Power transmission line bird-related risk monitoring method, system and equipment based on multi-behavior identification

Abstract

The invention discloses a power transmission line bird-related risk monitoring method, system and device based on multi-behavior recognition, and relates to the technical field of power transmission line protection monitoring, and the method comprises the following steps: obtaining a multi-modal monitoring data flow of bird activities in a target area; identifying a bird basic behavior event related to the power transmission line risk from the multi-modal monitoring data stream; based on the basic behavior event, generating a bird-related risk quantitative index curve with continuous time in real time; performing time sequence analysis on the bird-related risk quantitative index curve, and dynamically delimiting a risk time period matched with a bird activity rhythm by detecting a waveform interval conforming to a preset risk mode characteristic in the curve; and adjusting a protection strategy of the power transmission line based on the dynamically delimited risk time period. According to the invention, continuous quantitative characterization of the bird line-related risk and dynamic delimitation of the risk time period are realized, the timeliness and pertinence of bird-related risk monitoring are improved, and active protection and safe operation of a power transmission line are facilitated.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power transmission line protection monitoring, and more particularly, to a power transmission line bird-related risk monitoring method, system and device based on multi-behavior identification. BACKGROUND

[0002] To ensure the safe and stable operation of the power grid, the precise early warning and active protection of bird-related faults of power transmission lines have become an urgent need of the power system. Bird activities have significant seasonal regularity, and different behavior patterns such as breeding, clustering and migration can directly cause differentiated line risks (such as short circuit caused by nest building and flashover caused by excretion). Therefore, developing a technical method that can automatically and finely identify bird behavior and analyze its seasonal dynamics is the key to realizing the transition from passive disposal to active early warning and precise protection mode, and is of great significance to improving the reliability of power supply of the power grid.

[0003] At present, although automatic monitoring means based on sensors (such as acoustic recorders, visible light and thermal imaging cameras) have been widely used, existing technologies mainly rely on the statistics of single and isolated behaviors (such as the number of calls), and cannot identify composite behavior sequences composed of multiple basic behaviors (such as specific calls, display postures and pursuit flying), resulting in a lag or rough determination of the high-risk activity stage of birds (such as the breeding period clustering and migration peak). At the same time, due to the lack of dynamic identification ability of composite behavior sequences, existing methods cannot real-time define the risk time period according to the real rhythm of bird activities, but only rely on fixed astronomical or meteorological seasons as monitoring reference, resulting in lack of pertinence of inspection and protection strategy, and difficulty in meeting the requirements of high timeliness and accuracy of power grid safety. SUMMARY

[0004] In order to overcome the above-mentioned defects of the prior art, the embodiments of the present application provide a power transmission line bird-related risk monitoring method, system and device based on multi-behavior identification, which generates a continuous risk curve in real time by identifying composite behavior sequences composed of multiple basic behaviors, and dynamically defines a risk time period matched with the real activity rhythm of birds accordingly, and finally realizes active and accurate adjustment of protection strategy, to solve the technical problems of inaccurate early warning and lack of timeliness of response of existing methods.

[0005] To achieve the above-mentioned purpose, the present application provides the following technical solutions: The application discloses a power transmission line bird-related risk monitoring method based on multi-behavior identification, which comprises the following steps: acquiring a multi-modal monitoring data stream of bird activities in a target area; identifying bird basic behavior events related to power transmission line risks from the multi-modal monitoring data stream; generating a time-continuous bird-related risk quantitative index curve in real time based on the basic behavior events; performing time series analysis on the bird-related risk quantitative index curve, dynamically determining a risk time period matched with the bird activity rhythm by detecting a waveform interval in the curve that meets a preset risk mode characteristic; and adjusting a protection strategy of the power transmission line based on the dynamically determined risk time period.

[0006] In a preferred embodiment, the multi-modal monitoring data stream of bird activities in the target area is acquired by continuously collecting audio data through an acoustic sensor and comparing the audio data with a preset bird key behavior acoustic feature library in real time to generate a trigger instruction; the trigger instruction synchronously wakes up a visible light camera and a thermal infrared sensor deployed in the same monitoring node to obtain visual-thermal infrared dual-modal behavior data; and the dual-modal behavior data is spatio-temporally aligned based on a time stamp of the acoustic sensor to obtain the multi-modal monitoring data stream.

[0007] In a preferred embodiment, the bird basic behavior events related to the power transmission line risks are identified from the multi-modal monitoring data stream by dynamically determining a behavior attention time window based on the instantaneous acoustic feature change detected by the acoustic sensor; performing candidate segment extraction on the visual-thermal infrared dual-modal behavior data in the behavior attention time window to obtain a time-constrained behavior candidate set; constructing a time folding sequence based on the behavior candidate set and identifying a basic behavior event combination meeting a preset time sequence logic relationship according to a behavior event sequence relationship in the time folding sequence.

[0008] In a preferred embodiment, the time folding sequence is constructed based on the behavior candidate set by cross-window associating the candidate segments identified in the multiple behavior attention time windows based on their time stamps; taking the instantaneous acoustic feature change point corresponding to each behavior attention time window as a reference origin point, folding and mapping the actual physical time axis occupied by the associated candidate segments to a normalized logical time axis with the change point as the origin; and generating the time folding sequence on the normalized logical time axis according to the types of the candidate segments and their relative positions.

[0009] In a preferred embodiment, the time-continuous bird-related risk quantification index curve is generated in real time based on the basic behavior events, specifically comprising: constructing a state set representing the evolution of bird behavior based on the identified combination of basic behavior events in the time-folded sequence; statistically analyzing the duration of each state on the normalized logical time axis and the transition relationship between adjacent states to obtain state transition probability characteristics; fusing the state transition probability characteristics with the synchronously obtained environmental information to form a state observation vector; inputting the state observation vector into a Bayesian state space model to online estimate the activation probability of the bird-related risk; taking the confidence upper limit of the activation probability as the risk quantification index and continuously outputting along the time axis to form the bird-related risk quantification index curve.

[0010] In a preferred embodiment, the risk time period matched with the bird activity rhythm is dynamically delimited, specifically comprising: performing second-order difference operation on the quantification index curve with respect to time to obtain an acceleration sequence, and locating the first significant positive jump point in the acceleration sequence, denoted as a potential risk starting point; determining a corresponding risk closing signal from the potential risk starting point; and defining, as a risk time period, the interval from the potential risk starting point to the risk closing signal on the time axis triggered by the risk closing signal.

[0011] In a preferred embodiment, the determination of the risk closing signal specifically comprises: synchronously monitoring the instantaneous acceleration and cumulative slope of the curve from the potential risk starting point; and generating the risk closing signal when the instantaneous acceleration decreases and stabilizes around zero within a preset monitoring period, and the growth of the cumulative slope remains a constant proportion.

[0012] In a preferred embodiment, the adjustment of the protection strategy of the power transmission line comprises: executing an initial protection strategy at the beginning of the risk time period based on the identified bird species and their expected interference response model; monitoring the change of the risk quantification index curve in real time, calculating a defense attenuation coefficient representing the degree of attenuation of protection effectiveness by comparing the actual decline rate with the model standard value; and determining the failure of the strategy and starting reconstruction if the defense attenuation coefficient continuously exceeds the failure threshold.

[0013] The system of the bird-related risk monitoring method for the power transmission line based on multi-behavior recognition comprises: a data sensing unit for acquiring multi-modal monitoring data streams of bird activities in a target area; a behavior recognition unit for identifying bird basic behavior events related to the risk of the power transmission line from the multi-modal monitoring data streams; a risk quantification unit for generating a time-continuous bird-related risk quantification index curve in real time based on the basic behavior events; an interval division unit for performing time series analysis on the bird-related risk quantification index curve, dynamically dividing a risk time period matched with the rhythm of bird activities by detecting waveform intervals in the curve that meet preset risk pattern characteristics; and a strategy adjustment unit for adjusting the protection strategy of the power transmission line based on the dynamically divided risk time period.

[0014] The bird-related risk monitoring device for the power transmission line based on multi-behavior recognition comprises a memory and a processor: the memory is used for storing programs; and the processor is used for executing the programs to realize each step of the bird-related risk monitoring method for the power transmission line based on multi-behavior recognition.

[0015] The technical effects and advantages of the bird-related risk monitoring method, system and device for the power transmission line based on multi-behavior recognition of the present application are as follows: 1. The present application solves the problems of risk early warning lag and rough judgment caused by only being able to count single and isolated behaviors in the prior art by fusing the identified multiple bird basic behavior events and generating a time-continuous risk quantification index curve, upgrades discrete and fuzzy bird-related risk perception to continuous and quantitative situation assessment of the dynamic evolution process of the risk, and thus realizes the technical effect of active and accurate early warning.

[0016] 2. The present application solves the problems of disconnection between the protection strategy and the actual risk and lack of pertinence caused by rigidly relying on fixed astronomical or meteorological seasons in the prior art by dynamically dividing a risk time period matched with the bird activities by analyzing the waveform characteristics of the risk quantification index curve, changes the core time reference of risk protection from a static calendar to a dynamic real activity rhythm of birds, and thus drives the accurate delivery of protection resources in the best time window, greatly improves the protection efficiency and pertinence. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 It is a flowchart of the bird-related risk monitoring method for the power transmission line based on multi-behavior recognition of the present application; Figure 2 It is a curve graph of the quantification index changing with time in the present application; Figure 3 It is a time sequence diagram of the multi-modal monitoring data stream cooperative collection in the present application; Figure 4 It is a system structure schematic diagram of the bird-related risk monitoring method for the power transmission line based on multi-behavior recognition of the present application; Figure 5 A structural block diagram of an exemplary electronic device that can be used to implement embodiments of the present disclosure is provided for embodiments of the present application. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the scope of the present application.

[0019] Embodiment 1, Figure 1 The bird risk monitoring method for a power transmission line based on multi-behavior recognition is given, including the following steps: S1, obtaining a multi-modal monitoring data stream of bird activities in a target area; In this embodiment, the multi-modal monitoring data stream of bird activities in the target area is specifically obtained as follows: The multi-modal monitoring data stream of bird activities in the target area is obtained by deploying acoustic sensors, visible light cameras and thermal infrared sensors on the monitoring nodes. The acoustic sensors continuously collect audio data of the target area with a sampling frequency of 44.1 kHz and a resolution of 16 bits, and calculate acoustic features such as short-time energy, spectral centroid and mel-frequency cepstral coefficient through a built-in feature extraction module. The system compares the real-time extracted acoustic features with a pre-constructed bird key behavior acoustic feature library. The feature library is obtained by feature extraction (such as mel-frequency cepstral coefficient MFCC, spectral centroid) and machine learning training on the recordings of the target bird species (such as white egret, red-crowned crane) such as calling, courtship sound and alarm sound. When the matching degree exceeds the preset threshold, the acoustic sensor generates a trigger instruction to immediately wake up the visible light camera and thermal infrared sensor on the same monitoring node for data collection.

[0020] The collection resolutions of the visible light camera and the thermal infrared sensor are 1920x1080 and 640x480 respectively, and the frame rates are both 30 fps. Under the action of the trigger instruction, the camera and the infrared sensor start collecting almost synchronously, obtain visual-thermal infrared dual-modal behavior data, and attach the time stamp of the acoustic trigger event to the collected image frames to ensure the accurate correspondence of the visual and infrared data with the acoustic event in time. During the collection process, the monitoring node performs basic denoising and enhancement processing on the visual and infrared frames, and buffers the frames in the local cache to ensure data integrity under short time delay. After the collection is completed, the visible light camera and the thermal infrared sensor enter the sleep state again and wait for the next trigger.

[0021] Further, the system maps the visual and infrared frames onto a unified timeline with the timestamps of the acoustic sensor as the reference. For continuous behaviors across frames, the system aligns the time difference by interpolation, achieving a precision of ±10 milliseconds. Spatially, the system corrects the field-of-view difference between the camera and the infrared sensor, making the two- or three-dimensional location of each behavior event correspond to the time information. Finally, the system generates a structured multi-modal monitoring data stream, where each data unit includes the timestamp, acoustic features, visible light image frames, infrared thermal map, and location information.

[0022] S2, identifying bird basic behavior events related to power line risks from the multi-modal monitoring data stream; In this embodiment, the identification of bird basic behavior events related to power line risks from the multi-modal monitoring data stream includes: S21, dynamically determining a behavior attention time window. The exact moment when the acoustic sensor generates the trigger instruction is recorded as , i.e., the point of instantaneous acoustic feature change, and a behavior attention time window is dynamically determined based on :

[0023] wherein is the pre-observation duration (e.g., 1.0 second) for capturing the preparatory posture before the acoustic behavior, is the subsequent observation duration (e.g., 5.0 seconds) for capturing the visual feedback triggered by the acoustic behavior.

[0024] S22, extracting a time-constrained behavior candidate set. Within the behavior attention time window , the synchronized visual and thermal infrared video frames are analyzed. A pre-trained dual-stream neural network (e.g., one branch processes RGB images and the other branch processes thermal infrared temperature matrices) is used to classify the basic behaviors of each frame or a short segment, identifying visual behavior primitives such as “flapping wings”, “lowering head to forage”, “flying”, “courtship dance”, etc. If a behavior primitive is identified with high confidence in consecutive frames (e.g., more than 5 frames), it is extracted as a candidate segment , each candidate segment contains the behavior type, the start time and end time on the physical time axis, the confidence, etc. All candidate segments extracted from the same window constitute a time-constrained behavior candidate set .

[0025] S23, Construct a time-folded sequence. Within a relatively long timeframe (e.g., a morning), multiple behavioral attention time windows, triggered by various acoustic events and arranged chronologically, are captured. Let the behavioral candidate set for the k-th behavioral attention time window be denoted as . If two segments and If two candidate segments are of the same type and their physical time interval is less than a preset global behavior maximum interval (e.g., 5 minutes), then these two candidate segments are considered to be part of the same continuous behavior and are logically associated. For any associated candidate segment... Regardless of which behavioral focus window it originates from All of them are time-normalized relative to the acoustic trigger point of their respective windows, and their logical time is calculated. :

[0026]

[0027] in, Candidate segments The absolute start time, Candidate segments The end of absolute time, For the k-th window The absolute time of the corresponding acoustic trigger point Candidate segments The logical start time, Candidate segments The logical end time. This operation projects all segments from the absolute physical timeline onto a relative logical timeline with their respective acoustic events as zero points.

[0028] Logical start time of all related segments Sort them in ascending order, and generate a time-folded sequence based on their behavior type and logical sequence relationship. Time-folded sequence This describes the order in which behavioral segments spanning multiple physical windows are arranged on a unified behavioral logical timescale. For example, a time-folded sequence of a courtship behavior might be represented as:

[0029] in, , , It is the logical time relative to each trigger point, and .

[0030] S24, identify the basic behavior event combination. A number of target composite behavior event templates (such as complete courtship display, cooperative nest building) are pre-set, each of which specifies a set of logical time sequence relationships that must be met by a group of basic behavior types. The is matched with these templates, if all behavior types specified in the template appear in sequence in the video, and the logical time interval between adjacent behaviors falls within the range allowed by the template, it is determined that an occurrence of the composite behavior event has occurred, and this event is taken as the basic behavior event combination.

[0031] The time folding sequence analysis method proposed in the present application takes the acoustic feature point as the unified time reference origin, and maps the visual behavior segments occurring at different times to the normalized logical time axis for comparison. This method enables the accurate identification of complex risk behavior patterns with fixed timing logic, such as "1-3 seconds after issuing an alarm call, a dive behavior towards the insulator appears". The time folding sequence analysis method strips away the absolute time of behavior occurrence and magnifies the relative timing relationship. The system can automatically discover that although a certain type of risk behavior occurs on different dates and at different times, they exhibit similar segment sequences on the "logical time axis". This enables the system not only to identify known patterns, but also to have the potential to cluster and warn new, unknown but fixed-logic risk behavior patterns, improving the system's adaptability and intelligence level.

[0032] S3, based on the basic behavior events, real-time generation of time-continuous bird-related risk quantification index curve; In the present embodiment, the calculation step of the quantification index specifically comprises: Assuming that the system has identified a sequence of composite behavior events arranged in chronological order, for example (courtship), (nest building), (incubation), as the basic behavior event combination in the current analysis period, the transition frequency between each type of event in the basic behavior event combination is counted, for example, if the number of times that the "courtship" event is followed by the "nest building" event is , and the total number of times that the "courtship" event occurs is , then the empirical transition probability from state "courtship" to state "nest building" can be estimated as . This calculation is performed for all possible event type pairs in the basic behavior event combination to obtain a state transition probability matrix. The upper triangular or lower triangular elements (excluding the diagonal line) are extracted from the matrix to form a one-dimensional feature vector representing the uncertainty of event transition. For each type of event The duration of all observed instances is calculated, and the variance of these durations is determined. A larger variance indicates greater instability and uncertainty in the execution duration of this type of behavior. The duration variances of all event types are then combined into a feature vector. Simultaneously collect environmental sensor data, such as daily average temperature, precipitation, and sunshine duration. During analysis, these environmental parameters within the current observation window are normalized to form an environmental feature vector. .Will , and The three vectors are concatenated to form the final multidimensional uncertainty feature vector. , where the superscript T indicates transpose.

[0033] The pre-trained Bayesian state-space model treats the "bird-related risk activation level," which cannot be directly observed, as a hidden state. (A latent variable between 0 and 1) will be observed These are considered as observations. The model defines state transition equations (describing...) How it potentially evolves over time) and observation equations (describing a given...) Observed at time (The probability). The current... The data is input into a pre-trained Bayesian state-space model, based on all historical observations. Hiding the state using Bayesian filtering algorithms (such as particle filtering) Online, recursive posterior probability distribution estimation is performed, typically approximated by a beta distribution. The activation probability of bird-related risks is then calculated from the obtained posterior probability distribution. , Typically, the mean or mode of the beta distribution is taken. To provide a conservative and robust risk measure, the system uses the upper confidence limit of the posterior probability distribution (e.g., the upper bound of the 95% confidence interval) as the final risk quantification indicator. .

[0034] The system outputs at a fixed frequency (e.g., once per minute). The values ​​are calculated and connected chronologically to generate a continuous temporal risk quantification curve for bird-related activities. This curve smoothly reflects the dynamic trend of risk levels, providing a direct quantitative basis for subsequent dynamic time period delineation.

[0035] S4. Perform time-series analysis on the bird-related risk quantification index curve. By detecting waveform intervals in the curve that conform to the characteristics of a preset risk pattern, dynamically define the risk time period that matches the activity rhythm of birds. In this embodiment, the dynamic delineation of risk time periods that match the bird's activity rhythm specifically includes: Discrete sampling (such as one mean point per day) is performed on the risk quantification index curve to form a smooth mean time series , where N is the number of days. A second-order difference operation is performed on the mean time series, specifically a central difference method, to approximately calculate the daily instantaneous acceleration , and then the acceleration sequence is obtained:

[0036] , where reflects the change in the growth rate of the risk itself, and when , it indicates that the risk is in an accelerated accumulation state.

[0037] A sliding window with a length of days is set, and within the sliding window, the acceleration mean of the current window and the acceleration mean of the past window with the same length are calculated to determine whether is significantly greater than , and itself is greater than a small positive threshold to ensure that it is indeed a positive acceleration. The time when the above conditions are first met in the entire acceleration sequence is recorded as the potential risk starting point . Starting from the potential risk starting point , the instantaneous acceleration and the cumulative slope of the curve are monitored simultaneously, and the average slope of the risk accumulation growth from the starting point to the current day relative to time, i.e., the cumulative slope , is calculated:

[0038] When the instantaneous acceleration decreases and stabilizes near zero within a preset monitoring period (such as 3 days), and the growth of the cumulative slope remains constant, it is determined that the process enters a steady state and generates a risk closure signal, and the first day (or the day after the last day) that meets the conditions is recorded as the end point of the risk . With the risk closure signal as the trigger, the interval from the potential risk starting point to the end point determined by the risk closure signal is defined as a risk time period on the time axis.

[0039] S5, based on the dynamically defined risk time period, adjusts the protection strategy of the power transmission line.

[0040] In this embodiment, the dynamically defined risk time period is used to adjust the protection strategy of the power transmission line, specifically: At the starting moment of the risk period, according to the bird species corresponding to the identified basic behavior event combination, a preset expected interference response model of the bird species is called, and an initial protection strategy is generated and executed based on the model. After executing the initial protection strategy, the change of the bird-related risk quantification index curve is monitored in real time, and the actual decline rate thereof is calculated; the actual decline rate is compared with the standard decline rate defined in the expected interference response model, and a real-time defense attenuation coefficient is calculated, which is used to quantify the actual performance attenuation degree of the current protection strategy. The greater the value (tending to 1), the higher the bird's tolerance to the current protection means, and the strategy tends to be ineffective. If the real-time defense attenuation coefficient continuously exceeds the preset invalidation threshold, the system determines that the current protection strategy is invalid, and automatically executes strategy reconstruction, including: in space, controlling adjacent tower equipment to form a coordinated interference field to drive or guide the bird group away from the core risk area; in time, switching the device triggering mode to a non-periodic triggering mode to break the adaptability formed by the birds.

[0041] Embodiment 2, Figure 4 A system of a power transmission line bird-related risk monitoring method based on multi-behavior identification is given, including: A data perception unit is configured to acquire multi-modal monitoring data streams of bird activities in a target area. A behavior identification unit is configured to identify bird basic behavior events related to power transmission line risks from the multi-modal monitoring data streams. A risk quantification unit is configured to generate a time-continuous bird-related risk quantification index curve in real time based on the basic behavior events. An interval division unit is configured to perform time series analysis on the bird-related risk quantification index curve, and dynamically divide a risk period matching the bird activity rhythm by detecting waveform intervals in the curve that meet preset risk pattern characteristics. A strategy adjustment unit is configured to adjust the protection strategy of the power transmission line based on the dynamically divided risk period.

[0042] Embodiment 3, a power transmission line bird-related risk monitoring device based on multi-behavior identification, as shown in Figure 5 It includes a memory and a processor: the memory is used to store programs; the processor is used to execute the programs, and realize any of the embodiments in embodiment 1.

[0043] Since the bird-related risk monitoring device for power transmission lines based on multi-behavior recognition introduced in this embodiment is the device used to implement the method in Embodiment 1 of the present application, based on the method introduced in Embodiment 1 of the present application, those skilled in the art can understand the specific implementation of the electronic device of this embodiment and its various forms of changes, so the method in this embodiment is not described in detail. As long as the device used to implement the method in this embodiment is implemented by those skilled in the art, it belongs to the scope of protection of the present application.

[0044] In Embodiment 4, from March 1, 2024 to June 30, 2024, three intelligent monitoring nodes integrating acoustics, visible light, and thermal infrared were deployed in the 500kV power transmission line corridor near a certain important wetland where bird droppings flashover and nesting short circuit accidents frequently occurred in history. The system monitored the bird species such as white storks and other flocking birds. The system identified and dynamically delineated the high-risk time period caused by bird nesting and flocking activities, and verified the adaptive protection effect. The system generated 832 effective acoustic triggers (mainly corresponding to alarm calls and flocking noise), successfully awakened and collected dual-mode video clips, and cumulatively identified the following basic behavior event combinations directly related to power transmission line risks, as shown in Table 1: Table 1

[0045] The system fused the above risk behavior event frequency, flock size (thermal infrared identification), and environmental humidity (easy to cause flashover) in "hours" as a unit, and calculated the "bird-related risk quantification index (range 0-100) in real time through the Bayesian state space model. The time-varying curve is shown in Figure 2 . The curve fluctuates smoothly in the low position (15-30 interval), corresponding to general bird activity, and the risk is controllable. The curve steeply rises around April 5, forming the first significant peak. This inflection point marks the beginning of accelerated risk accumulation, and the system can accurately locate this day as the starting point of potential risk through the second-order difference algorithm, corresponding to the beginning of the spring nesting and flocking activity peak. From late April to May, the curve fluctuates in the high position of 75-90, indicating that the risk is in a state of sustained high incidence. High-intensity protection is maintained during this period. The curve starts to show a stable downward trend around May 25, and the acceleration of the decline tends to zero. At this time, the risk closure signal is generated, and the end of the high-risk period is dynamically determined.

[0046] The system automatically outputs the time interval from April 5, 2024 to May 25, 2024 as the "high-risk period for bird-related activities in spring 2024". Compared with the traditional fixed-month-based (such as April to June as the bird prevention period) or manual inspection-based experience judgment, the dynamic determination based on data-driven is more accurate (avoiding too early or too late), and the resource allocation period is shortened by about 30%, and the efficiency is significantly improved.

[0047] The above formulas are dimensionless values calculated, and the formulas are obtained by collecting a large amount of data to simulate a formula of the latest real situation, and the preset parameters in the formula are set by a person skilled in the art according to the actual situation.

[0048] The above embodiments can be realized wholly or partially by software, hardware, firmware or any combination thereof. When realized by software, the above embodiments can be realized wholly or partially in the form of a computer program product.

[0049] Those skilled in the art can realize that the modules and algorithm steps of the examples described in combination with the embodiments disclosed herein can be realized by electronic hardware or a combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solutions. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0050] In addition, the functional modules in each embodiment of the present application can be integrated in one processing module, or each module can exist physically alone, or two or more modules can be integrated in one module.

[0051] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art can easily think of changes or replacements within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

[0052] Finally, the above is only a preferred embodiment of the present application and is not used to limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. A method for monitoring bird-related risks on power transmission lines based on multi-behavior recognition, characterized in that, Includes the following steps: Acquire multimodal monitoring data streams of bird activity within the target area; Identify basic bird behavioral events associated with transmission line risks from multimodal monitoring data streams; Based on basic behavioral events, a time-continuous quantitative indicator curve for bird-related risks is generated in real time. Time-series analysis of bird-related risk quantification index curves is performed. By detecting waveform intervals in the curves that conform to the characteristics of preset risk patterns, risk time periods that match the activity rhythms of birds are dynamically defined. Based on dynamically defined risk time periods, adjust the protection strategy for transmission lines.

2. The method for monitoring bird-related risks on power transmission lines based on multi-behavior recognition according to claim 1, characterized in that, The acquisition of the multimodal monitoring data stream of bird activity within the target area specifically includes: Audio data is continuously collected by acoustic sensors and compared in real time with a pre-set acoustic feature library of key bird behaviors to generate trigger commands. The trigger command synchronously wakes up the visible light camera and thermal infrared sensor deployed on the same monitoring node to obtain visual-thermal infrared dual-modal behavioral data; Spatiotemporal alignment of dual-modal behavior data is performed based on timestamps from acoustic sensors to obtain a multimodal monitoring data stream.

3. The method for monitoring bird-related risks on power transmission lines based on multi-behavior recognition according to claim 2, characterized in that, The identification of basic bird behavioral events related to transmission line risks from multimodal monitoring data streams includes: Dynamically determine the time window of interest for the behavior based on the instantaneous changes in acoustic features detected by acoustic sensors; Within the behavior attention time window, candidate fragment extraction is performed on the visual-thermal infrared dual-modal behavior data to obtain a temporally constrained behavior candidate set; A time-folded sequence is constructed based on the candidate behavior set, and the combination of basic behavior events that satisfies the preset temporal logical relationship is identified based on the sequential relationship of the behavior events in the time-folded sequence.

4. The method for monitoring bird-related risks on power transmission lines based on multi-behavior recognition according to claim 3, characterized in that, The construction of time-folded sequences based on the behavior candidate set specifically involves: Candidate segments identified within multiple time windows of behavioral focus will be associated across windows based on their timestamps. Using the instantaneous acoustic feature change point corresponding to the attention time window of each behavior as the reference origin, the actual physical time axis occupied by the associated candidate segments is folded and mapped to the normalized logical time axis with the change point as the origin. On the normalized logical time axis, a time-folded sequence is generated based on the type of candidate segments and their relative positions.

5. The method for monitoring bird-related risks on transmission lines based on multi-behavior recognition according to claim 4, characterized in that, The method of generating a time-continuous quantitative indicator curve for bird-related risks in real time based on basic behavioral events specifically includes: Based on the combination of basic behavioral events identified in the time-folded sequence, a set of states representing the evolution of bird behavior is constructed; By statistically analyzing the duration of each state on the normalized logic time axis and the transition relationship between adjacent states, the state transition probability characteristics are obtained. The state transition probability features are fused with synchronously acquired environmental information to form a state observation vector; The state observation vector is input into the Bayesian state space model to estimate the activation probability of bird-related risks online. The upper confidence limit of the activation probability is used as a risk quantification indicator and continuously output along the time axis to form a bird-related risk quantification indicator curve.

6. The method for monitoring bird-related risks on power transmission lines based on multi-behavior recognition according to claim 5, characterized in that, The dynamic delineation of risk time periods that match bird activity rhythms specifically includes: The second-order difference operation is performed on the curve of the quantitative indicator changing over time to obtain the acceleration sequence, and the first significant positive jump point in the acceleration sequence is located and recorded as the starting point of potential risk. Starting from the potential risk point, determine a corresponding risk closure signal; Triggered by a risk closure signal, a risk time period is defined by reversing the timeline from the potential risk starting point to the risk closure signal.

7. The method for monitoring bird-related risks on transmission lines based on multi-behavior recognition according to claim 6, characterized in that, The determination of the risk closure signal specifically includes: Starting from the potential risk point, the instantaneous acceleration and cumulative slope of the curve are monitored simultaneously; A risk closure signal is generated when the instantaneous acceleration decreases and stabilizes near zero within a preset monitoring period, and the cumulative slope increases at a constant rate.

8. The method for monitoring bird-related risks on transmission lines based on multi-behavior recognition according to claim 7, characterized in that, The adjusted protection strategy for transmission lines includes: Based on the identified bird species and their expected disturbance response models, an initial protection strategy is implemented at the start of the risk period. Real-time monitoring of changes in risk quantification indicator curves; by comparing the actual rate of decline with the model standard value, calculation of the defense attenuation coefficient, which characterizes the degree of attenuation of protective effectiveness; If the defense attenuation coefficient continues to exceed the failure threshold, the strategy is deemed to have failed and reconstruction is initiated.

9. A system for monitoring bird-related risks on power transmission lines based on multi-behavior recognition as described in any one of claims 1-8, characterized in that, include: The data sensing unit is used to acquire multimodal monitoring data streams of bird activity within the target area; A behavior recognition unit is used to identify basic bird behavioral events related to transmission line risks from multimodal monitoring data streams; The risk quantification unit is used to generate a time-continuous risk quantification indicator curve for bird-related activities in real time based on basic behavioral events. Interval division units are used to perform time-series analysis on bird-related risk quantification index curves. By detecting waveform intervals in the curve that conform to the characteristics of a preset risk pattern, risk time periods that match the activity rhythms of birds are dynamically defined. The strategy adjustment unit is used to adjust the protection strategy of transmission lines based on dynamically defined risk time periods.

10. A transmission line bird risk monitoring device based on multi-behavior recognition, characterized in that, Including memory and processor: The memory is used to store programs; The processor is used to execute the program to implement each step of the method for monitoring bird-related risks on transmission lines based on multi-behavior recognition as described in any one of claims 1-8.

Images