Parameter updating method of digital twin model applied to power communication optical cable
Through the least squares method and the gradual update algorithm, combined with the forgetting factor and mean square error weight coefficient, the problem of untimely update of the digital twin model of the power communication optical cable is solved, the system stability and consistency is improved, maintenance costs are reduced, and maintenance costs are achieved. Rapid response and maximization of resource utilization.
Patent Information
- Application Number
- CN202410432363.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-11
- Publication Date
- 2024-11-01
- Estimated Expiration
- 2044-04-11
AI Technical Summary
The existing power communication cable digital twin model is not updated in time, resulting in model stability and synchronization problems, affecting system performance.
The least squares method is used to update the online parameter, introduce the forgetting factor and mean square error weight coefficient, and realize the adaptive update of the model through the progressive update algorithm to ensure the synchronization between the model and the actual system.
It improves the stability and accuracy of the system model, reduces maintenance costs, and maximizes rapid response and resource utilization, ensuring that the model is updated in time when changes, maintaining high consistency and prediction capabilities.
Smart Images

Figure CN118227953B_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the field of power communication, and in particular to a method for updating a digital twin model of a power communication optical cable based on a tuna swarm optimization (TSO) least squares method. Background Art
[0002] Digital Twin (DT) is a simulation process that fully utilizes data such as physical models, sensors, and operation history, integrates multi-disciplinary, multi-physical quantities, and multi-scales, and reflects the entire life cycle of physical equipment and products by mapping the physical space in the digital space. Digital Twin solves the bottleneck that information space and physical space cannot interact and merge in intelligent manufacturing. Online identification of controlled object parameters and accurate judgment of model mismatch are the keys to realizing digital twins.
[0003] The least squares method is a basic estimation method in the field of system identification. It can be used for both offline and online estimation. Since the identification algorithm proposed by the least squares principle is relatively simple to implement, the least squares method is often used to provide a solution to the problem when other parameter identification methods are difficult to use.
[0004] Power communication optical cables are the basic guarantee for the orderly operation of the power system. As the working time of power optical cables increases, the failure rate gradually increases, which greatly increases the cost of repairing and maintaining optical cables. The power communication optical cable status monitoring system driven by digital twin technology can effectively ensure the stable operation of the power system, but changes in working conditions and environmental changes may lead to model mismatch. Therefore, it is of great practical significance to ensure that the model is adaptively updated without disturbance when mismatch occurs. Summary of the invention
[0005] In order to solve the problems that the existing digital twin model of power communication optical cables is not updated in time, which in turn affects the stability of the model and further affects its synchronization with the actual system, the present invention provides a digital twin model parameter updating method applied to power communication optical cables.
[0006] The parameter updating method of the digital twin model applied to the power communication optical cable is implemented by the following steps:
[0007] Step 1: Online identification and parameter update of the digital twin model of power communication optical cable;
[0008] The twin model parameters are updated in real time using the least squares method, a forgetting factor is introduced, and the corrected identification parameters of the twin model are obtained in real time;
[0009] Step 2: Use the mean square error as the twin model performance evaluation index, and add a weight coefficient to determine whether the mean square error exceeds the mean square error threshold range. If yes, return to step 1; otherwise, execute step 3;
[0010] Step three: Use a progressive update algorithm to transfer the corrected identification parameters obtained in real time back to the twin model; and realize the online update of the digital twin model of the power communication optical cable.
[0011] Beneficial effects of the present invention: The model updating method described in the present invention is implemented through three steps: online identification of digital twin model parameter update; selection of model performance mismatch evaluation indicators; and model parameter update and maintenance.
[0012] In the method described in the present invention, the parameters are updated online by the least squares method, and after determining the mismatch, the parameters are passed back to the model by the progressive update algorithm. The method described in the present invention can realize the online iterative update of the system model parameters, effectively solve the system model performance degradation, improve the stability of the system, and make the digital twin model of the power communication optical cable more reliable.
[0013] In the method described in the present invention, by updating the model parameters in real time, the high consistency between the model and the actual system can be maintained, the accuracy and predictive ability of the model can be improved, and the digital twin model can be made adaptive, that is, the digital twin model does not require manual intervention when the actual system data continues to increase and maintains accuracy and predictive ability. In addition, the digital twin model parameter update method obtains the data of the physical system in real time and synchronizes it with the virtual digital twin model. The parameter update method can ensure that the model parameters are updated in time when the physical system changes to maintain the synchronization of the model with the actual system.
[0014] The present invention uses the least squares method to update the model parameters in real time, and introduces the forgetting factor to eliminate the influence of data redundancy. Through the online identification algorithm, the correction parameters of the model are obtained in real time; the mean square error (MSE) is used as the performance evaluation index, and a weight coefficient is introduced to reduce the influence of historical data. The coefficient is obtained by the TSO algorithm through historical data; under the premise that the model determines that the parameters are mismatched, a progressive update algorithm is used to feed back the parameters obtained by online identification to the digital twin model, so as to realize the update and maintenance of the digital twin model parameters.
[0015] The present invention can timely discover potential problems and take preventive measures to improve the reliability and stability of system operation. At the same time, the parameter update method described in the present invention can achieve rapid response and adjustment of the digital twin model, providing timely support and optimization for the system. The digital twin model after parameter update can more accurately diagnose system faults and provide maintenance suggestions, thereby reducing maintenance costs and downtime, improving system efficiency, and optimizing the operation of the power communication system and maximizing resource utilization. BRIEF DESCRIPTION OF THE DRAWINGS
[0016] Figure 1 This is a framework diagram of the method for updating the digital twin model of the power communication optical cable described in the present invention.
[0017] Figure 2 This is a flow chart of the method for updating the digital twin model of the power communication optical cable described in the present invention.
[0018] Figure 3 This is a flow chart of the least squares method in the method for updating the digital twin model of the power communication optical cable described in the present invention.
[0019] Figure 4 This is a flow chart of the TSO algorithm in the method for updating the digital twin model of the power communication optical cable described in the present invention.
[0020] Figure 5 This is a comparison chart between the forgetting factor least squares method and the least squares method in the power communication optical cable digital twin model updating method described in the present invention.
[0021] Figure 6 This is a comparison chart of the convergence curves of the TSO algorithm and other algorithms in the method for updating the digital twin model of the power communication optical cable described in the present invention. DETAILED DESCRIPTION
[0022] Combination Figures 1 to 4 This embodiment is described as a method for updating parameters of a digital twin model applied to a power communication optical cable. The method is implemented by the following steps:
[0023] Step 1: Online identification and parameter update of the digital twin model of power communication optical cables;
[0024] S11: Establish a digital twin model of power communication optical cable;
[0025] In one embodiment, according to the specific form of operation of the power communication optical cable and the physical laws between various parameters, the mechanism twin model of the power optical fiber communication line state monitoring system can be expressed as:
[0026] (y1(k),y2(k),…y n (k))=f(x1(k),x2(k),…x n(k)) (1)
[0027] Among them, x1(k),x2(k),…x n (k) is the input parameter of the model, including temperature change, strain change, inherent properties of the optical fiber line, etc.; f is the corresponding relationship between the model input and output; y1(k), y2(k),…y n (k) is the output of the model, including Brillouin frequency shift, Brillouin gain peak, optical power value, and other environmental information data.
[0028] S12: Online identification of power communication optical cable digital twin model parameter update;
[0029] Online identification, also known as real-time identification, refers to the continuous collection of current input and output data while the system is already running, and the correction of the parameters of the previous model through the newly introduced data based on the results of the previous parameter estimation. The least squares method is widely used in real-time correction of model parameters. The basic idea can be summarized as follows: the current estimated value in is the kth observation value.
[0030] The digital twin model of power communication optical cable can be written as:
[0031] A(z -1 )(y1(k),y2(k),…,y i (k)) = B(z -1 )(u1(kd),u2(kd),…,u i (kd))+ξ(k)(2)
[0032] Right now:
[0033] A(z -1 )y(k)=B(z -1 )u(kd)+ξ(k) (3)
[0034] In the formula, ξ(k) has a mean of 0 and a variance of Uncorrelated random noise that follows a normal distribution; u(kd) represents the kdth input value; A(z -1 )、B(z -1 ) is the delay operator polynomial, denoted as:
[0035]
[0036] In the formula, They are A(z -1 )、B(z -1) The coefficients na and nb after expansion represent the number of parameters; parameter estimation determines n based on the input data such as temperature change and strain change obtained from the current measurement, as well as the output data such as Brillouin frequency shift and optical power value. a +n b +1 parameter.
[0037] Formula (3) can be written in the following least squares form:
[0038]
[0039] for and θ, generally:
[0040]
[0041] in, is the data vector; θ is the parameter vector to be estimated.
[0042] Assuming there are L sets of input and output data {y(k),u(k),k=1,2,...L}, the least squares estimate can be obtained as:
[0043]
[0044] in, Φ k and Y k The intermediate variables have no practical significance. and y(k) are obtained recursively.
[0045] The least squares estimate The recursive formula is:
[0046]
[0047] Among them, the least squares method is to estimate the model value through the matrix K(k) and P(k) (K(k) and P(k) are intermediate variables with no practical significance) As the number of iterations increases, The correction ability will become weaker and weaker as K(k) and P(k) gradually decrease, and cause data redundancy, that is, data saturation will make the newly introduced data unable to be effectively updated. Therefore, this embodiment introduces a forgetting factor, also called a forgetting factor least squares algorithm (FFRLS).
[0048] The estimation formula after introducing the forgetting factor is:
[0049]
[0050] Among them, λ is the forgetting factor, which is a positive number close to 1.
[0051] like Figure 5 As shown, Figure 5 It is a comparison chart of FFRLS and RLS. It can be seen from the figure that FFRLS has a better fitting effect with the true value. The specific test results are compared in Table 1. Table 1 is the test comparison results of FFRLS and RLS.
[0052] Table 1
[0053]
[0054] It can be seen from Table 1 that the FFRLS algorithm is superior to the RLS algorithm in terms of RMSE, ME and MAE indicators.
[0055] Step 2: Select the performance evaluation indicators of the digital twin model of power communication optical cable;
[0056] In the online update of the digital twin model, the performance should not be focused on a certain moment, but on the comprehensive deviation of each moment within a certain interval, and the performance at the current moment should be more valued. Based on the above characteristics, the mean square error is used as the performance evaluation indicator. The general mean square error calculation formula is:
[0057]
[0058] Where N is the data length; y pv (k) is the feedback value of the kth data; y sp (k) is the set value of the kth data; e 2 (k) is the deviation estimate of the kth data.
[0059] The model described in the present invention is a multi-output system, and the output variables have different weights, so the mean square error can be expressed as:
[0060]
[0061] In the formula, is the weight of each output variable; is the deviation estimate of the kth data, n i is the number of variables; MSE is the mean square error weight coefficient. In order to weaken the data saturation phenomenon, the present invention introduces a weight coefficient to reduce the influence of historical data. The new mean square error can be expressed as:
[0062]
[0063] Among them, λ MSEis the mean square error weight coefficient, which is obtained by the TSO algorithm through historical data. If it exceeds the mean square error MSE threshold range, it is determined that the parameters updated in step 1 are returned to the digital twin model of the power communication optical cable.
[0064] In this embodiment, the mean square error weight coefficient is determined by using the TSO optimization algorithm. The algorithm optimizes the problem by simulating the foraging behavior of a tuna group. It has the characteristics of strong optimization ability and fast convergence speed.
[0065] In this embodiment, the specific process of simulating the foraging behavior of a tuna group is as follows:
[0066] 1. Population initialization;
[0067]
[0068] in, is the initial position of the i-th individual at the t-th iteration; ub and lb are the upper and lower bounds of the search space, respectively. MSE It is usually a number close to 1, so ub can be 1 and lb can be 0.9; NP is the tuna population; rand is a random vector uniformly distributed in the range [0,1].
[0069] 2. Spiral foraging;
[0070] Tuna populations chase prey by forming tight spirals. In addition to chasing prey, groups of tuna also exchange information with each other. Each tuna follows the previous one, so information can be shared between neighboring tunas. Based on the above principle, the mathematical formula of the spiral foraging strategy is as follows:
[0071] when When:
[0072]
[0073] when When:
[0074]
[0075]
[0076]
[0077] β=e hl ·cos(2πh) (18)
[0078]
[0079] in, is the position of the i-th individual at the t+1-th iteration; is the best individual position at the tth iteration; is a randomly generated reference point in the search space; is the position of the i-1th individual in the tth iteration; α1 and α2 are weight coefficients that control the tendency of the individual to move toward the best individual and the previous individual; β and l are intermediate variables with no practical significance; r is a constant; h is a random number that obeys a uniform distribution between [0,1].
[0080] 3. Parabolic foraging;
[0081] In addition to forming spiral foraging, tuna also forms parabolic cooperative foraging. Tuna forms a parabola shape with food as the reference point. In addition, tuna finds food by searching around. Assuming that the selection probability of both methods is 50%, the two methods are carried out simultaneously. The specific mathematical model is described as follows:
[0082]
[0083]
[0084] Here, TF is a random number with a value of 1 or -1.
[0085] In this embodiment, the tuna cooperates in hunting through the above two foraging strategies and then finds prey. During the entire optimization process, all individuals of TSO are continuously updated and calculated until the final condition is met, and then the optimal individual and the corresponding fitness value are returned. In this embodiment, the mean square error is used as the fitness function, and the mean square error weight coefficient that minimizes the fitness function is found through the TSO algorithm.
[0086] In this implementation, in order to verify the superiority of the TSO algorithm, the benchmark function is selected x s ∈[-1.28,1.28] for verification, and compare the TSO algorithm with the gray wolf algorithm, particle swarm algorithm, sparrow algorithm, and whale algorithm. The results are as follows Figure 6 As shown. Figure 6 It can be seen that compared with other intelligent search algorithms, the TSO algorithm improves the global search capability and accelerates the convergence speed, showing the best effect.
[0087] In order to update the parameter model in time and reduce the amount of mean square error calculation, the mean square error is set to a sliding window calculation method. If the window size does not reach the maximum value, then:
[0088]
[0089] Among them, N size is the number of data in the window at the current moment.
[0090] If the maximum window is reached, the mean square error update iteration formula becomes:
[0091]
[0092] Step 3: Use the progressive update algorithm to pass the real-time parameters back to the model;
[0093] After the parameters updated by the identification algorithm in step 1 are determined to be mismatched by step 2, the new parameters need to be passed back to the system model to improve the control performance. If the identification parameters are directly passed back to the model, large fluctuations often occur, affecting the stability of the system. In order to avoid such situations, a gradual method is adopted to gradually correct the parameter changes, dynamically select the corresponding rate value for iterative update, and feed the corrected parameters back to the model to realize the online update of the digital twin model of the power optical communication line monitoring system.
[0094] Therefore, this embodiment adopts a progressive update algorithm, which can effectively avoid interference and fluctuation during parameter update.
[0095] Step 1: The algorithm model parameters before and after updating the identification are as follows:
[0096]
[0097] The parameters before identification are:
[0098]
[0099] After identification, the parameters are:
[0100]
[0101] Updated at a fixed rate, parameter a i 、b i The update is shown in formula (13):
[0102]
[0103] Where, ν ai , ν bi They are the identification parameters a i 、b i Update rate; a i (k) is the parameter value of the kth data; b i (k) is the kth data parameter value.
[0104] Since a fixed update rate may result in a long update time and untimely parameter correction, this implementation changes the fixed update rate to an iterative update mode, that is, the farther away from the identification parameter, the faster the update rate, and the closer to the identification parameter, the slower the update rate. i, the variable rate update formula is:
[0105]
[0106] In the formula, γ is the iterative update coefficient; The lower limit of the rate update; is the upper limit of the rate update. The correction parameters are fed back to the model through the progressive update algorithm to realize the online update of the digital twin model of the power communication optical cable.
[0107] In this embodiment, when the identification parameters of the model are updated, the update rate v ai (k) i (k) With approach First increase and then decrease, and The limitation prevents the rate update from exceeding or falling below the limit, so that the update rate is always within a safe range, ensuring the rapidity and stability of parameter updates.
[0108] The technical features of the above-mentioned embodiments can be combined arbitrarily. In order to make the description concise, not all possible combinations of the technical features in the above-mentioned embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0109] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.
Claims
1. A digital twin model parameter updating method applied to power communication optical cables, characterized by: The method is implemented by the following steps: Step 1: Online identification and parameter update of the digital twin model of power communication optical cable; The twin model parameters are updated in real time using the least squares method, a forgetting factor is introduced, and the corrected identification parameters of the twin model are obtained in real time; Step 2: Use the mean square error as the twin model performance evaluation index, and add a weight coefficient to determine whether the mean square error exceeds the mean square error threshold range. If yes, return to step 1; Otherwise, go to step 3; Step 3: Using a progressive update algorithm to transfer the corrected identification parameters obtained in real time back to the twin model; realizing online update of the digital twin model of the power communication optical cable; Specifically, the identification parameters are gradually corrected using a progressive update algorithm, and the corresponding rate values are dynamically selected for iterative updates; for the identification parameter a i , the variable rate update formula is: Where, ν ai (0) is the initial value of the rate update, ν ai (k) is the update rate at time k, ν ai (k-1) is the update rate at time k-1, γ is the iterative update coefficient, is the lower limit of the rate update, The upper limit of the rate update.
2. The digital twin model parameter updating method applied to power communication optical cable according to claim 1 is characterized in that: The least square method is used to correct the twin model identification parameters in real time, and the forgetting factor is introduced. The estimation formula after the forgetting factor is introduced is: in, is the least squares estimate, K(k) and P(k) are matrices, y(k) is the least squares form of the twin model, is the data vector, and λ is the forgetting factor.
3. The digital twin model parameter updating method applied to power communication optical cable according to claim 2 is characterized in that: In step 2, the mean square error is expressed as follows: In the formula, is the weight of each output variable; is the deviation estimate of the kth data, n i is the number of variables; MSE is the mean square error weight coefficient.
4. The method for updating parameters of a digital twin model applied to a power communication optical cable according to claim 3 is characterized in that: The mean square error weight coefficient is determined by a TSO optimization algorithm; specifically, by simulating the foraging behavior of a tuna school, taking the mean square error as a fitness function, and finding the mean square error weight coefficient that minimizes the fitness function through the TSO algorithm.
5. The method for updating parameters of a digital twin model applied to a power communication optical cable according to claim 3 is characterized in that: Set the mean square error to the sliding window calculation method. If the window size does not reach the maximum value, then: Where N size is the number of data in the window at the current moment; If the sliding window reaches the maximum value, the mean square error update iteration formula becomes:
