Early warning method and system for icing galloping of power transmission line based on data joint characteristics

Abstract

The invention discloses a power transmission line icing galloping early warning method and a power transmission line icing galloping early warning system based on data joint characteristics, and relates to the technical field of power transmission line risk prediction. Extracting various characteristics such as wind direction frequency, wind speed and wind direction joint probability, humidity fluctuation amplitude, air pressure and wind speed correlation and tension fluctuation amplitude as environment states, inputting the environment states into a DQN algorithm model, outputting the algorithm according to action definition and training through a reward mechanism, and finally feeding back a deicing effect and a line galloping condition by operation and maintenance personnel. Safe and stable operation of a power transmission line is effectively guaranteed, early warning accuracy and effectiveness are improved, operation and maintenance cost is reduced, and overall safety of a power system is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of transmission line risk prediction, and in particular to a data-joint feature transmission line ice dancing warning method and system. Background Art

[0002] Transmission lines, as key infrastructure for power transmission, are widely distributed in various complex geographical environments. In cold weather conditions, ice dancing becomes an important factor threatening the safe and stable operation of transmission lines.

[0003] Traditional transmission line ice dancing monitoring and early warning technologies are mainly concentrated in several aspects. First, some early technologies simply rely on basic meteorological data provided by meteorological stations, such as single parameters such as wind speed and temperature to judge the possibility of ice dancing. However, this method ignores the complex interactions between meteorological factors and the combined influence of environmental factors. For example, warnings based only on whether the wind speed exceeds a certain threshold may lead to misjudgments or missed judgments because the synergy of factors such as wind direction, humidity, and air pressure is not considered. When the humidity is high and the air pressure changes are special, even if the wind speed does not reach the traditional warning value, the line may dance due to ice formation and special aerodynamics.

[0004] Second, some technologies focus on monitoring line tension. By installing tension sensors on transmission lines and monitoring changes in line tension, the icing situation and the possibility of dancing can be inferred. However, changes in line tension may be affected by a variety of factors, such as thermal expansion and contraction of the line itself, elastic changes caused by line aging, etc. It is difficult to accurately distinguish between normal line status changes and abnormal changes caused by icing and dancing based solely on tension data. The lack of comprehensive consideration of other related factors greatly reduces the accuracy of the early warning.

[0005] Third, some existing early warning methods are relatively simple in terms of algorithmic models. For example, judgment models with fixed thresholds or simple linear regression models are used. These models are difficult to handle nonlinear relationships between multiple variables and complex dynamic changes in the environment. When faced with diverse meteorological and line conditions in different regions and seasons, they have poor adaptability and cannot accurately adjust early warning strategies according to actual conditions. They are prone to untimely or excessive warnings, which brings troubles to power operation and maintenance work and increases operation and maintenance costs.

[0006] With the continuous growth of electricity demand and the continuous expansion of the power grid, the requirements for the safety and reliability of transmission lines are increasing. The limitations of traditional technologies in dealing with ice-covered and dancing warnings of transmission lines are becoming increasingly prominent. There is an urgent need for an early warning method and system that can integrate multi-source data, accurately extract features, make intelligent decisions, and have self-optimization capabilities to effectively protect transmission lines from the harm of ice-covered and dancing in complex environments, improve the stability and safety of power supply, and reduce the economic losses and social impacts caused by line failures. Summary of the invention

[0007] The purpose of the present invention is to provide a data-joint feature-based transmission line ice dancing warning method and system. Through the warning system, meteorological and line tension data are first collected and multiple types of features are extracted as environmental state inputs to the DQN algorithm. The algorithm is output according to the action definition and trained through the reward mechanism. The action definition is then optimized through operation and maintenance feedback, and the cycle is improved to ensure the safety of the transmission line.

[0008] To achieve the above object, the present invention provides the following technical solutions:

[0009] A method for early warning of ice dancing on a power transmission line based on data joint features, characterized in that it comprises the following steps:

[0010] S1: Data collection;

[0011] S2: Feature extraction;

[0012] S3: Action definition;

[0013] S4: Algorithm training;

[0014] S5: Establish a feedback mechanism.

[0015] The meteorological data includes wind speed, wind direction, temperature, humidity, and air pressure information along the transmission line obtained from the local meteorological department, and recorded at certain time intervals to form time series data; the line data collects the line tension through the tension sensor installed on the transmission line.

[0016] Feature extraction includes the frequency of wind direction, the joint probability of wind speed and wind direction interval, humidity fluctuation amplitude, the correlation between air pressure and wind speed, and tension fluctuation amplitude;

[0017] The specific calculation process of the frequency of occurrence of the wind direction is as follows:

[0018] The wind direction data is divided into certain angle intervals, and the wind direction angle 0°-360° is divided into 16 intervals, each interval is 22.5°; the frequency of wind direction in each interval is counted f i , and its calculation formula is:

[0019]

[0020] where n i is the number of data points whose wind direction falls within the ith interval, and n is the total number of data points. Through wind direction distribution statistics, we can intuitively understand the main wind direction and its probability of occurrence, providing a basis for subsequent analysis of the relationship between wind direction and ice dancing.

[0021] The specific calculation process of the joint probability of the wind speed and wind direction interval is as follows:

[0022] According to the definition of probability density function, for any wind speed interval [a, b] and wind direction interval [c, d], the following formula is implemented:

[0023]

[0024] Where v represents wind speed, θ represents wind direction, a and b are the lower and upper limits of the wind speed interval, c and d are the lower and upper limits of the wind direction interval, which are used to define the range of wind directions for which the probability is to be calculated. f(v,θ) is the joint probability distribution function of wind speed v and direction θ. By performing double integration on the corresponding interval, Get the joint probability of the specified wind speed and direction interval;

[0025] The kernel density estimation method is used to estimate the specific form of the joint probability distribution function f(v,θ), and its calculation formula is:

[0026]

[0027] Where n is the number of data points, K h (x,y) is the kernel function, v i and θ i are the wind speed and direction values ​​of the ith data point.

[0028] The specific calculation process of the humidity fluctuation amplitude is as follows:

[0029] Define the humidity fluctuation amplitude σ H , the formula is:

[0030]

[0031] Among them, σ H is the humidity function, which indicates the change of humidity over time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean value of humidity. Humidity fluctuations, especially the duration and amplitude of high humidity, have an important impact on the formation of ice. In a high humidity environment, water vapor is more likely to condense into ice on transmission lines. When the humidity fluctuates greatly, it may cause changes in the growth rate and morphology of ice, which in turn affects the possibility of ice dancing.

[0032] Considering the correlation between air pressure and wind speed, their covariance calculation formula is:

[0033]

[0034] Among them, P(t) is the air pressure function, which indicates the change of air pressure with time t, v(t) is the wind speed function, which indicates the change of wind speed with time t, t 1 and t 2 are the start time and end time of the time period considered, which are used to define the time range for calculating the covariance. is the mean air pressure, is the mean of wind speed, Cov(P,v) is the covariance of air pressure and wind speed, which reflects the coordinated changes of the two variables of air pressure and wind speed in the same time period.

[0035] Defining the tension fluctuation amplitude The formula is:

[0036]

[0037] Among them, F tENSION (t) is the line tension function, which indicates the change of line tension with time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean value of the tension. Changes in line tension may be caused by factors such as increased ice weight and line dancing. By analyzing the fluctuation amplitude of line tension and other characteristics, we can understand the stress of the line and then determine the possibility of ice dancing.

[0038] The early warning measures taken are taken as actions, issuing early warning signals of different levels, arranging operation and maintenance personnel to conduct on-site inspections, and initiating emergency protection measures. The actions use discrete values ​​to represent different actions, which are specifically:

[0039]

[0040] The specific algorithm training is as follows: the previous calculated values ​​of the transmission line, including the joint probability distribution of wind speed and direction, humidity fluctuation amplitude, the correlation between air pressure and wind speed, and the line temperature and tension parameter characteristics are imported into the DQN algorithm model as environmental states, and the defined actions are used as outputs; by continuously inputting environmental states and corresponding actions, training is performed according to the set reward mechanism. Specifically, when the environmental state at a certain moment is input, the algorithm selects an action according to the current strategy. If the subsequent actual situation shows that this action is correct, the algorithm is given a corresponding reward, otherwise it is punished; through such a training process, the DQN algorithm can optimally select appropriate actions according to different environmental states to achieve the best early warning effect.

[0041] The feedback mechanism is established specifically for the operation and maintenance personnel to feed back to the early warning system the de-icing effect and whether the line is still dancing after taking countermeasures. If it is found that the countermeasures taken by the operation and maintenance personnel are not effective and the line is still frequently dancing after de-icing, the action definition of the early warning system can be optimized and more effective countermeasure options can be added to enable the algorithm to better select appropriate actions. By continuously optimizing based on feedback information, the entire early warning process and methods can be continuously improved to improve the accuracy and effectiveness of the ice-covered transmission line dancing warning.

[0042] A transmission line ice dancing warning system based on data joint features, characterized in that it includes a server and a processor, wherein a system program is stored in the server, and characterized in that when the processor executes the computer program, the steps of the method described in any one of the above claims are implemented.

[0043] The transmission line ice dancing warning method and system based on data joint features first collects meteorological data and line tension data, and then extracts wind direction frequency, wind speed and wind direction joint probability, humidity fluctuation amplitude, air pressure and wind speed correlation, tension fluctuation amplitude and other features, and inputs these features into the DQN algorithm model as environmental states. The algorithm outputs response actions according to the set action definition, and at the same time trains the algorithm according to the reward mechanism so that it can optimally select actions according to the environmental state to achieve the best warning effect. Finally, the operation and maintenance personnel will feedback the de-icing effect and line dancing after taking measures to the system, and the system will optimize the action definition accordingly. This cycle repeats continuously to continuously improve the warning process and methods, improve the accuracy and effectiveness of warnings, and ensure the safe and stable operation of transmission lines.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] The integration of meteorological data and line tension data is richer and more comprehensive than traditional technical data sources. The various parameters in the meteorological data are interrelated and have different degrees of influence on the dancing of ice. The combined effect of wind speed and wind direction can determine the aerodynamic conditions of the line. Changes in humidity and air pressure affect the conditions for ice formation, and line tension data directly reflects the changes in the stress state of the line. These data together constitute a comprehensive and detailed information network, laying a solid foundation for accurate early warning.

[0046] The extraction of multiple features, which are interrelated and complementary to each other, can analyze the potential mechanism of ice dancing more deeply and comprehensively compared with existing technologies of single feature or simple data processing, and provide more valuable decision-making basis.

[0047] The extracted features are imported into the algorithm model as environmental states, combined with clear action definitions, and through the continuous input of large amounts of data and training based on reward mechanisms, the algorithm can gradually learn and optimize decision-making strategies. When faced with complex and changing line environments, it is no longer limited to fixed rules or simple threshold judgments.

[0048] The establishment of a feedback mechanism further improves the adaptability and reliability of the system. After taking countermeasures, the operation and maintenance personnel will feed back information such as the de-icing effect and whether the line is still dancing to the early warning system. If it is found that the countermeasures in the existing action definition are not effective, the system can make adjustments quickly. Through this cycle of continuous self-learning and optimization, the system can continue to adapt to changes in different line environments, different climatic conditions, and different ice-covered dancing characteristics. It also helps to optimize the allocation of operation and maintenance resources, reduce long-term operation and maintenance costs, and improve the overall safety and stability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 A flow chart of a data-joint feature transmission line ice dancing warning method of the present invention; DETAILED DESCRIPTION

[0050] The technical solutions in the embodiments of the present invention will be fully described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0051] like Figure 1 As shown, a method for early warning of ice dancing of power transmission lines based on data joint features is characterized by comprising the following steps:

[0052] S1: Data collection;

[0053] S2: Feature extraction;

[0054] S3: Action definition;

[0055] S4: Algorithm training;

[0056] S5: Establish a feedback mechanism.

[0057] Step S1 data collection includes meteorological data and environmental data; the meteorological data includes wind speed, wind direction, temperature, humidity, and air pressure information along the transmission line obtained from the local meteorological department, and recorded at certain time intervals to form time series data; the line data collects the line tension through the tension sensor installed on the transmission line.

[0058] Step S2: feature extraction includes the frequency of wind direction, the joint probability of wind speed and wind direction interval, humidity fluctuation amplitude, correlation between air pressure and wind speed, and tension fluctuation amplitude;

[0059] The specific calculation process of the frequency of occurrence of the wind direction is as follows:

[0060] The wind direction data is divided into certain angle intervals, and the wind direction angle 0°-360° is divided into 16 intervals, each interval is 22.5°; the frequency of wind direction in each interval is counted f i , and its calculation formula is:

[0061]

[0062] where n i is the number of data points whose wind direction falls within the i-th interval, and n is the total number of data points;

[0063] The specific calculation process of the joint probability of the wind speed and wind direction interval is as follows:

[0064] According to the definition of probability density function, for any wind speed interval [a, b] and wind direction interval [c, d], the following formula is implemented:

[0065]

[0066] Where v represents wind speed, θ represents wind direction, a and b are the lower and upper limits of the wind speed interval, c and d are the lower and upper limits of the wind direction interval, which are used to define the range of wind directions for which the probability is to be calculated. f(v,θ) is the joint probability distribution function of wind speed v and direction θ. By performing double integration on the corresponding interval, Get the joint probability of the specified wind speed and direction interval;

[0067] The kernel density estimation method is used to estimate the specific form of the joint probability distribution function f(v,θ), and its calculation formula is:

[0068]

[0069] Where n is the number of data points, K h (x,y) is the kernel function, v i and θi are the wind speed and direction values ​​of the ith data point.

[0070] Step S2: The specific calculation process of the humidity fluctuation amplitude is as follows:

[0071] Define the humidity fluctuation amplitude σ H , the formula is:

[0072]

[0073] Among them, σ H is the humidity function, which indicates the change of humidity over time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean humidity.

[0074] Step S2: The specific calculation process of the correlation between air pressure and wind speed is as follows:

[0075] Considering the correlation between air pressure and wind speed, their covariance calculation formula is:

[0076]

[0077] Among them, P(t) is the air pressure function, which indicates the change of air pressure with time t, v(t) is the wind speed function, which indicates the change of wind speed with time t, t 1 and t 2 are the start time and end time of the time period considered, which are used to define the time range for calculating the covariance. is the mean air pressure, is the mean of wind speed, Cov(P,v) is the covariance of air pressure and wind speed, which reflects the coordinated changes of the two variables of air pressure and wind speed in the same time period.

[0078] Step S2: Feature extraction: The specific calculation process of the tension fluctuation amplitude is as follows:

[0079] Defining the tension fluctuation amplitude The formula is:

[0080]

[0081] Among them, F tENSION (t) is the line tension function, which indicates the change of line tension with time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean tension.

[0082] The action definition of step S3 is as follows:

[0083] The early warning measures taken are taken as actions, issuing early warning signals of different levels, arranging operation and maintenance personnel to conduct on-site inspections, and initiating emergency protection measures. The actions use discrete values ​​to represent different actions, which are specifically:

[0084]

[0085] Step S4 algorithm training is specifically as follows: the previously calculated values ​​of the transmission line, including the joint probability distribution of wind speed and wind direction, humidity fluctuation amplitude, correlation between air pressure and wind speed, and line temperature and tension parameter characteristics are imported into the DQN algorithm model as environmental states, and the defined actions are output; by continuously inputting environmental states and corresponding actions, training is performed according to the set reward mechanism, specifically, when the environmental state at a certain moment is input, the algorithm selects an action according to the current strategy, and if the subsequent actual situation shows that this action is correct, the algorithm is given a corresponding reward, otherwise it is punished; through such a training process, the DQN algorithm can optimally select appropriate actions according to different environmental states to achieve the best early warning effect.

[0086] Step S5 establishes a feedback mechanism, specifically, after the operation and maintenance personnel take countermeasures, they feed back the de-icing effect and whether the line is still dancing to the early warning system. If it is found that the countermeasures taken by the operation and maintenance personnel are not effective and the line still frequently dances after de-icing, the action definition of the early warning system can be optimized, and more effective countermeasure options can be added to enable the algorithm to better select appropriate actions. By continuously optimizing according to feedback information, the entire early warning process and methods are continuously improved, thereby improving the accuracy and effectiveness of the early warning of ice-covered transmission lines.

[0087] A transmission line ice dancing warning system based on data joint features, characterized in that it includes a server and a processor, wherein a system program is stored in the server, and characterized in that when the processor executes the computer program, the steps of the method described in any one of claims 1 to 9 are implemented.

[0088] The specific implementation method is:

[0089] S1: Data collection; the meteorological data includes wind speed, wind direction, temperature, humidity, and air pressure information along the transmission line obtained from the local meteorological department, and recorded at certain time intervals to form time series data; the line data collects the line tension through the tension sensor installed on the transmission line.

[0090] S2: Feature extraction; including the frequency of wind direction, the joint probability of wind speed and wind direction range, humidity fluctuation amplitude, the correlation between air pressure and wind speed, and tension fluctuation amplitude;

[0091] The specific calculation process of the frequency of occurrence of the wind direction is as follows:

[0092] The wind direction data is divided into certain angle intervals, and the wind direction angle 0°-360° is divided into 16 intervals, each interval is 22.5°; the frequency of wind direction in each interval is counted f i , and its calculation formula is:

[0093]

[0094] where n i is the number of data points whose wind direction falls within the i-th interval, and n is the total number of data points;

[0095] According to the definition of probability density function, for any wind speed interval [a, b] and wind direction interval [c, d], the following formula is implemented:

[0096]

[0097] Where v represents wind speed, θ represents wind direction, a and b are the lower and upper limits of the wind speed interval, c and d are the lower and upper limits of the wind direction interval, which are used to define the range of wind directions for which the probability is to be calculated. f(v,θ) is the joint probability distribution function of wind speed v and direction θ. By performing double integration on the corresponding interval, Get the joint probability of the specified wind speed and direction interval;

[0098] The kernel density estimation method is used to estimate the specific form of the joint probability distribution function f(v,θ), and its calculation formula is:

[0099]

[0100] Where n is the number of data points, K h (x,y) is the kernel function, v i and θ i are the wind speed and direction values ​​of the ith data point.

[0101] Define the humidity fluctuation amplitude σ H , the formula is:

[0102]

[0103] Among them, σ H is the humidity function, which indicates the change of humidity over time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean humidity.

[0104] Step S2: The specific calculation process of the correlation between air pressure and wind speed is as follows:

[0105] Considering the correlation between air pressure and wind speed, their covariance calculation formula is:

[0106]

[0107] Among them, P(t) is the air pressure function, which indicates the change of air pressure with time t, v(t) is the wind speed function, which indicates the change of wind speed with time t, t 1 and t 2 are the start time and end time of the time period considered, which are used to define the time range for calculating the covariance. is the mean air pressure, is the mean of wind speed, Cov(P,v) is the covariance of air pressure and wind speed, which reflects the coordinated changes of the two variables of air pressure and wind speed in the same time period.

[0108] Define the tension fluctuation amplitude σ Ftension , the formula is:

[0109]

[0110] Among them, F tENSION (t) is the line tension function, which indicates the change of line tension with time t. 1 and t 2 are the start and end time of the time period under consideration, is the mean tension.

[0111] S3: Action definition; specifically:

[0112] The early warning measures taken are taken as actions, issuing early warning signals of different levels, arranging operation and maintenance personnel to conduct on-site inspections, and initiating emergency protection measures. The actions use discrete values ​​to represent different actions, which are specifically:

[0113]

[0114] S4: Algorithm training; specifically: import the previous calculated values ​​of the transmission line, including the joint probability distribution of wind speed and direction, humidity fluctuation amplitude, correlation between air pressure and wind speed, and line temperature and tension parameter characteristics as environmental states into the DQN algorithm model, and take the defined actions as outputs; by continuously inputting environmental states and corresponding actions, training is performed according to the set reward mechanism. Specifically, when the environmental state at a certain moment is input, the algorithm selects an action according to the current strategy. If the subsequent actual situation shows that this action is correct, the algorithm is given a corresponding reward, otherwise it is punished; through such a training process, the DQN algorithm can optimally select appropriate actions according to different environmental states to achieve the best early warning effect.

[0115] S5: Establish a feedback mechanism. Specifically, after the operation and maintenance personnel take countermeasures, they will feed back the de-icing effect and whether the line is still dancing to the early warning system. If it is found that the countermeasures taken by the operation and maintenance personnel are not effective and the line is still frequently dancing after de-icing, the action definition of the early warning system can be optimized and more effective countermeasure options can be added to enable the algorithm to better select appropriate actions. By continuously optimizing according to feedback information, the entire early warning process and methods are continuously improved to improve the accuracy and effectiveness of the ice-covered transmission line dancing warning.

Claims

1. A method for early warning of ice dancing on power transmission lines based on data joint features, characterized in that: The following steps are involved: S1: Data collection; S2: Feature extraction; S3: Action definition; S4: Algorithm training; S5: Establish a feedback mechanism.

2. According to claim 1, a method for early warning of ice dancing of power transmission lines based on data joint features is characterized in that: Step S1 data collection includes meteorological data and environmental data; the meteorological data includes wind speed, wind direction, temperature, humidity, and air pressure information along the transmission line obtained from the local meteorological department, and recorded at certain time intervals to form time series data; the line data collects the line tension through the tension sensor installed on the transmission line.

3. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 1 is characterized in that: Step S2: feature extraction includes the frequency of wind direction, the joint probability of wind speed and wind direction interval, humidity fluctuation amplitude, correlation between air pressure and wind speed, and tension fluctuation amplitude; The specific calculation process of the frequency of occurrence of the wind direction is as follows: The wind direction data is divided into certain angle intervals, and the wind direction angle 0°-360° is divided into 16 intervals, each interval is 22.5°; the frequency of wind direction in each interval is counted f i , and its calculation formula is: where n i is the number of data points whose wind direction falls within the i-th interval, and n is the total number of data points; The specific calculation process of the joint probability of the wind speed and wind direction interval is as follows: According to the definition of probability density function, for any wind speed interval [a, b] and wind direction interval [c, d], the following formula is implemented: Where v represents wind speed, θ represents wind direction, a and b are the lower and upper limits of the wind speed interval, c and d are the lower and upper limits of the wind direction interval, which are used to define the range of wind directions for which the probability is to be calculated. f(v,θ) is the joint probability distribution function of wind speed v and direction θ. By performing double integration on the corresponding interval, Get the joint probability of the specified wind speed and direction interval; The kernel density estimation method is used to estimate the specific form of the joint probability distribution function f(v,θ), and its calculation formula is: Where n is the number of data points, Kh(x,y) is the kernel function, and v i and θ i are the wind speed and direction values ​​of the ith data point.

4. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 3 is characterized in that: Step S2: The specific calculation process of the humidity fluctuation amplitude is as follows: Define the humidity fluctuation amplitude σ H , the formula is: Among them, σ H is the humidity function, which indicates the change of humidity with time t. t1 and t2 are the start and end time of the time period considered, respectively. is the mean humidity.

5. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 3 is characterized in that: Step S2: The specific calculation process of the correlation between air pressure and wind speed is as follows: Considering the correlation between air pressure and wind speed, their covariance calculation formula is: Among them, P(t) is the air pressure function, which indicates the change of air pressure with time t, v(t) is the wind speed function, which indicates the change of wind speed with time t, t1 and t2 are the start time and end time of the time period considered, which are used to define the time range for calculating covariance. is the mean air pressure, is the mean of wind speed, Cov(P,v) is the covariance of air pressure and wind speed, which reflects the coordinated changes of the two variables of air pressure and wind speed in the same time period.

6. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 3 is characterized in that: Step S2: Feature extraction: The specific calculation process of the tension fluctuation amplitude is as follows: Defining the tension fluctuation amplitude The formula is: Among them, F tENSION (t) is the line tension function, which indicates the variation of line tension with time t. t1 and t2 are the start and end time of the time period considered, respectively. is the mean tension.

7. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 1 is characterized in that: The action definition of step S3 is as follows: The early warning measures taken are taken as actions, issuing early warning signals of different levels, arranging operation and maintenance personnel to conduct on-site inspections, and initiating emergency protection measures. The actions use discrete values ​​to represent different actions, which are specifically:

8. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 1 is characterized in that: Step S4 algorithm training is specifically as follows: the previously calculated values ​​of the transmission line, including the joint probability distribution of wind speed and wind direction, humidity fluctuation amplitude, correlation between air pressure and wind speed, and line temperature and tension parameter characteristics are imported into the DQN algorithm model as environmental states, and the defined actions are output; by continuously inputting environmental states and corresponding actions, training is performed according to the set reward mechanism, specifically, when the environmental state at a certain moment is input, the algorithm selects an action according to the current strategy, and if the subsequent actual situation shows that this action is correct, the algorithm is given a corresponding reward, otherwise it is punished; through such a training process, the DQN algorithm can optimally select appropriate actions according to different environmental states to achieve the best early warning effect.

9. The method for early warning of ice dancing of power transmission lines based on data joint features according to claim 1 is characterized in that: Step S5 establishes a feedback mechanism, specifically, after the operation and maintenance personnel take countermeasures, they feed back the de-icing effect and whether the line is still dancing to the early warning system. If it is found that the countermeasures taken by the operation and maintenance personnel are not effective and the line still frequently dances after de-icing, the action definition of the early warning system can be optimized, and more effective countermeasure options can be added to enable the algorithm to better select appropriate actions. By continuously optimizing according to feedback information, the entire early warning process and methods are continuously improved, thereby improving the accuracy and effectiveness of the early warning of ice-covered transmission lines.

10. A transmission line ice dancing warning system based on data joint features, characterized in that: The method comprises a server and a processor, wherein the server stores a system program, and is characterized in that the processor implements the steps of the method according to any one of claims 1 to 9 when executing the computer program.

Images