Intelligent Fusion Terminal Based on Dynamic Weighting of Multiple Clock Sources and Time Synchronization Method

Abstract

The present invention discloses an intelligent fusion terminal and a time synchronization method based on dynamic weighting of multiple clock sources, including: First, the intelligent fusion terminal establishes a time synchronization model based on multiple clock sources; Second, the intelligent fusion terminal obtains synchronization signals from multiple clock sources such as Beidou satellites, GPS, and ground stations by optimizing the clock source selection strategy for large time scales, and then realizes low-latency and high-precision time synchronization by optimizing the clock source weight allocation strategy for small time scales; Finally, the intelligent fusion terminal uses the time service module and the communication module to provide time service for unsynchronized power equipment. The advantages of the present invention are: realizing low-latency and high-precision time synchronization of power equipment in the distribution system, reducing the time synchronization error through weighted summation, and improving the time synchronization accuracy.

Description

Technical Field

[0001] The present invention relates to the technical field of distribution system time synchronization, and particularly to an intelligent fusion terminal and a time synchronization method based on dynamic weighting of multiple clock sources. Background Art

[0002] With the large-scale access of renewable energy such as distributed photovoltaic and distributed wind power to the distribution system, the distribution system requires more precise energy management to ensure the balance between energy supply and demand. Among them, high-precision time synchronization of power equipment is the basis for realizing precise energy management. In the traditional distribution system time synchronization method, the intelligent fusion terminal obtains the time synchronization signal from the satellite or the ground station to achieve time synchronization between intelligent fusion terminals. On this basis, each intelligent fusion terminal sends time synchronization data to the power equipment deployed on a large scale within its coverage area. However, due to the complex electromagnetic environment of the distribution system and serious channel attenuation, the synchronization delay of the time synchronization method based on a single clock source is large, and the reliability of time synchronization cannot be guaranteed. In addition, the clock drift of the single clock source also increases the synchronization error and affects the time synchronization accuracy. Therefore, it is necessary to study a low-delay and high-precision time synchronization method based on multiple clock sources.

[0003] In the high-precision time synchronization method based on multiple clock sources, the intelligent fusion terminal deployed in the distribution system obtains the synchronization signal from multiple clock sources such as Beidou satellite, Global Positioning System (GPS), and ground station, and then performs time synchronization on the power equipment. Compared with the time synchronization method based on a single clock source, the time synchronization method based on multiple clock sources has the following advantages. On the one hand, when a certain clock source fails to work properly, or the channel quality between the intelligent fusion terminal and the clock source is poor, the terminal can obtain the time synchronization signal from other clock sources to ensure the reliability of time synchronization. On the other hand, by weighted summation of the time signals from multiple clock sources, the synchronization error caused by the frequency drift of the clock source can be greatly reduced, and high-precision time synchronization can be achieved. However, the high-precision time synchronization technology for distribution systems based on multiple clock sources still faces many challenges.

[0004] First, the time synchronization delay and the time synchronization error are coupled with each other. The channel condition between the clock source with a small time synchronization error and the intelligent fusion terminal may be poor. Selecting the clock source with a small time synchronization error will increase the time synchronization delay, and vice versa. Therefore, it is necessary to jointly consider the minimization problem of time synchronization error and time synchronization delay.

[0005] Second, the clock source selection strategy and the clock source weight allocation strategy need to be optimized on different time scales. Since the communication overhead for the intelligent fusion terminal to request time synchronization information from the clock source is relatively large, the selection of the clock source should not be switched frequently and should be optimized on a large time scale. However, the frequency drift of the clock source changes rapidly, and the clock source weight allocation strategy needs to be optimized on a small time scale to adapt to the dynamic changes of the synchronization error.

[0006] Prior Art I

[0007] Chinese Invention Patent: A multi-clock source cooperative time synchronization system and its time synchronization method, with the application number 202111272124.1 and the application date February 11, 2022;

[0008] First, the remote host coordinates multi-clock source information to allocate clock sources for in-station power equipment, and then the clock sources are used to provide a unified time signal for in-station equipment. Prior Art I cannot solve the coupling problem between time synchronization delay and time synchronization error, and cannot reduce the communication overhead caused by frequent switching of the clock source, resulting in the difficulty for the selected clock source to simultaneously meet the low-latency and high-precision time synchronization requirements of the intelligent fusion terminal.

[0009] Prior Art II

[0010] Chinese Invention Patent: A multi-clock source highly reliable time synchronization method, with the application number 202110674315.4 and the application date June 17, 2021;

[0011] By calculating the time difference between the previous and the current time synchronization services to determine whether the current clock source is trustworthy, and screening out the qualified clock sources. Prior Art II cannot reduce the time synchronization error through weighted summation of multiple clock sources to meet the high-precision time synchronization requirements of power equipment in the distribution system. In addition, this technology cannot solve the coupling problem between time synchronization error and time synchronization delay, and cannot optimize the clock source selection strategy and the clock source weight allocation strategy from different scales, making it difficult to achieve low-latency and high-precision time synchronization of the intelligent fusion terminal. Summary of the Invention

[0012] In view of the deficiencies of the prior art, the present invention provides an intelligent fusion terminal and a time synchronization method based on dynamic weighting of multi-clock sources.

[0013] To achieve the above invention objectives, the technical solutions adopted by the present invention are as follows:

[0014] An intelligent fusion terminal based on dynamic weighting of multi-clock sources includes: a power supply module, a communication module, a storage module, a multi-clock source dynamic weighting time synchronization module, a processor module, and a timing module.

[0015] The power supply module is responsible for providing continuous and stable power supply inside the intelligent fusion terminal to support the continuous operation of the intelligent fusion terminal.

[0016] The communication module supports multiple communication methods including satellite communication, 5G, optical fiber, AC / DC power line carrier, and industrial Ethernet. On the one hand, the communication module obtains time synchronization information of Beidou and GPS satellites through satellite communication, and obtains time synchronization information of ground stations through 5G and optical fiber communication; on the other hand, the communication module uses AC / DC power line carrier and industrial Ethernet to time the power terminal. This module can support multiple communication methods simultaneously to avoid abnormal information transmission caused by link interruption.

[0017] The storage module stores the code related to the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy, stores the time synchronization information of satellites and ground stations, and the data required to execute the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy.

[0018] The multi-clock source dynamic weighted time synchronization module optimizes the clock source selection strategy for large time scales and the clock source weight allocation strategy for small time scales by using the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy according to the time synchronization information obtained by the communication module and the data stored in the storage module, so as to achieve low-latency and high-precision time synchronization.

[0019] The processor module provides computing services for the operation of the optimization algorithm.

[0020] The time synchronization module responds to and manages the time synchronization requests of power equipment, and based on the high-precision time synchronization information obtained by the multi-clock source dynamic weighted time synchronization module, it times the un-timed power equipment through the communication module.

[0021] The present invention also discloses a time synchronization method based on multi-clock source dynamic weighting, including the following steps:

[0022] S1: The intelligent fusion terminal receives the time synchronization request of the power terminal with large-scale deployment and is ready to receive the time synchronization information of the external clock source;

[0023] S2: The intelligent fusion terminal confirms the currently available clock sources;

[0024] S3: The intelligent fusion terminal extracts the data information in the storage module and establishes a joint optimization model of the clock source selection strategy for large time scales and the weight allocation strategy for small time scales;

[0025] S4: Based on the strategy and the joint optimization model of the weight allocation strategy for small time scales, the intelligent fusion terminal runs the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy;

[0026] S5: The intelligent fusion terminal completes time synchronization with the clock source and uses the time service module to provide time service to unsynchronized power terminals.

[0027] Further, the step S3 includes:

[0028] Suppose there are N external clock sources, 1 local clock source, 1 intelligent fusion terminal, and multiple unsynchronized power terminal devices. Let the set of N external clock sources be represented as C = {C1, C2,..., C n ,..., C N}, which includes N1 Beidou satellites, and the corresponding clock sources are represented as C n , n = 1,..., N1; N2 GPS satellites, and the corresponding clock sources are represented as C n , n = N1 + 1,... N1 + N2; N3 ground stations, and the corresponding clock sources are represented as C n , n = N1 + N2 + 1,..., N. In addition, N = N1 + N2 + N3.

[0029] Divide the total time into I time periods, and its set is represented as I = {1, 2,..., i,..., I}. Each time period is divided into T time slots, and its set is represented as T = {1, 2,..., t,..., T}. On a large time scale, the intelligent fusion terminal selects M clock sources and receives time synchronization signals from the selected clock sources. On a small time scale, the intelligent fusion terminal assigns weights to the selected M clock sources to optimize the time synchronization accuracy. After the multi-time scale clock source selection and weight assignment strategy optimization are completed, the intelligent fusion terminal provides time service to unsynchronized power equipment, thereby achieving low-latency and high-precision time synchronization for large-scale deployed power equipment.

[0030] Further, the specific sub-steps of the step S3 are as follows:

[0031] S3.1 Create a time synchronization delay model

[0032] Define the clock source selection strategy in the i-th time period as x n (i) ∈ {0, 1}, where x n (i) = 1 indicates that the intelligent fusion terminal selects the clock source C n , otherwise x n (i) = 0. In each time period, the intelligent fusion terminal selects M clock sources, that is In the t-th time slot, the intelligent fusion terminal assigns weights to the selected M clock sources. The clock source weight set is represented as And it is stipulated that

[0033] For the clock sources of Beidou satellites and GPS satellites, the intelligent fusion terminal and the clock source C nThe signal-to-interference-plus-noise ratio between is expressed as

[0034] SINR(i,t) = P n + h n (i,t) - L D - 10lg(kTem n B n ) - 10lg(χ n (t)) (1)

[0035] Where P n is the transmission power of the time synchronization signal (dBW), h n (i,t) is the channel gain, L D is the downlink loss of the satellite, k is the Boltzmann constant, Tem n is the equivalent thermal noise base temperature, B n is the bandwidth, χ n (t) is the electromagnetic interference, expressed as

[0036]

[0037] Where γ ∈ (0, 2] is the characteristic exponent, determining the degree of pulse characteristics, κ ∈ [-1, 1] is the symmetry parameter, ζ ≥ 0 is the distribution parameter, μ ∈ R represents the location parameter. When γ ∈ (0, 1], the location parameter μ is the median of the distribution function. When γ ∈ (1, 2], the location parameter μ is the mean of the distribution function.

[0038] For the ground station clock source, the signal-to-interference-plus-noise ratio is expressed as

[0039]

[0040] Where represents the noise power.

[0041] Therefore, the transmission rate from the clock source to the intelligent fusion terminal is expressed as

[0042] R n (i,t) = B n log2(1 + SINR(i,t)) (4)

[0043] Assume that the time synchronization data packet size of the clock source C n is D n (i,t), then the transmission delay from the clock source to the intelligent fusion terminal is expressed as

[0044]

[0045] Define the time synchronization delay of the intelligent fusion terminal as the maximum transmission delay among M clock sources, that is

[0046] T(i,t) = max{x1(i)T i (i,t),...,x N (i)T N (i,t)} (6)

[0047] S3.2 Create a time synchronization error model

[0048] Define the clock source C n The time synchronization error at the t-th time slot in the i-th period is E n (i,t), and it follows a normal distribution where α n represents the mean value, represents the variance

[0049] Define the weight assigned to the clock source C by the intelligent fusion terminal within the t-th time slot in the i-th period as ω n (i,t), and ω n (i,t) ∈ Ω. Therefore, the time synchronization error of the intelligent fusion terminal is expressed as n (i,t) ∈ Ω. Therefore, the time synchronization error of the intelligent fusion terminal is expressed as

[0050]

[0051] Furthermore, the step S4 includes:

[0052] By jointly optimizing the clock source selection strategy at a large time scale and the clock source weight allocation strategy at a small time scale, minimize the weighted sum of the time synchronization delay and time synchronization error of the intelligent fusion terminal. The optimization objective is modeled as

[0053]

[0054] where V is the weight. C1 represents the value range of the clock source selection strategy; C2 represents that within each period, the intelligent fusion terminal selects M clock sources; C3 represents the value range of the clock source weight; C4 represents that within each time slot, the sum of the clock source weights is 1

[0055] To solve the problem of minimizing the weighted sum of the time synchronization delay and time synchronization error of the intelligent fusion terminal, the present invention proposes a joint optimization algorithm for multi-time scale clock source selection and weight allocation based on reinforcement learning, which first optimizes the clock source selection strategy at a large time scale and then optimizes the clock source weight allocation strategy at a small time scale

[0056] Furthermore, the optimization of the clock source selection strategy at a large time scale is specifically as follows:

[0057] Define the clock source C n The total time synchronization delay and time synchronization error within the i-th period are expressed as

[0058]

[0059] Define that within the $i$-th time period, the intelligent fusion terminal selects clock source $C$. n The obtained reward is

[0060]

[0061] Therefore, as of the $i$-th time period, the intelligent fusion terminal selects clock source $C$. n The average reward obtained is expressed as

[0062]

[0063] Assume that as of the $i$-th time period, the set of average rewards of all clock sources is expressed as Define the set Denote as $R$ L the largest $M$ average rewards in (i).

[0064] Furthermore, based on the $\epsilon$-greedy algorithm, optimize the clock source selection strategy on a large time scale, including the following steps:

[0065] Step 1: Initialization. The intelligent fusion terminal traverses all clock sources to obtain the initial reward For all clock sources $C$ n $\in C$, initialize the clock source selection strategy $x$ n (0) = 1, initialize the average reward Initialize the threshold $\epsilon\in(0,1)$.

[0066] Step 2: Policy optimization. The intelligent fusion terminal generates a random number $a\in(0,1)$ and compares the random number $a$ with the threshold $\epsilon$. If $a\leq\epsilon$, the intelligent fusion terminal randomly selects $M$ clock sources; otherwise, the intelligent fusion terminal selects the clock sources with the largest $M$ average rewards, that is

[0067]

[0068] Step 3: Iterative update. The intelligent fusion terminal executes the optimal strategy obtained by the above policy optimization method, calculates the reward according to the execution result according to formula (11) and updates the average reward according to formula (12).

[0069] Furthermore, the optimization of the clock source weight allocation strategy on a small time scale is specifically as follows:

[0070] There are a total of $K = M!$ weight allocation methods based on the set $\Omega$, and define the $k$-th weight allocation method as $A$ k , and the set of weight allocation methods is $A=\{A_1,A_2,\cdots,A$ k ,\cdots,A$ K}. For ease of description, define the weight allocation selection variable z k (i,t) ∈ {0,1}, z k (i,t) = 1 indicates that the intelligent fusion terminal selects weight allocation method A k for time synchronization, otherwise z k (i,t) = 0.

[0071] Define that the intelligent fusion terminal selects weight allocation method A k The obtained reward is

[0072]

[0073] Furthermore, the exploration and exploitation exponential weight algorithm is adopted to optimize the clock source weight allocation strategy on a small time scale, including the following steps:

[0074] Step 1: Initialization. Initialize the uniform distribution parameter ξ ∈ (0,1], and initialize the empirical distribution parameter λ k (i,1) = 1,

[0075] Step 2: Policy optimization. First, calculate the probability of selecting weight allocation method A k which is expressed as

[0076]

[0077] Then calculate the cumulative probability, which is expressed as

[0078]

[0079] Finally, generate a random number p0(i,t) ∈ [0,1], and optimize the clock source weight allocation strategy according to the relationship between the random number p0 and the cumulative probability F i,t which is expressed as

[0080]

[0081] Step 3: Iterative update. The intelligent fusion terminal performs time synchronization according to the clock source weight allocation strategy, calculates the reward according to formula (14), and updates the empirical distribution parameter, which is expressed as

[0082]

[0083] where γ is the weight adjustment factor, is the reward evaluation value, which is expressed as

[0084]

[0085] Compared with the prior art, the advantages of the present invention are as follows:

[0086] 1. The present invention comprehensively considers the impact of time synchronization delay and time synchronization error on the time synchronization of the distribution system, and reduces the weighted sum of time synchronization delay and time synchronization error by jointly optimizing the clock source selection strategy and the clock source weight allocation strategy, avoiding high synchronization delay caused by poor channel conditions between the selected clock source and the intelligent fusion terminal and low synchronization accuracy caused by severe frequency drift of the selected clock source, so as to achieve low-delay and high-precision time synchronization of power equipment in the distribution system.

[0087] 2. On the one hand, the present invention iteratively optimizes the clock source selection strategy on a large time scale, enabling the intelligent fusion terminal to learn the channel states and frequency drift conditions of multiple clock sources, and using the threshold ε to balance the exploration and utilization of multiple clock sources; on the other hand, by iteratively optimizing the clock source weight allocation strategy on a small time scale, the intelligent fusion terminal continuously learns the time synchronization error information of multiple clock sources in a high-dynamic environment, and reduces the time synchronization error through weighted summation to improve the time synchronization accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 is a schematic structural diagram of an intelligent fusion terminal based on dynamic weighting of multiple clock sources according to an embodiment of the present invention;

[0089] Figure 2 is a schematic structural diagram of a multi-clock source time synchronization scenario of a distribution system according to an embodiment of the present invention;

[0090] Figure 3 is a flow chart of a joint optimization algorithm for multi-time scale clock source selection and weight allocation based on reinforcement learning according to an embodiment of the present invention;

[0091] Figure 4 is a schematic diagram of the weighted sum of time synchronization delay and time synchronization error according to an embodiment of the present invention;

[0092] Figure 5 is a schematic diagram of the weighted sum of time synchronization delay and time synchronization error after a certain satellite becomes unavailable according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0093] In order to make the objectives, technical solutions and advantages of the present invention more clearly understood, the following further describes the present invention in detail with reference to the drawings and by way of examples.

[0094] As Figure 1 shown, an intelligent fusion terminal based on dynamic weighting of multiple clock sources includes a power supply module, a communication module, a storage module, a multi-clock source dynamic weighting time synchronization module, a processor module, and a timing module in the terminal.

[0095] The power supply module is connected to other modules and is responsible for providing continuous and stable power supply inside the terminal to support the continuous operation of the intelligent fusion terminal.

[0096] The communication module supports multiple communication methods such as satellite communication, 5G, optical fiber, AC / DC power line carrier, and industrial Ethernet. On the one hand, the communication module obtains the time synchronization information of Beidou and GPS satellites through satellite communication, and obtains the time synchronization information of the ground station through 5G and optical fiber communication; on the other hand, the communication module uses AC / DC power line carrier and industrial Ethernet to time the power terminal. This module can support multiple communication methods simultaneously to avoid abnormal information transmission caused by link interruption.

[0097] The storage module stores the code related to the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy, stores the time synchronization information of satellites and ground stations, and the data required to execute the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy.

[0098] The multi-clock source dynamic weighted time synchronization module optimizes the clock source selection strategy for large time scales and the clock source weight allocation strategy for small time scales by using the joint optimization algorithm of multi-time scale clock source selection and weight allocation strategy based on the time synchronization information obtained by the communication module and the data stored in the storage module, so as to achieve low-latency and high-precision time synchronization.

[0099] The processor module provides computing services for the operation of the optimization algorithm.

[0100] The timing module responds to and manages the time synchronization requests of power equipment, and based on the high-precision time synchronization information obtained by the multi-clock source dynamic weighted time synchronization module, it times the un-timed power equipment through the communication module.

[0101] The present invention proposes a joint optimization algorithm for multi-time scale clock source selection and weight allocation strategy based on reinforcement learning for the multi-clock source time synchronization scenario of the distribution system. First, the intelligent fusion terminal establishes a time synchronization model based on multiple clock sources; second, the intelligent fusion terminal obtains synchronization signals from multiple clock sources such as Beidou satellites, GPS, and ground stations by optimizing the clock source selection strategy for large time scales, and then realizes low-latency and high-precision time synchronization by optimizing the clock source weight allocation strategy for small time scales; finally, the intelligent fusion terminal times the unsynchronized power equipment through the timing module and the communication module.

[0102] The specific implementation steps of a time synchronization method based on multi-clock source dynamic weighting are as follows:

[0103] S1: The intelligent fusion terminal receives the time synchronization request of the power terminals deployed on a large scale and is ready to receive the time synchronization information of external clock sources;

[0104] S2: The intelligent fusion terminal confirms the currently available clock sources;

[0105] S3: The intelligent fusion terminal extracts the data information in the storage module and establishes a joint optimization model for the large-time-scale clock source selection strategy and the small-time-scale weight allocation strategy;

[0106] S4: Based on the above model, the intelligent fusion terminal runs a joint optimization algorithm for multi-time-scale clock source selection and weight allocation strategy;

[0107] S5: The intelligent fusion terminal completes the time synchronization with the clock source and uses the time service module to time service the unsynchronized power terminals.

[0108] In a further embodiment, step S3 includes:

[0109] As Figure 2 shown, assume that there are N external clock sources, 1 local clock source, 1 intelligent fusion terminal and multiple unsynchronized power terminal devices in the system. Assume that the set of N external clock sources is represented as C = {C1, C2,..., C n ,..., C N}, which includes N1 Beidou satellites, and the corresponding clock sources are represented as C n , n = 1,..., N1; N2 GPS satellites, and the corresponding clock sources are represented as C n , n = N1 + 1,... N1 + N2; N3 ground stations, and the corresponding clock sources are represented as C n , n = N1 + N2 + 1,..., N. In addition, N = N1 + N2 + N3.

[0110] The embodiment of the present invention considers a multi-time-scale model, divides the total time into I time periods, and its set is represented as I = {1, 2,..., i,..., I}. Each time period is divided into T time slots, and its set is represented as T = {1, 2,..., t,..., T}. On the large time scale, the intelligent fusion terminal selects M clock sources and receives time synchronization signals from the selected clock sources. On the small time scale, the intelligent fusion terminal assigns weights to the selected M clock sources to optimize the time synchronization accuracy. After the optimization of the multi-time-scale clock source selection and weight allocation strategy is completed, the intelligent fusion terminal performs time service for the unsynchronized power equipment, so as to achieve low-delay and high-precision time synchronization of large-scale deployed power equipment.

[0111] S3.1 Time synchronization delay model

[0112] The embodiment of the present invention defines the clock source selection strategy in the i-th time period as x n (i) ∈ {z, 1}, where x n (i) = 1 means that the intelligent fusion terminal selects the clock source C n , otherwise x n(i) = 0. In each time period, the intelligent fusion terminal selects M clock sources, that is In the t-th time slot, the intelligent fusion terminal assigns weights to the selected M clock sources. The clock source weight set is denoted as And it is stipulated that

[0113] For the clock sources of Beidou satellites and GPS satellites, the signal-to-interference-plus-noise ratio (SINR) between the intelligent fusion terminal and the clock source C n is denoted as

[0114] SINR(i,t) = P n + h n (i,t) - L D - 10lg(kTem n B n ) - 10lg(χ n (t)) (1)

[0115] where, P n is the transmission power of the time synchronization signal (dBW), h n (i,t) is the channel gain, L D is the downlink loss of the satellite, k is the Boltzmann constant, Tem n is the equivalent thermal noise base temperature, B n is the bandwidth, χ n (t) is the electromagnetic interference, denoted as

[0116]

[0117] where, γ ∈ (0, 2] is the characteristic exponent, which determines the degree of pulse characteristics, κ ∈ [-1, 1] is the symmetry parameter, ζ ≥ 0 is the distribution parameter, μ ∈ R represents the position parameter. When γ ∈ (0, 1], the position parameter μ is the median of the distribution function. When γ ∈ (1, 2], the position parameter μ is the mean value of the distribution function.

[0118] For the ground station clock source, the signal-to-interference-plus-noise ratio is denoted as

[0119]

[0120] where, represents the noise power.

[0121] Therefore, the transmission rate from the clock source to the intelligent fusion terminal is denoted as

[0122] R n (i,t) = B n log2(1 + SINR(i,t)) (4)

[0123] Assume the clock source C nThe size of the time synchronization data packet is D n (i, t), then the transmission delay from the clock source to the intelligent fusion terminal is expressed as

[0124]

[0125] The present invention defines the time synchronization delay of the intelligent fusion terminal as the maximum transmission delay among M clock sources, that is

[0126] T(i, t) = max{x1(i)T i (i, t),..., x N (i)T N (i, t)} (6)

[0127] S3.2 Time synchronization error model

[0128] The embodiment of the present invention defines the clock source C n The time synchronization error at the t-th time slot in the i-th period is E n (i, t), and it follows a normal distribution where α n represents the mean value, represents the variance

[0129] Define the weight assigned to the clock source C by the intelligent fusion terminal at the t-th time slot in the i-th period as ω n (i, t), and ω n (i, t) ∈ Ω. Therefore, the time synchronization error of the intelligent fusion terminal is expressed as n (i, t) ∈ Ω. Therefore, the time synchronization error of the intelligent fusion terminal is expressed as

[0130]

[0131] In a further embodiment, step S4 includes:

[0132] By jointly optimizing the clock source selection strategy at a large time scale and the clock source weight allocation strategy at a small time scale, minimize the weighted sum of the time synchronization delay and time synchronization error of the intelligent fusion terminal. The optimization objective is modeled as

[0133]

[0134] where V is the weight. C1 represents the value range of the clock source selection strategy; C2 represents that within each period, the intelligent fusion terminal selects M clock sources; C3 represents the value range of the clock source weight; C4 represents that within each time slot, the sum of the clock source weights is 1

[0135] To solve the problems of time synchronization delay and minimization of the weighted sum of time synchronization errors in the above-mentioned intelligent fusion terminal, the present invention proposes a joint optimization algorithm for multi-time scale clock source selection and weight allocation based on reinforcement learning. First, the clock source selection strategy at the large time scale is optimized, and then the clock source weight allocation strategy at the small time scale is optimized. The algorithm flow is as Figure 3 shown.

[0136] S4.1 Optimization of the clock source selection strategy at the large time scale

[0137] Define the clock source C n The total time synchronization delay and time synchronization error within the i-th time period are expressed as

[0138]

[0139] Define that within the i-th time period, the intelligent fusion terminal selects the clock source C n The obtained reward is

[0140]

[0141] Therefore, as of the i-th time period, the intelligent fusion terminal selects the clock source C n The average reward obtained is expressed as

[0142]

[0143] Assume that as of the i-th time period, the average reward set of all clock sources is expressed as Define the set Denote the largest M average rewards in R L (i).

[0144] The present invention optimizes the clock source selection strategy at the large time scale based on the ε-greedy algorithm, including three steps: initialization, strategy optimization, and iterative update. The specific introduction is as follows.

[0145] Step 1: Initialization. The intelligent fusion terminal traverses all clock sources to obtain the initial reward For all clock sources C n ∈C, initialize the clock source selection strategy x n (0)=1, initialize the average reward Initialize the threshold ε∈(0,1).

[0146] Step 2: Strategy optimization. The intelligent fusion terminal generates a random number a∈(0,1) and compares the random number a with the threshold ε. If a≤ε, the intelligent fusion terminal randomly selects M clock sources; otherwise, the intelligent fusion terminal selects the clock sources with the largest M average rewards, that is

[0147]

[0148] Step 3: Iterative update. The intelligent fusion terminal executes the optimal policy obtained by the above policy optimization method, calculates the reward according to the execution result according to formula (11), and updates the average reward according to formula (12).

[0149] S4.2 Optimization of clock source weight allocation strategy at small time scales

[0150] There are a total of K = M! weight allocation methods based on the set Ω, and the k-th weight allocation method is defined as A k , and the weight allocation method set is A = {A1, A2..., A k ,..., A K}. For ease of description, define the weight allocation selection variable z k (i,t) ∈ {0,1}, z k (i,t) = 1 means that the intelligent fusion terminal selects the weight allocation method A k for time synchronization, otherwise z k (i,t) = 0.

[0151] Define the reward obtained by the intelligent fusion terminal when selecting the weight allocation method A k as

[0152]

[0153] The present invention uses the exponential-weight algorithm for exploration and exploitation (EXP3) to optimize the clock source weight allocation strategy at small time scales, mainly including three steps: initialization, policy optimization, and iterative update:

[0154] Step 1: Initialization. Initialize the uniform distribution parameter ξ ∈ (0,1], and initialize the empirical distribution parameter λ k (i,1) = 1,

[0155] Step 2: Policy optimization. First, calculate the probability of selecting the weight allocation method A k , which is expressed as

[0156]

[0157] Then calculate the cumulative probability, which is expressed as

[0158]

[0159] Finally, generate a random number p0(i,t) ∈ [0,1], and according to the random number p0 and the cumulative probability F i,t(k) Optimize the clock source weight allocation strategy, expressed as

[0160]

[0161] Step 3: Iterative update. The intelligent fusion terminal performs time synchronization according to the clock source weight allocation strategy, calculates the reward according to formula (14), and updates the experience distribution parameters, expressed as

[0162]

[0163] where γ is the weight adjustment factor, is the reward evaluation value, expressed as

[0164]

[0165] Simulation verification

[0166] The embodiment of the present invention considers the time synchronization scenario of a power distribution system with multiple satellite and multi-ground station clock sources, including 4 Beidou satellites, 4 GPS satellites and 2 ground stations. In each time period, the intelligent fusion terminal can select 4 clock sources. Other simulation parameters are shown in Table 1. The present invention is compared with the multi-clock source time synchronization algorithm with preset priorities (PMST) to verify the performance of the proposed algorithm. In the PMST algorithm, the gateway selects the clock source according to the preset clock source priority.

[0167] Table 1 Simulation parameters

[0168]

[0169] Figure 4 Describes the variation of the weighted sum of time synchronization delay and time synchronization error with the time period. Compared with the PMST algorithm, the proposed algorithm in the present invention increases the weighted sum by 45.30%. This is because the PMST selects the clock source and allocates the clock source weight according to the preset clock source priority, and cannot dynamically select the clock source, nor can it adjust the weight when the synchronization accuracy of the clock source and the channel link quality deteriorate.

[0170] Figure 5 Describes the variation of the weighted sum of time synchronization delay and time synchronization error with the time period when a high-priority satellite with better performance suddenly becomes unavailable after the 60th time period. Compared with the PMST, the weighted sum of the proposed method is reduced by 54.78%. When the satellite suddenly becomes unavailable, the time synchronization delay and time synchronization error of both the proposed method and the PMST will increase because they must switch from a clock source with better performance to other clock sources, which results in an increase in the weighted sum. However, since the proposed method can quickly find other clock sources with good performance through exploration, the time synchronization delay and time synchronization error decrease with the increase of the time period.

[0171] Those of ordinary skill in the art will realize that the embodiments described herein are to assist the reader in understanding the implementation methods of the present invention, and it should be understood that the protection scope of the present invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific deformations and combinations that do not deviate from the essence of the present invention based on these technical revelations disclosed in the present invention, and these deformations and combinations are still within the protection scope of the present invention.

Claims

1. A time synchronization method based on dynamic weighting of multiple clock sources, characterized in that It includes the following steps: S1: The intelligent fusion terminal receives a time synchronization request from a large-scale deployed power terminal and prepares to receive time synchronization information from an external clock source. S2: The intelligent fusion terminal confirms the currently available clock sources. S3: The intelligent fusion terminal extracts data information from the storage module and establishes a joint optimization model for the large-time-scale clock source selection strategy and the small-time-scale weight allocation strategy. The specific sub-steps are as follows: S3.1 Create a time synchronization delay model Define the clock source selection strategy in the $i$-th time period as $x^{(i)} \in \{0, 1\}$, where $x^{(i)} = 1$ indicates that the intelligent fusion terminal selects clock source $C$, otherwise $x^{(i)} = 0$. In each time period, the intelligent fusion terminal selects $M$ clock sources, that is, in the $t$-th time slot, the intelligent fusion terminal assigns weights to the selected $M$ clock sources. The clock source weight set is denoted as $\mathbf{w}_t$, and it is stipulated that $\sum_{m = 1}^{M} w_{t,m} = 1$. n $(i) \in \{0, 1\}$, where, $x$ n $(i) = 1$ indicates that the intelligent fusion terminal selects clock source $C$ n , otherwise $x$ n (i) = 0; In each time period, the intelligent fusion terminal selects $M$ clock sources, that is In the $t$-th time slot, the intelligent fusion terminal assigns weights to the selected $M$ clock sources; The clock source weight set is denoted as and it is stipulated that For the clock sources of Beidou satellites and GPS satellites, the signal-to-interference-plus-noise ratio between the intelligent fusion terminal and clock source C n is expressed as SINR(i,t) = P n + h n (i,t) - L D - 10lg(kTem n B n ) - 10lg(χ n (t)) (1) where, P n is the transmission power of the time synchronization signal (dBW), h n (i, t) is the channel gain, L D is the downlink loss of the satellite, k is the Boltzmann constant, Tem n is the basic temperature of the equivalent thermal noise, B n is the bandwidth, χ n (t) is the electromagnetic interference, expressed as Among them, γ∈(0,2] is the characteristic exponent, which determines the degree of pulse characteristics, κ∈[-1,1] is the symmetry parameter, ζ≥0 is the distribution parameter, and μ∈R represents the position parameter. When γ∈(0,1], the position parameter μ is the median of the distribution function. When γ∈(1,2], the position parameter μ is the average value of the distribution function; For the ground station clock source, the signal-to-interference-plus-noise ratio is expressed as Among them, represents the noise power; Therefore, the transmission rate from the clock source to the intelligent fusion terminal is expressed as R n (i,t) = B n log2(1 + SINR(i,t)) (4) Assume clock source C n The size of the time synchronization data packet is D n (i, t), then the transmission delay from the clock source to the intelligent fusion terminal is expressed as Define the time synchronization delay of the intelligent fusion terminal as the maximum transmission delay among M clock sources, that is T(i,t) = max{x1(i)T i (i,t),...,x N (i)T N (i,t)} (6) S3.2 Create a time synchronization error model Define clock source C n The time synchronization error at the t-th time slot in the i-th period is E n (i, t), and it follows a normal distribution where α n represents the mean value, represents the variance; Define the weight assigned to clock source C by the intelligent fusion terminal in the t-th time slot of the i-th period as ω n (i, t), and ω n (i, t) ∈ Ω; therefore, the time synchronization error of the intelligent fusion terminal is expressed as n (i, t) ∈ Ω; thus, the time synchronization error of the intelligent fusion terminal is expressed as S4: Based on the joint optimization model of the large-time-scale clock source selection strategy and the small-time-scale weight allocation strategy, the intelligent fusion terminal runs a joint optimization algorithm for multi-time-scale clock source selection and weight allocation. S5: The intelligent fusion terminal completes time synchronization with the clock source and uses the time service module to time the unsynchronized power terminals.

2. The time synchronization method according to claim 1, characterized in that: The step S3 includes: Suppose there are N external clock sources, 1 local clock source, 1 intelligent fusion terminal, and multiple unsynchronized power terminal devices; the set of N external clock sources is represented as C = {C1, C2,..., C n ,..., C N}, which includes N1 Beidou satellites, and the corresponding clock sources are represented as C n , n = 1,..., N1; N2 GPS satellites, and the corresponding clock sources are represented as C n , n = N1 + 1,... N1 + N2; N3 ground stations, and the corresponding clock sources are represented as C n , n = N1 + N2 + 1,..., N; in addition, N = N1 + N2 + N3; Divide the total time into I time periods, and its set is expressed as I = {1, 2,..., i,..., I}; each time period is divided into T time slots, and its set is expressed as T = {1, 2,..., t,..., T}; at the large time scale, the intelligent fusion terminal selects M clock sources and receives time synchronization signals from the selected clock sources; at the small time scale, the intelligent fusion terminal assigns weights to the selected M clock sources to optimize the time synchronization accuracy; after the optimization of the multi-time-scale clock source selection and weight allocation strategy is completed, the intelligent fusion terminal times the unsynchronized power equipment, so as to achieve low-delay and high-precision time synchronization of large-scale deployed power equipment.

3. The time synchronization method according to claim 1, wherein: The step S4 includes: By jointly optimizing the clock source selection strategy at the large time scale and the clock source weight allocation strategy at the small time scale, minimize the weighted sum of the time synchronization delay and the time synchronization error of the intelligent fusion terminal; the optimization objective is modeled as Among them, V is the weight; C1 represents the value range of the clock source selection strategy; C2 represents that within each time period, the intelligent fusion terminal selects M clock sources; C3 represents the value range of the clock source weights; C4 represents that within each time slot, the sum of the clock source weights is 1; To solve the problem of minimizing the weighted sum of the time synchronization delay and the time synchronization error of the intelligent fusion terminal, the present invention proposes a joint optimization algorithm for multi-time-scale clock source selection and weight allocation based on reinforcement learning. First, optimize the clock source selection strategy at the large time scale, and then optimize the clock source weight allocation strategy at the small time scale.

4. The time synchronization method according to claim 3, wherein: The optimization of the clock source selection strategy at the large time scale is specifically as follows: Define clock source C n The total time synchronization delay and time synchronization error within the i-th period are expressed as Define that in the $i$-th time period, the intelligent fusion terminal selects clock source $C$. n The obtained reward is Therefore, as of the i-th period, the intelligent fusion terminal selects clock source C n The average reward obtained is expressed as Suppose that as of the $i$-th period, the average reward set of all clock sources is denoted as Define the set Denote $R$ L the $M$ largest average rewards in (i).

5. The time synchronization method according to claim 4, characterized in that: Based on the ε-greedy algorithm, optimize the clock source selection strategy at the large time scale, including the following steps: Step 1: Initialization; The intelligent fusion terminal traverses all clock sources to obtain the initial reward For all clock sources C n ∈ C, initialize the clock source selection strategy x n (0) = 1, initialize the average reward Initialize the threshold ε ∈ (0, 1); Step 2: Policy Optimization; The intelligent fusion terminal generates a random number \(a\in(0,1)\) and compares the random number \(a\) with the threshold \(\varepsilon\); if \(a\leq\varepsilon\), the intelligent fusion terminal randomly selects \(M\) clock sources; Otherwise, the intelligent fusion terminal selects the clock sources with the largest \(M\) average rewards, that is Step 3: Iterative Update; The intelligent fusion terminal executes the optimal policy obtained by the above policy optimization method, calculates the reward according to the execution result according to formula (11) and updates the average reward according to formula (12).

6. The time synchronization method according to claim 3, characterized in that: The optimization of the clock source weight allocation strategy on a small time scale is as follows: There are a total of K = M! weight assignment methods based on the set Ω, and the k-th weight assignment method is defined as A k , and the set of weight assignment methods is A = {A1, A2..., A k ,..., A K}; for the sake of description, the weight assignment selection variable z k (i,t) ∈ {0, 1}, z k (i,t) = 1 means that the intelligent fusion terminal selects the weight assignment method A k for time synchronization, otherwise z k (i,t) = 0; Define the weight assignment method A for intelligent fusion terminal selection k The obtained reward is 7. The time synchronization method according to claim 6, characterized in that: The exploration-exploitation exponential weight algorithm is used to optimize the clock source weight allocation strategy on a small time scale, including the following steps: Step 1: Initialization; Initialize the uniform distribution parameter ξ ∈ (0, 1], and initialize the empirical distribution parameter λ k (i, 1) = 1, Step 2: Policy Optimization; First, calculate the probability of selecting the weight allocation method A k , denoted as Then calculate the cumulative probability, denoted as Finally, generate a random number p0(i,t) ∈ [0,1], and optimize the clock source weight allocation strategy according to the relationship between the random number p0 and the cumulative probability F i,t (k), which is expressed as Step 3: Iterative Update; The intelligent fusion terminal performs time synchronization according to the clock source weight allocation strategy and calculates the reward according to formula (14), and updates the empirical distribution parameters, denoted as where γ is a weight adjustment factor, is the reward evaluation value, expressed as 8. An intelligent fusion terminal based on dynamic weighting of multiple clock sources, characterized in that: The intelligent fusion terminal is implemented based on the time synchronization method described in any one of claims 1 to 7, and specifically includes: a power supply module, a communication module, a storage module, a multi-clock source dynamic weighted time synchronization module, a processor module, and a timing module; The power supply module is responsible for providing continuous and stable power supply inside the intelligent fusion terminal to support the continuous operation of the intelligent fusion terminal; The communication module supports multiple communication methods including satellite communication, 5G, optical fiber, AC / DC power line carrier, and industrial Ethernet; on the one hand, the communication module obtains the time synchronization information of Beidou and GPS satellites through satellite communication, and obtains the time synchronization information of ground stations through 5G and optical fiber communication; on the other hand, the communication module uses AC / DC power line carrier and industrial Ethernet to time the power terminals; this module can support multiple communication methods at the same time to avoid abnormal information transmission caused by link interruption; The storage module stores the code related to the joint optimization algorithm of the multi-time scale clock source selection and weight allocation strategy, stores the time synchronization information of satellites and ground stations, and the data required to execute the joint optimization algorithm of the multi-time scale clock source selection and weight allocation strategy; The multi-clock source dynamic weighted time synchronization module optimizes the clock source selection strategy on a large time scale and the clock source weight allocation strategy on a small time scale by using the joint optimization algorithm of the multi-time scale clock source selection and weight allocation strategy according to the time synchronization information obtained by the communication module and the data stored in the storage module, so as to achieve low-latency and high-precision time synchronization; The processor module provides computing services for the operation of the optimization algorithm; The timing module responds to and manages the time synchronization requests of power equipment, and based on the high-precision time synchronization information obtained by the multi-clock source dynamic weighted time synchronization module, times the un-timed power equipment through the communication module.

Images