Distributed multi-sensing optical cable state monitoring method

Abstract

The invention discloses a distributed multi-sensing optical cable state monitoring method. According to the method, optical cable state data information is collected through various sensors, data fusion is carried out at the same time, the data information is transmitted to a data monitoring center in real time, and partial data is unloaded to an edge server in a partial unloading mode, so that the processing efficiency is improved. Wherein a multi-agent game reinforcement learning algorithm is utilized to optimize the system performance.

Description

Technical Field

[0001] The invention belongs to the field of power communication, and in particular relates to a distributed multi-sensor optical cable status monitoring method. Background Art

[0002] With the comprehensive construction of the power grid, the power transmission and distribution and operation business of the power grid are growing, and the business carried by a single optical cable is becoming increasingly concentrated. The operation risk of the communication system that may be caused by the optical cable failure has greatly increased. The means of monitoring and maintaining optical cable faults have developed in parallel with the scale application of optical communications, and have gone through a process from simple to perfect, from manual to gradually automated, and from mainly based on personal experience to more and more auxiliary tools. The earliest optical cable fault maintenance mainly relied on users to discover and report faults, and maintenance personnel to locate the optical cable and fiber core where the fault point was located; with the increase in the demand for operation and maintenance of optical communication networks, physical methods such as "fiber cutting method", "insertion method", "bending method" and "freezing method" have emerged for fault monitoring. The disadvantage is that there is physical destruction or damage to the optical fiber, and the workload is large and the positioning time is long.

[0003] Therefore, considering the complexity of the current power system optical cable laying environment, the traditional optical cable status monitoring method obviously cannot meet the needs well. Summary of the invention

[0004] The present invention aims to solve one of the technical problems existing in the prior art or related technology. The inventor found that multi-sensing technology can sense the state of the optical cable in real time by using various sensors, and through the data fusion of multiple sensors, comparative analysis can greatly improve the accuracy of the data, enhance the ability of fault diagnosis and prevention, and avoid the drawbacks of consuming a lot of manpower in traditional optical cable monitoring. At the same time, multi-sensing technology can make the system more adaptable to various working conditions, making the system more intelligent and improving the operating efficiency of the system. The integrated sensing and computing technology can transmit and calculate data while sensing the optical cable status data, and with the assistance of edge computing, it can greatly prevent and handle power system failures in advance, improve the safety and reliability of the system, and realize automatic monitoring and control of the power system, reduce manual intervention, and greatly improve the degree of automation.

[0005] Based on this, the present invention proposes a distributed multi-sensor optical cable status monitoring method, which uses multiple sensors to collect optical cable status data information and simultaneously performs data fusion, and transmits it to the data monitoring center in real time. By partially unloading, part of the data is unloaded to the edge server to improve processing efficiency. The multi-agent game reinforcement learning algorithm is used to optimize system performance.

[0006] The embodiment of the present invention provides the following technical solutions:

[0007] Step A, fusion of multi-sensor perception data of the optical cable status monitoring system;

[0008] Step B, modeling the communication process of perceptual data transmission;

[0009] Step C, calculating the probability of safe interruption of communication;

[0010] Step D: The data monitoring center processes the data and calculates the delay by partial offloading;

[0011] Step E, optimizing system parameters using multi-agent game reinforcement learning algorithm;

[0012] Wherein, step A specifically includes:

[0013] A1, by deploying multiple sensors at the sensing terminal of the optical cable: ice sensor, temperature and humidity sensor, tilt sensor, voltage sensor to respectively sense the ice condition of the optical cable, the temperature and humidity of the environment, the tilt angle of the optical cable, and the voltage data.

[0014] A2, construct the sensing echo signal received by the sensor as:

[0015]

[0016] Where η is the path loss factor, are the guidance vectors for the sensor to transmit and receive sensing signals, To perceive the azimuth and elevation of the target relative to the sensor, s(t) is the sensing signal transmitted in time slot t, ω s is the transmit beamforming vector, n s ~CN(0,σ s ) is the variance σ s Additive white Gaussian noise.

[0017] A3, forming the data set S(x1,...,x i ,y1,...,y j ,z1,...z m ,g1,...,g n ), where x i ,y j 、z m , g n They respectively represent the data monitored by the i-th ice sensor, the j-th temperature and humidity sensor, the m-th tilt sensor, and the n-th voltage sensor.

[0018] A4, calculate the probability of perception error:

[0019]

[0020] The set t corresponds to the set δ one by one, that is, when t takes x, δ takes i, and so on.

[0021] A5, define the fusion calculation method of multi-sensor data, and assume that the trust function of the data is:

[0022]

[0023] Combination rules for multiple data: Therefore, a new data set is constructed: S * (X,Y,Z,G), where X,Y,Z,G correspond to the synthesized ice cover, temperature and humidity, tilt angle, and voltage data respectively.

[0024] Wherein, step B specifically includes:

[0025] B1, the communication signal received by the monitoring center is: y c (t) = ω c H c x c (t)+n c (t), where x c It is sent by the sensor. * The communication signal composed of (X,Y,Z,G), ω c The monitoring control center receives the beamforming vector, H c Channel matrix for signal transmission to the monitoring and control center, The mean is 0 and the variance is Gaussian white noise.

[0026] B2, considering the security of the system, it is assumed that there is a passive eavesdropper (Eve) eavesdropping on the communication signal, so the eavesdropped communication signal is expressed as: y e (t) = H e x c (t)+n e (t), H e Channel matrix for signal transmission to Eve, The mean is 0 and the variance is Gaussian white noise.

[0027] B3, Assumption It follows the Rice distribution, They are line-of-sight transmission channel and non-line-of-sight transmission channel, α is the Rice factor, and v is the channel H c The path loss coefficient of channel H e Assume that the monitoring center cannot obtain accurate channel state information and the eavesdropper’s location is unknown, that is, it is impossible to obtain H c , H e Accurate channel state information, so the line-of-sight transmission channel is modeled as:

[0028]

[0029] are the estimated channel state information, ε c , ε e are the estimated error values ​​respectively.

[0030] Wherein, step C specifically includes:

[0031] C1, the instantaneous signal-to-noise ratios of the communication signals received by the data monitoring center and the eavesdropper are:

[0032]

[0033] C2, calculate the secure transmission rate of the communication, which is the data center transmission rate minus the eavesdropper's transmission rate:

[0034]

[0035] where [z] + represents max{0,z}.

[0036] C3, calculate the interruption probability of secure transmission of communication: P sop =p r (R s <R min ), according to the channel obeying the Rice distribution, and the Rice fading is a special case of the Nakagami fading channel, so γ c The cumulative distribution function of is:

[0037]

[0038] in is the Laguerre polynomial.

[0039] Same reason e The cumulative distribution function of is:

[0040] in

[0041] C4, calculate γ c , γ e The probability distribution function of is:

[0042] C5, the closed expression of the safety interruption probability is:

[0043] Wherein, step D specifically comprises:

[0044] D1, starting at time slot t, the data monitoring center generates a computing task of Ξ(t) = {L(t), F(t), T(t)}, where L(t) is the amount of data to be calculated, F(t) is the number of CPU cycles required to process a unit of data, and T(t) is the total delay for processing data.

[0045] D2 uses a partial offloading strategy to offload ηL(t) data to the edge server, and (1-η)L(t) is calculated locally. η is the offloading scale factor. Therefore, the total delay required for the task with a data volume of L(t) can be calculated as:

[0046]

[0047] g1(t) and g2(t) are the computing rates of the edge server and the local respectively.

[0048] Wherein, step E specifically includes:

[0049] E1, in order to ensure the safety, accuracy and timeliness of optical cable status monitoring, the multi-agent game reinforcement learning algorithm is used to achieve the normalization of the system's safe interruption probability, calculation delay, and minimization of the weighted sum of the perception error probability:

[0050]

[0051] Among them, λ1, λ2, λ3 are weight factors and satisfy λ1+λ2+λ3=1, is the normalized safety interruption probability, the calculated total delay, and the perception error probability.

[0052] E2, abstract the problem into a Markov game process and express it as a tuple:

[0053] Where n is the number of agents, S = {P sop ,T(t),P t} is the state space, A i (i=1,..,n)={ω s ,ω c ,η} is the action space of agent i, P(S,A i , S') is the transition probability of agent i executing action A in state S and transferring to the next state S', is the discount factor, R i is the reward function of agent i.

[0054] E3, Definition The execution strategies of all agents in state S are The discounted benefit of agent i at this time.

[0055] E4, define the reward function for agent i to perform action A in state S as:

[0056]

[0057] E6, define the state-action value function as:

[0058] E7, update makes Q i (s,a) achieves the maximum strategy, namely:

[0059]

[0060] E8, enter the next iteration and output the optimal value that meets the conditions.

[0061] Compared with the prior art, the above technical solution has the following advantages:

[0062] The present invention uses a variety of sensors to collect optical cable status information, and considers using the integrated technology of synaesthesia and calculation to improve the real-time synchronization of data collection, transmission and calculation, in which a multi-agent game reinforcement learning algorithm is used for calculation. Specifically, the system performance is improved in an all-round way by continuously optimizing the probability of perception error in the perception process, the probability of safe interruption in the communication process, and the processing delay of calculation. BRIEF DESCRIPTION OF THE DRAWINGS

[0063] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0064] Figure 1 A schematic flow chart of a distributed multi-sensor optical cable status monitoring method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0065] As described in the background technology section, how to ensure the normal operation of the power system while comprehensively improving the overall performance of communication perception computing in the process of optical cable status monitoring is an urgent problem to be solved by people in this field.

[0066] The core idea of ​​the present invention is to analyze and calculate the real-time data of the optical cable status through the integrated function of synaesthesia and resource sharing in the monitoring of the optical cable status, wherein the multi-agent game reinforcement learning algorithm is used to optimize the system performance.

[0067] See also Figure 1The embodiment of the present invention provides a distributed multi-point sensing optical cable status monitoring method, the method comprising:

[0068] Step A, fusion of multi-sensor perception data of the optical cable status monitoring system;

[0069] Step B, modeling the communication process of perceptual data transmission;

[0070] Step C, calculating the probability of safe interruption of communication;

[0071] Step D: The data monitoring center processes the data and calculates the delay by partial offloading;

[0072] Step E, optimizing system parameters using multi-agent game reinforcement learning algorithm;

[0073] Wherein, step A specifically includes:

[0074] A1, by deploying multiple sensors at the sensing terminal of the optical cable: ice sensor, temperature and humidity sensor, tilt sensor, voltage sensor to respectively sense the ice condition of the optical cable, the temperature and humidity of the environment, the tilt angle of the optical cable, and the voltage data.

[0075] A2, construct the sensing echo signal received by the sensor as:

[0076]

[0077] Where η is the path loss factor, are the guidance vectors for the sensor to transmit and receive sensing signals, θ, To perceive the azimuth and elevation of the target relative to the sensor, s(t) is the sensing signal transmitted in time slot t, ω s is the transmit beamforming vector, n s ~CN(0,σ s ) is the variance σ s Additive white Gaussian noise.

[0078] A3, forming the data set S(x1,...,x i ,y1,...,y j ,z1,...z m ,g1,...,g n ), where x i ,y j 、z m , g n They respectively represent the data monitored by the i-th ice sensor, the j-th temperature and humidity sensor, the m-th tilt sensor, and the n-th voltage sensor.

[0079] A4, calculate the probability of perception error:

[0080]

[0081] The set t corresponds to the set δ one by one, that is, when t takes x, δ takes i, and so on.

[0082] A5, define the fusion calculation method of multi-sensor data, and assume that the trust function of the data is:

[0083]

[0084] Combination rules for multiple data:

[0085] Therefore, a new data set is constructed: S * (X,Y,Z,G), where X,Y,Z,G correspond to the synthesized ice cover, temperature and humidity, tilt angle, and voltage data respectively.

[0086] Wherein, step B specifically includes:

[0087] B1, the communication signal received by the monitoring center is: y c (t) = ω c H c x c (t)+n c (t), where x c It is sent by the sensor. * The communication signal composed of (X,Y,Z,G), ω c The monitoring control center receives the beamforming vector, H c Channel matrix for signal transmission to the monitoring and control center, The mean is 0 and the variance is Gaussian white noise.

[0088] B2, considering the security of the system, it is assumed that there is a passive eavesdropper (Eve) eavesdropping on the communication signal, so the eavesdropped communication signal is expressed as: y e (t) = H e x c (t)+n e (t), H e Channel matrix for signal transmission to Eve, The mean is 0 and the variance is Gaussian white noise.

[0089] B3, Assumption It follows the Rice distribution, They are line-of-sight transmission channel and non-line-of-sight transmission channel, α is the Rice factor, and v is the channel H c The path loss coefficient of channel H eAssume that the monitoring center cannot obtain accurate channel state information and the eavesdropper’s location is unknown, that is, it is impossible to obtain H c , H e Accurate channel state information, so the line-of-sight transmission channel is modeled as:

[0090]

[0091] are the estimated channel state information, ε c , ε e are the estimated error values ​​respectively.

[0092] Wherein, step C specifically includes:

[0093] C1, the instantaneous signal-to-noise ratios of the communication signals received by the data monitoring center and the eavesdropper are:

[0094]

[0095] C2, calculate the secure transmission rate of the communication, which is the data center transmission rate minus the eavesdropper's transmission rate:

[0096]

[0097] where [z] + represents max{0,z}.

[0098] C3, calculate the interruption probability of secure transmission of communication: P sop =p r (R s <R min ), according to the channel obeying the Rice distribution, and the Rice fading is a special case of the Nakagami fading channel, so γ c The cumulative distribution function of is:

[0099]

[0100] in is the Laguerre polynomial.

[0101] Same reason e The cumulative distribution function of is:

[0102] in

[0103] C4, calculate γ c , γ e The probability distribution function of is:

[0104] C5, the closed expression of the safety interruption probability is:

[0105] Wherein, step D specifically comprises:

[0106] D1, starting at time slot t, the data monitoring center generates a computing task of Ξ(t) = {L(t), F(t), T(t)}, where L(t) is the amount of data to be calculated, F(t) is the number of CPU cycles required to process a unit of data, and T(t) is the total delay for processing data.

[0107] D2 uses a partial offloading strategy to offload ηL(t) data to the edge server, and (1-η)L(t) is calculated locally. η is the offloading scale factor. Therefore, the total delay required for the task with a data volume of L(t) can be calculated as:

[0108]

[0109] g1(t) and g2(t) are the computing rates of the edge server and the local respectively.

[0110] Wherein, step E specifically includes:

[0111] E1, in order to ensure the safety, accuracy and timeliness of optical cable status monitoring, the multi-agent game reinforcement learning algorithm is used to achieve the normalization of the system's safe interruption probability, calculation delay, and minimization of the weighted sum of the perception error probability:

[0112]

[0113] Among them, λ1, λ2, λ3 are weight factors and satisfy λ1+λ2+λ3=1, is the normalized safety interruption probability, the calculated total delay, and the perception error probability.

[0114] E2, abstract the problem into a Markov game process and express it as a tuple:

[0115] Where n is the number of agents, S = {P sop ,T(t),P t} is the state space, A i (i=1,..,n)={ω s ,ω c ,η} is the action space of agent i, P(S,A i , S') is the transition probability of agent i executing action A in state S and transferring to the next state S', is the discount factor, R i is the reward function of agent i.

[0116] E3, Definition The execution strategies of all agents in state S are The discounted benefit of agent i at this time.

[0117] E4, define the reward function for agent i to perform action A in state S as:

[0118]

[0119] E6, define the state-action value function as:

[0120] E7, update makes Q i (s,a) achieves the maximum strategy, namely:

[0121]

[0122] E8, enter the next iteration and output the optimal value that meets the conditions.

[0123] Compared with the prior art, the above technical solution has the following advantages:

[0124] The present invention uses a variety of sensors to collect the status information of the optical cable and integrates multiple data, which greatly improves the comprehensiveness and efficiency of the collection of the optical cable status information. The perceived data is transmitted to the data monitoring center in real time and unloaded to the edge server by partial unloading, thereby greatly improving the efficiency of calculation and realizing the mutual integration of synaesthesia and calculation. The multi-agent game reinforcement learning algorithm is used to continuously optimize the performance in the perception, communication and calculation processes.

[0125] The above are only embodiments of the present application and are not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included within the scope of the claims of the present application.

Claims

1. A distributed multi-sensor optical cable status monitoring method, characterized in that: The steps include: Step A, fusion of multi-sensor perception data of the optical cable status monitoring system; Step B, modeling the communication process of perceptual data transmission; Step C, calculating the probability of safe interruption of communication; Step D: The data monitoring center processes the data and calculates the delay by partial offloading; Step E, optimizing system parameters using multi-agent game reinforcement learning algorithm; Wherein, step A specifically includes: A1, by deploying multiple sensors at the sensing terminal of the optical cable: ice sensor, temperature and humidity sensor, tilt sensor, voltage sensor to sense the ice condition of the optical cable, the temperature and humidity of the environment, the tilt angle of the optical cable, and the voltage data; A2, construct the sensing echo signal received by the sensor as: Where η is the path loss factor, are the guidance vectors for the sensor to transmit and receive sensing signals, θ, To perceive the azimuth and elevation of the target relative to the sensor, s(t) is the sensing signal transmitted in time slot t, ω s is the transmit beamforming vector, n s ~CN(0,σ s ) is the variance σ s Additive Gaussian white noise; A3, forming the data set S(x1,...,x i ,y1,...,y j ,z1,...z m ,g1,...,g n ), where x i ,y j 、z m , g n Respectively represent the data monitored by the i-th ice sensor, the j-th temperature and humidity sensor, the m-th tilt sensor, and the n-th voltage sensor; A4, calculate the probability of perception error: The set t corresponds to the set δ one by one, that is, when t takes x, δ takes i, and so on; A5, define the fusion calculation method of multi-sensor data, and assume that the trust function of the data is: Combination rules for multiple data: Therefore, a new data set is constructed: S * (X, Y, Z, G), where X, Y, Z, and G correspond to the synthesized ice cover, temperature and humidity, inclination angle, and voltage data respectively; Wherein, step B specifically includes: B1, the communication signal received by the monitoring center is: y c (t) = ω c H c x c (t)+n c (t), where x c It is sent by the sensor. * The communication signal composed of (X,Y,Z,G), ω c The monitoring control center receives the beamforming vector, H c Channel matrix for signal transmission to the monitoring and control center, The mean is 0 and the variance is Gaussian white noise; B2, considering the security of the system, it is assumed that there is a passive eavesdropper (Eve) eavesdropping on the communication signal, so the eavesdropped communication signal is expressed as: y e (t) = H e x c (t)+n e (t), H e Channel matrix for signal transmission to Eve, The mean is 0 and the variance is Gaussian white noise; B3, Assumption It follows the Rice distribution, They are line-of-sight transmission channel and non-line-of-sight transmission channel, α is the Rice factor, and v is the channel H c The path loss coefficient of channel H e Path loss coefficient; Assume that the monitoring center cannot obtain accurate channel state information and the eavesdropper’s location is unknown, that is, it is impossible to obtain H c , H e Accurate channel state information, so the line-of-sight transmission channel is modeled as: are the estimated channel state information, ε c , ε e are the estimated error values ​​respectively; Wherein, step C specifically includes: C1, the instantaneous signal-to-noise ratios of the communication signals received by the data monitoring center and the eavesdropper are: C2, calculate the secure transmission rate of the communication, which is the data center transmission rate minus the eavesdropper's transmission rate: where [z] + represents max{0,z}; C3, calculate the interruption probability of secure transmission of communication: P sop =p r (R s <R min ), according to the channel obeying the Rice distribution, and the Rice fading is a special case of the Nakagami fading channel, so γ c The cumulative distribution function of is: in is the Laguerre polynomial; Same reason e The cumulative distribution function of is: in C4, calculate γ c , γ e The probability distribution function of is: C5, the closed expression of the safety interruption probability is: Wherein, step D specifically comprises: D1, starting at time slot t, the data monitoring center generates a computing task of Ξ(t) = {L(t), F(t), T(t)}, where L(t) is the amount of data to be calculated, F(t) is the number of CPU cycles required to process a unit of data, and T(t) is the total delay for processing data; D2 uses a partial offloading strategy to offload ηL(t) data to the edge server, and (1-η)L(t) is calculated locally. η is the offloading scale factor. Therefore, the total delay required for the task with a data volume of L(t) can be calculated as: g1(t) and g2(t) are the computing rates of the edge server and the local server, respectively; Wherein, step E specifically includes: E1, in order to ensure the safety, accuracy and timeliness of optical cable status monitoring, the multi-agent game reinforcement learning algorithm is used to achieve the normalization of the system's safe interruption probability, calculation delay, and minimization of the weighted sum of the perception error probability: Among them, λ1, λ2, λ3 are weight factors and satisfy λ1+λ2+λ3=1, is the normalized safety interruption probability, the total calculated delay, and the perception error probability; E2, abstract the problem into a Markov game process and express it as a tuple: Where n is the number of agents, S = {P sop ,T(t),P t } is the state space, A i (i=1,..,n)={ω s ,ω c ,η} is the action space of agent i, P(S,A i , S') is the transition probability of agent i executing action A in state S and transferring to the next state S', is the discount factor, R i is the reward function of agent i; E3, Definition The execution strategies of all agents in state S are The discounted benefit of agent i at time E4, define the reward function for agent i to perform action A in state S as: E6, define the state-action value function as: E7, update makes Q i (s,a) achieves the maximum strategy, namely: E8, enter the next iteration and output the optimal value that meets the conditions.

Images