-time risk decision-making, thereby reducing false alarm rates and improving response speed.
It enables timely risk assessment of cableway operations, reduces false alarm and missed alarm rates, and improves operational efficiency and safety.
Smart Images
Figure CN120804949A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of meteorological monitoring, and particularly relates to a design of a method for judging meteorological conditions affecting the operation safety of a cableway. BACKGROUND
[0002] The cableway undertakes the dual functions of passenger transport and material circulation in mountainous scenic areas. However, due to the steep terrain, turbulent airflow and frequent local convection, disaster weather such as strong gusts, thunderstorms and low-temperature icing directly threatens the cableway cable tension, electrical control system and support structure. At present, the industry generally adopts a single wind speed threshold or manual experience shutdown standard. This method lacks quantitative description of the coupling effect of complex meteorological elements, is easy to miss the report of mountain valley wind shear or lightning precursor, and is also easy to cause high-frequency false alarms due to generalized threshold, thereby affecting the operation efficiency. In addition, the existing cableway meteorological monitoring technology is mostly an independent system, and data cannot be fused in real time to a unified judgment process, resulting in information island and response lag. SUMMARY
[0003] The purpose of the application is to solve the problems of the existing cableway meteorological monitoring technology, such as relying on single-element static threshold judgment, high false alarm rate and lag response to small-scale sudden weather in mountainous scenic areas, and a method for judging meteorological conditions affecting the operation safety of a cableway is proposed.
[0004] The technical scheme of the application is as follows: a method for judging meteorological conditions affecting the operation safety of a cableway, comprising the following steps: S1, collecting multi-source original meteorological data along the cableway and transmitting to the Kafka platform.
[0005] S2, performing data cleaning on the multi-source original meteorological data in the Kafka platform to obtain cleaned data.
[0006] S3, obtaining key meteorological features reflecting the risk of the cableway according to the cleaned data, and performing normalization processing on the key meteorological features to obtain normalized features.
[0007] S4, constructing an input tensor according to the normalized features and inputting the deep time series model to output a three-class probability vector.
[0008] S5, judging the meteorological conditions along the cableway according to the three-class probability vector by using a dynamic threshold correction method, and making a risk decision on the operation of the cableway.
[0009] S6, feeding back the risk decision result to the cableway operation and maintenance personnel.
[0010] S7, iteratively optimizing the deep time series model.
[0011] Further, step S1 comprises the following sub-steps: S11, set observation points every 500 m along the cableway line and on the top of all towers, install ultrasonic three-dimensional anemometers, lightning radiometers, fiber grating vibration sensors and 10 m tower meteorological stations at each observation point, and collect multi-source raw meteorological data along the cableway line.
[0012] S12, send the multi-source raw meteorological data to the edge gateway through LoRa-WAN networking.
[0013] S13, upload the multi-source raw meteorological data to the cloud Kafka platform after encryption by the edge gateway using the MQTT protocol.
[0014] Further, step S2 includes the following sub-steps: S21, in the Kafka platform, obtain multi-source raw meteorological data from the sliding window every 30 s, and use the Tukey method to remove outliers in the multi-source raw meteorological data.
[0015] S22, based on the ARIMA(1,0,1) model, time series interpolation is performed on the missing values in the multi-source raw meteorological data ≤90 s: wherein represents the observation value at the current time, represents the observation value at the previous time, represents the random disturbance term at the current time, represents the random disturbance term at the previous time, represents the autoregressive coefficient, represents the moving average coefficient.
[0016] S23, referring to the ground wind speed data from the WRF 5km wind field, correcting the elevation difference between observation points by Cressman interpolation, and the interpolation weight is calculated by the formula: wherein represents the interpolation action radius, represents the distance between the observation point and the target grid.
[0017] Further, the key meteorological features reflecting the risk of the cableway in step S3 include maximum gust intensity, wind direction rotation rate, Froude number, Brent-Weissell frequency, SWEAT index, frozen potential integral, electric field jump rate and cable vibration spectrum energy density.
[0018] The maximum gust intensity represents the rapid change of wind speed within one minute.
[0019] The wind direction rotation rate is used to evaluate whether the wind direction changes rapidly.
[0020] Froude number and Brunt-Vaisala frequency are used to identify the risk of strong wind crossing a valley or air flow convection instability.
[0021] SWEAT index is used to determine the likelihood of thunderstorm occurrence.
[0022] Freezing potential integral is used to measure the risk of icing when temperature is below 0℃.
[0023] Electric field jump rate is a precursor signal before lightning occurs.
[0024] Cable vibration spectrum energy density is used to assess the risk of cable sway induced by resonance or strong wind.
[0025] Further, the formula for normalizing the key meteorological features in step S3 is: wherein represents the original key meteorological feature value, represents the mean value within the sliding window, represents the standard deviation within the sliding window, represents the normalized feature value.
[0026] Further, the deep time series model in step S4 includes two layers of stacked LSTM, each LSTM contains 64 units, and is followed by an attention mechanism module on the time axis, and the model output is a three-class probability vector , wherein represents the probability of safe operation of the cableway, represents the probability of recommended speed limit operation of the cableway, represents the probability of recommended shutdown of the cableway.
[0027] Further, the optimizer used in the training process of the deep time series model in step S4 is AdamW, with the learning rate initial value set to , epoch = 50, batch_size = 32, and the loss function is: wherein represents the target class prediction probability, represents the weight used to reduce the majority class samples.
[0028] Further, step S5 includes the following sub-steps: S51, calculate the root mean square error of WRF forecast field and measured wind speed within the last 30 minutes .
[0029] S52, construct a dynamic risk threshold according to : wherein represents a shutdown risk probability threshold value, represents a speed limit risk probability threshold value, and are both threshold values determined based on the ROC curve inflection point method on the training set, and are both empirical tuning weights.
[0030] S53, determining whether is satisfied, if yes, the risk decision result is cable shutdown, otherwise, step S54 is entered.
[0031] S54, determining whether is satisfied, if yes, the risk decision result is cable speed limit, otherwise, the risk decision result is cable safe operation.
[0032] Further, step S6 is specifically: the risk decision result is written into the cable PLC console through the RESTful API, and is pushed to the cable dispatching APP, the SMS gateway and the LED station screen, so as to feed back to the cable operation and maintenance personnel.
[0033] Further, step S7 is specifically: the past 24h new samples are called through the differential learning module on the cloud Kafka platform at 02:00-03:00 every day, the deep time sequence model is distilled and fine-tuned under the condition of keeping the core structure unchanged, the LSTM weight is automatically updated and the shutdown risk probability threshold value is recalibrated , realizing the self-adaptive iterative optimization of the model.
[0034] The beneficial effects of the present application are: (1) The present application makes full use of the existing wind speed, temperature and humidity, lightning, electric field and cable vibration along the cable line. Multisensor type, realize 1min data synchronization and space interpolation, more comprehensive and timely than traditional single point wind speed monitoring, can quickly capture the mountain valley wind shear and lightning precursor.
[0035] (2) The present application introduces LSTM-Attention deep time sequence model and uses FocalLoss loss function to solve the problem of extreme weather sample scarcity.
[0036] (3) The application realizes real-time coupling of numerical prediction error through dynamic threshold, online discrimination through Kafka platform, edge gateway and MQTT encryption and retransmission, reduces false discrimination caused by prediction deviation, reduces cloud load and ensures data link security, reduces overall outage false alarm rate and missed alarm rate of the cableway. BRIEF DESCRIPTION OF DRAWINGS
[0037] Figure 1 Fig. 1 is a flow chart of a method for discriminating weather conditions affecting the safety of a cableway operation according to an embodiment of the application. DETAILED DESCRIPTION
[0038] Exemplary embodiments of the application will now be described in detail with reference to the accompanying drawings. It should be understood that the embodiments shown and described herein are merely exemplary and are intended to illustrate the principles and features of the application, but are not intended to limit the scope of the application.
[0039] The embodiment of the application provides a method for discriminating weather conditions affecting the safety of a cableway operation, as shown in Figure 1 The method comprises the following steps S1-S7: S1, collecting multi-source original weather data along the cableway and transmitting the multi-source original weather data to a Kafka platform.
[0040] Step S1 comprises the following sub-steps S11-S13: S11, setting observation points every 500m along the cableway and on the top of all towers, installing ultrasonic three-dimensional anemometers, lightning radiometers, fiber grating vibration sensors and 10m tower weather stations at each observation point, collecting multi-source original weather data along the cableway, and collecting data at a frequency of 0.5ms-1min.
[0041] S12, sending the multi-source original weather data to an edge gateway through LoRa-WAN networking.
[0042] S13, uploading the multi-source original weather data to a cloud Kafka platform after encryption by the edge gateway using the MQTT protocol.
[0043] S2, performing data cleaning on the multi-source original weather data in the Kafka platform to obtain cleaned data.
[0044] Step S2 comprises the following sub-steps S21-S23: S21, in the Kafka platform, obtaining multi-source original weather data from a sliding window every 30s, and using the Tukey method to remove outliers in the multi-source original weather data, i.e., setting values exceeding Q3+1.5IQR as NaN, wherein Q3 (third quartile) represents the value corresponding to the position of 75% of the data, IQR (interquartile range) represents the span of 50% of the data (i.e., the range of the middle half of the data), and NaN represents missing.
[0045] S22, based on the ARIMA (1, 0, 1) model, the missing values of the multi-source original meteorological data ≤90s are time series interpolation: Wherein represents the observation value at the current time, represents the observation value at the previous time, represents the random disturbance term at the current time, represents the random disturbance term at the previous time, represents the autoregressive coefficient (representing the influence of historical data), represents the moving average coefficient (representing the influence of historical error).
[0046] S23, referring to the ground wind speed data from WRF 5km wind field, the elevation difference between observation points is corrected by Cressman interpolation, and the interpolation weight The calculation formula is: Wherein represents the interpolation action radius, represents the distance between the observation point and the target grid.
[0047] S3, according to the cleaning data, the key meteorological characteristics reflecting the risk of cableway are obtained, and the key meteorological characteristics are normalized to obtain normalized features.
[0048] In the embodiment of the application, the cleaning data is used to calculate eight key meteorological characteristics that can reflect the risk of cableway, namely maximum gust intensity, wind direction rotation rate, Froude number, Brent-Weissell frequency, SWEAT index, freezing potential integral, electric field jump rate and cable vibration spectrum energy density.
[0049] Wherein, the maximum gust intensity represents the violent change of wind speed in one minute.
[0050] The wind direction rotation rate is used to evaluate whether the wind direction changes rapidly.
[0051] Froude number and Brent-Weissell frequency are used to identify the risk of strong wind passing through a valley or airflow convection instability.
[0052] The SWEAT index judges the possibility of thunderstorm occurrence.
[0053] The freezing potential integral is used to measure the risk of icing when the temperature is lower than 0℃.
[0054] The electric field jump rate is a precursor signal before lightning occurs.
[0055] Cable vibration spectrum energy density is used to assess the risk of cable sway caused by resonance or strong wind.
[0056] In order to facilitate subsequent deep time sequence model learning and processing, the key meteorological features are uniformly normalized to transform features of different units into the same scale range, so that the subsequent deep time sequence model training is more stable and reliable. The normalization processing formula is: wherein represents the original key meteorological feature value, represents the mean value in the sliding window, represents the standard deviation in the sliding window, represents the normalized feature value.
[0057] S4, constructing an input tensor according to the normalized features and inputting the deep time sequence model to output a three-class probability vector.
[0058] In the embodiment of the application, an input tensor with a dimension of 120x8 is constructed according to the normalized features, wherein 120 represents the length of the time sequence, i.e. the meteorological features of the past 120 consecutive time steps are selected, and 8 represents 8 key meteorological features. The deep time sequence model includes two layers of stacked LSTM, each layer of LSTM contains 64 units (unit), each unit can be understood as a memory block, responsible for receiving a time sequence information, maintaining its memory state, and responding to the current input. The deep time sequence model is followed by an attention mechanism module (Additive Attention is used in the embodiment of the application) in the time axis (the direction of data evolution over time). The model output is a three-class probability vector , wherein represents the probability of safe operation of the cableway, represents the probability of recommended speed limit operation of the cableway, represents the probability of recommended shutdown of the cableway.
[0059] In the embodiment of the application, the optimizer used in the training process of the deep time sequence model is AdamW, the learning rate initial value is set to , epoch=50, batch_size=32, and the loss function is: wherein represents the target class prediction probability, represents the weight used to reduce the majority class sample.
[0060] The SHAP values of the first five important features and the importance change trend are recorded automatically during model training, and are used for model interpretation. With the help of the SHAP output feature contribution degree, the early warning result is traceable and easy to explain, and the defects of the experience threshold method, such as black box and high false alarm rate, are overcome.
[0061] S5, according to the three-class probability vector, a dynamic threshold correction method is used to distinguish the weather conditions along the cableway, and a risk decision is made for the operation of the cableway.
[0062] Step S5 includes the following sub-steps S51-S54: S51, calculate the root mean square error of the WRF prediction field and the measured wind speed in the last 30 minutes .
[0063] S52, according to A dynamic risk threshold is constructed: Among them represents the shutdown risk probability threshold, represents the speed limit risk probability threshold, and are the thresholds determined based on the ROC curve inflection point method on the training set, and are empirical tuning weights, used to relax or tighten the threshold in real time according to the prediction error.
[0064] S53, judge whether ≥ , if yes, the risk decision result is cableway shutdown, otherwise go to step S54.
[0065] S54, judge whether ≥ , if yes, the risk decision result is cableway speed limit, otherwise the risk decision result is cableway safe operation.
[0066] S6, the risk decision result is fed back to the cableway operation and maintenance personnel.
[0067] In the embodiment of the application, the risk decision result is written into the cableway PLC console through the RESTful API, and is pushed to the cableway dispatching APP, the SMS gateway and the LED station screen, so as to be fed back to the cableway operation and maintenance personnel. The operation and maintenance personnel can confirm or correct the disposal result in the APP, and the system records and returns to the cloud for subsequent model evaluation.
[0068] S7, iterative optimization is performed on the deep time series model.
[0069] In the embodiment of the application, the cloud Kafka platform calls the newly added samples of the past 24 hours through a differential learning module at 02:00-03:00 every day, performs model distillation and fine-tuning on the deep time sequence model under the condition of keeping the core structure unchanged, automatically updates the LSTM weight and recalibrates the outage risk probability threshold , and realizes self-adaptive iterative optimization of the model.
[0070] Those skilled in the art will appreciate that the embodiments described herein are presented for the purpose of helping the reader understand the principles of the application and should be understood as not limiting the scope of protection of the application to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations according to the technical inspiration disclosed in the application without departing from the essence of the application, and these modifications and combinations are still within the scope of protection of the application.
Claims
1. A method for determining meteorological conditions that affect cableway operation safety, characterized in that: The following steps are involved: S1, collects multi-source raw meteorological data along the cableway and transmits it to the Kafka platform; S2. Clean the multi-source raw meteorological data in the Kafka platform to obtain cleaned data. S3. Obtain key meteorological features reflecting cableway risks based on the cleaned data, and normalize the key meteorological features to obtain normalized features; S4. Construct an input tensor based on the normalized features and input it into the deep time series model, outputting a three-class probability vector. S5. Based on the three-class probability vector, a dynamic threshold correction method is used to identify the meteorological conditions along the cableway and make risk decisions for the cableway operation. S6. Feedback the risk decision results to the cableway operation and maintenance personnel; S7. Iteratively optimize the deep timing model.
2. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The step S1 includes the following sub-steps: S11. Set up observation points every 500 m along the cableway and on top of all towers. Install an ultrasonic three-dimensional anemometer, a lightning radiometer, a fiber Bragg grating vibration sensor, and a 10-m tower meteorological station at each observation point to collect multi-source raw meteorological data along the cableway. S12, sending multi-source raw meteorological data to the edge gateway via LoRa-WAN networking; S13. Use the MQTT protocol at the edge gateway to encrypt the multi-source raw meteorological data and upload it to the cloud Kafka platform.
3. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The step S2 comprises the following sub-steps: S21. In the Kafka platform, obtain multi-source raw meteorological data from the sliding window every 30 seconds and use the Tukey method to remove outliers in the multi-source raw meteorological data; S22. Based on the ARIMA (1,0,1) model, time series interpolation is performed on missing values ≤90s in multi-source original meteorological data: in represents the observation value at the current moment, represents the observation value at the previous moment, represents the random disturbance term at the current moment, represents the random disturbance term at the previous moment, represents the autoregressive coefficient, represents the sliding mean coefficient; S23, referring to the surface wind speed data from the WRF 5km wind field, using Cressman interpolation to correct the elevation difference between observation points, interpolation weight The calculation formula is: in represents the interpolation radius, Indicates the distance between the observation point and the target grid point.
4. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The key meteorological characteristics reflecting the cableway risk in step S3 include maximum gust intensity, wind direction rotation rate, Froude number, Brunt-Weisselay frequency, SWEAT index, freezing potential integral, electric field jump rate and cable vibration spectrum energy density; The maximum gust intensity indicates a dramatic change in wind speed within one minute; The wind direction rotation rate is used to evaluate whether the wind direction changes rapidly; The Froude number and Brunt-Weisselay frequency are used to identify risks such as strong winds crossing valleys or convective instabilities in airflow; The SWEAT index determines the likelihood of a thunderstorm; The freezing potential score is used to measure the risk of freezing when the temperature is below 0°C; The electric field transition rate is a precursor signal before lightning occurs; The cable vibration spectrum energy density is used to assess the risk of cable shaking caused by resonance or strong wind.
5. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The formula for normalizing the key meteorological features in step S3 is: in represents the original key meteorological characteristic value, represents the mean value within the sliding window, represents the standard deviation within the sliding window, represents the normalized eigenvalue.
6. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The deep time series model in step S4 includes two stacked LSTM layers, each layer of LSTM contains 64 units, and is connected to the time axis with an attention mechanism module. The model output is a three-class probability vector ,in represents the probability of safe operation of the cableway, represents the probability of the cableway recommending speed limit operation, Represents the probability that the cableway is recommended to be shut down.
7. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: In the training process of the deep time series model in step S4, the optimizer uses AdamW and sets the initial value of the learning rate to , epoch=50, batch_size=32, loss function for: in represents the target category prediction probability, Represents the weight used to reduce the majority class samples.
8. The method for determining meteorological conditions affecting cableway operation safety according to claim 6, characterized in that: The step S5 comprises the following sub-steps: S51. Calculate the root mean square error between the WRF forecast field and the measured wind speed in the last 30 minutes. ; S52, according to Constructing dynamic risk thresholds: in represents the outage risk probability threshold, represents the speed limit risk probability threshold, and Both are thresholds determined based on the ROC curve inflection point method on the training set. and All weights are adjusted by experience; S53, determine whether it is satisfied ≥ If yes, the risk decision result is that the cableway is shut down, otherwise, go to step S54; S54, determine whether it is satisfied ≥ If yes, the risk decision result is the speed limit of the cable car, otherwise the risk decision result is the safe operation of the cable car.
9. The method for determining meteorological conditions affecting cableway operation safety according to claim 1, characterized in that: The step S6 specifically includes: writing the risk decision result into the cableway PLC console via the RESTful API, and simultaneously pushing it to the cableway dispatching APP, SMS gateway and LED station screen, so as to provide feedback to the cableway operation and maintenance personnel.
10. The method for determining meteorological conditions affecting cableway operation safety according to claim 8, characterized in that: The step S7 is specifically as follows: the differential learning module is used to call the new samples of the past 24 hours from 02:00 to 03:00 every day on the cloud Kafka platform, and the deep time series model is distilled and fine-tuned while keeping the core structure unchanged, and the LSTM weight is automatically updated and the outage risk probability threshold is recalibrated. , realizing model adaptive iterative optimization.