Dynamic correction method and system for parameters of digital twin model of line equipment based on MPC

Abstract

A method and system for dynamic correction of parameters of a digital twin model of line equipment based on MPC. The method first collects historical operating data from transmission line construction to construct a digital twin model, then divides its system properties into static, dynamic and performance properties. Subsequently, a Seq2Seq algorithm is integrated with a multi-head self-attention mechanism Transformer model to construct a time series prediction framework, and an adaptive memory decay mechanism is introduced to optimize the output objective function. Finally, static properties are updated through event-driven operation, and dynamic and performance properties are corrected based on an MPC control strategy. The present invention effectively controls the convergence speed of property parameters by designing an adaptive memory decay mechanism and adding constraints, thereby reducing the computational complexity of the model. In addition, various properties of the system are corrected, adjusted and optimized, achieving a coordinated closed-loop matching between the output of the digital twin model and the operation of the physical system, thereby achieving higher prediction accuracy.

Description

Technical Field

[0001] The present invention relates to a method for correcting parameters of a digital twin model, and in particular to a method and system for dynamically correcting parameters of a digital twin model of line equipment based on MPC. Background Art

[0002] With the rapid development of science and technology, digital twin technology, as a cutting-edge information technology means, has made significant progress in industries such as industry, energy, and transportation. By establishing high-fidelity models to match specific physical entities, digital twin technology provides a new method for status assessment, operation monitoring, and performance prediction of physical equipment.

[0003] Transmission line equipment, as a core component of the power grid, is the driving force behind its digital and intelligent transformation. However, in existing digital construction of transmission line equipment, most have only achieved digital twin construction of the equipment's appearance, a panoramic display of its operating status, or simple equipment status evaluation. Construction is still at an early stage and cannot meet the practical application needs of key business scenarios in design, infrastructure construction, and operation and maintenance. Furthermore, the physical entities of the transmission line equipment considered during construction are often subject to multiple uncertainties during operation. This places higher demands on the synchronization accuracy of the parameters of the digital twin model of transmission line equipment. Current research on digital twin technology often fails to timely identify and adaptively adjust the digital twin model, resulting in a gradual deviation between the model's prediction results and actual operation, reducing the prediction accuracy of the digital twin model. Existing approaches, such as the RB unscented Kalman filter or methods based on the extended Kalman filter, although they have strong model convergence and good convergence effects, are computationally complex, time-consuming, and suffer from the sensitivity of initial settings to parameter solutions. Therefore, how to improve the prediction accuracy of digital twin models and reduce the computational complexity remains a problem that needs to be urgently solved by existing technologies. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above-mentioned defects and problems existing in the prior art and to provide a method and system for dynamic correction of parameters of a digital twin model of line equipment based on MPC, which is simple to calculate and has high prediction accuracy.

[0005] To achieve the above objectives, the technical solution of the present invention is: a method for dynamic correction of parameters of a digital twin model of line equipment based on MPC, comprising:

[0006] S1. Collect historical operation data during transmission line construction and build a digital twin model of transmission line construction equipment;

[0007] S2. Divide the system attributes of the digital twin model of transmission line construction equipment into static attributes, dynamic attributes, and performance attributes;

[0008] S3. In the digital twin model of transmission line construction equipment, the Seq2Seq algorithm is integrated with the Transformer model with a multi-head self-attention mechanism to build a time series prediction framework. An adaptive memory decay mechanism is introduced to optimize the output objective function of the time series prediction framework and control the parameter convergence speed of dynamic and performance attributes.

[0009] S4. Update static attributes in the digital twin model of transmission line construction equipment through event-driven;

[0010] S5. Modify the dynamic and performance attributes of the digital twin model of transmission line construction equipment based on the MPC control strategy.

[0011] The step S3 introduces an adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework, specifically:

[0012] For three different types of data points in historical operation data, namely stable data points, fluctuating data points and mutation data points, the memory decay factor λ(t) is introduced to optimize the output objective function to adapt to the dynamic characteristics of the data points; the details are as follows:

[0013] For stationary data points: keep the memory decay factor at a fixed constant value λ between [0, 1];

[0014] For fluctuating data points and mutation data points: introduce the correction term Δλ(t), the expression of the correction term is as follows:

[0015] λ(t)=λ(t-1)+Δλ(t);

[0016] Then for the fluctuating data points, the correction is: Δλ(t)=α(1-λ(t-1))α>0;

[0017] Then for the mutation data point, the correction is: Δλ(t)=β(1+λ(t-1))β>0;

[0018] Among them: α and β are adjustment coefficients, t is the current time step;

[0019] The optimized output objective function is as follows:

[0020]

[0021] Where: J is the output objective function, w1 and w2 are weight coefficients, y p is the predicted output, y s is the actual output, ΔU is the control increment, T is the total number of time steps, and t is the current time step;

[0022] The constraints of the output objective function are as follows:

[0023] V m_min ≤V m (t)≤V m_max

[0024] I m_min ≤I m (t)≤I m_max

[0025] T m_min ≤T m (t)≤T m_max

[0026] ΔU min ≤ΔU(t)≤ΔU max ;

[0027]

[0028] f min ≤f(t)≤f max

[0029] V min ≤V(t)≤V max

[0030] 0 <w1<1.0<w2<1;

[0031] w1+w2=1;

[0032] Where: V m , I m 、T m All are state quantities, V m_min 、V m_max , I m_min , I m_max 、T m_min 、T m_max The upper and lower limits of its state quantity, ΔU is the increment value, ΔU min , ΔU max are the upper and lower limits of the increment value, ε is the rate of change constraint, f is the frequency, V is the voltage, f min 、f max are the upper and lower limits of the frequency, and dt is the time increment.

[0033] The expression of the multi-head self-attention mechanism in step S3 is as follows:

[0034]

[0035] head i =Attention(QWi Q , KW i K , VW i V );

[0036] MultiHead(Q,K,V)=Concat(head1,...,head h )W ° ;

[0037] Among them: Attention(Q, K, V) is the output matrix of single-point attention, Q, K, V are all matrices, sofftmax is the activation function, d k is the key value of the input dimension, W i Q 、W i K 、W i V are the weight matrices corresponding to Q, K, and V in the i-th attention head, MultiHead(Q, K, V) is the output matrix of multi-head attention, and Concat is the concatenation operation.

[0038] The step S4, updating the static attributes in the digital twin model of the transmission line construction equipment through event-driven means: based on the changes in the construction site during the transmission line construction and the changes in the structural attributes corresponding to the construction site changes, triggering the re-identification and re-assignment of the parameters of the static attributes, and updating the assigned values ​​to the digital twin model of the transmission line construction equipment.

[0039] The step S5 specifically includes:

[0040] S51. Define error limits within different time steps t; the expression of the error limit is as follows:

[0041]

[0042] Where: θ(t) is the error limit within time step t, y p is the predicted output, y s is the actual output, t is the current time step;

[0043] S52. For dynamic attributes, periodically obtain measured data of the dynamic attributes at each time step t and use it as the actual output; for performance attributes, predict the parameters of the performance attributes based on the time series prediction framework and use the prediction results as the actual output;

[0044] S53. Based on the actual output historical data, a maximum error limit θ′ is set; when θ(t)>θ′, a parameter update is triggered, and the parameters of the dynamic attributes and performance attributes are updated respectively, and then step S53 is performed;

[0045] S53. Iterate the digital twin model of the transmission line construction equipment and set a value strategy for the output objective functions w1 and w2 in the model to balance the values ​​of the weight coefficients w1 and w2 and modify the performance attributes; the value strategy is as follows:

[0046]

[0047] Wherein: w1 and w2 are weight coefficients; lter is the number of iterations of the MPC control strategy; N1 is the iteration threshold; N is the maximum number of iterations; A1 is the value of w1 when the number of iterations is less than N1; A2 is the value of w2 when the number of iterations is less than N1; B1 is the value of w1 when the number of iterations is between N1 and N; B2 is the value of w2 when the number of iterations is between N1 and N.

[0048] The time series prediction framework includes an encoder and a decoder;

[0049] The stack structure of the Transformer in the encoder includes a first encoding sublayer and a second encoding sublayer arranged in sequence; the first encoding sublayer includes a multi-head self-attention mechanism and an accumulation layer normalization mechanism arranged in sequence; the second encoding sublayer includes a feedforward neural network and an accumulation layer normalization mechanism arranged in sequence;

[0050] The stack structure of the Transformer in the decoder includes a first decoding sublayer, a second decoding sublayer, and a third decoding sublayer arranged in sequence; the first decoding sublayer includes a masked multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the second decoding sublayer includes a multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the third decoding sublayer includes a feedforward neural network and a cumulative sum layer normalization mechanism arranged in sequence;

[0051] The encoding second sublayer is connected to the decoding second sublayer.

[0052] The digital twin model of the transmission line construction equipment includes a physical line and a digital twin line;

[0053] The physical circuit refers to: the entity reference target of digital twin circuit modeling; specifically including physical entities and IoT perception data;

[0054] The digital twin circuit refers to: the digital expression of the physical circuit in the virtual space; specifically, it includes the modeling of static properties, dynamic properties and performance properties.

[0055] The static attributes refer to: structural attributes of transmission line construction equipment; specifically, line length, conductor cross-sectional dimensions, tower parameters, insulator parameters, transformer capacity, supplier information of used materials, and specific information of construction equipment;

[0056] Dynamic attributes refer to dynamic indicators whose values ​​change with working conditions and reflect the real-time operating status of transmission line construction systems and equipment. Specifically, they include tower installation progress, material consumption, wind speed and direction, and equipment temperature.

[0057] The performance attributes refer to: performance indicators that reflect the operational stability of transmission line construction systems and equipment; specifically, they include maintenance records, life parameters, process parameters, work efficiency of workers or equipment, and quality control indicators.

[0058] A dynamic correction system for parameters of a line equipment digital twin model based on MPC, which is applied to the above method, comprises:

[0059] A digital twin model construction module is used to collect historical operation data during transmission line construction and build a digital twin model of transmission line construction equipment;

[0060] A system attribute division module is used to divide the system attributes of the digital twin model of transmission line construction equipment into static attributes, dynamic attributes, and performance attributes;

[0061] The digital twin model optimization module is used to integrate the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model in the initial digital twin model of transmission line construction equipment to build a time series prediction framework. It also introduces an adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework.

[0062] The attribute update module is used to update the static attributes in the digital twin model of transmission line construction equipment through event-driven;

[0063] The attribute correction module is used to correct the dynamic attributes and performance attributes in the digital twin model of transmission line construction equipment based on the MPC control strategy.

[0064] A device for dynamic correction of parameters of a digital twin model of line equipment based on MPC, the device comprising a processor and a memory;

[0065] The memory is used to store computer program code and transmit the computer program code to the processor;

[0066] The processor is used to execute the above-mentioned MPC-based line equipment digital twin model parameter dynamic correction method according to the instructions in the computer program code.

[0067] Compared with the prior art, the present invention has the following beneficial effects:

[0068] The present invention provides a method and system for dynamic correction of parameters of a digital twin model of line equipment based on MPC. The method first collects historical operation data in transmission line construction to construct a digital twin model of transmission line construction equipment, and then divides its system properties into static properties, dynamic properties and performance properties. Then, the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model are integrated to construct a time series prediction framework, and an adaptive memory decay mechanism is introduced to optimize the output objective function of the time series prediction framework. Finally, the static properties in the digital twin model of transmission line construction equipment are updated through event-driven, and the dynamic properties and performance properties in the digital twin model of transmission line construction equipment are corrected based on the MPC control strategy. In application, this design effectively controls the convergence speed of attribute parameters by designing an adaptive memory decay mechanism and adding constraints, thereby reducing the computational complexity of the model, and correcting, adjusting and optimizing various properties of the system, thereby realizing a coordinated closed-loop matching between the output of the digital twin model and the operation of the physical system, so that it has higher prediction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0069] Figure 1 It is a flow chart of the method steps of the present invention.

[0070] Figure 2 Schematic diagram of system attribute division of the present invention.

[0071] Figure 3 It is a structural diagram of the time series prediction framework of the present invention.

[0072] Figure 4 3 is a schematic diagram comparing the model frequencies before and after parameter correction of the present invention.

[0073] Figure 5 2 is a schematic diagram of error comparison of different methods of the present invention.

[0074] Figure 6 It is a schematic diagram of the system structure of the present invention.

[0075] Figure 7 It is a schematic diagram of the device structure of the present invention.

[0076] In the figure: digital twin model construction module 1, system attribute division module 2, digital twin model optimization module 3, attribute update module 4, attribute correction module 5, processor 6, memory 7, computer program code 71. DETAILED DESCRIPTION

[0077] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0078] Example 1:

[0079] See also Figure 1 , a method for dynamic correction of parameters of a line equipment digital twin model based on MPC, comprising:

[0080] S1. Collect historical operation data during transmission line construction and build a digital twin model of transmission line construction equipment;

[0081] Furthermore, the digital twin model of the transmission line construction equipment includes a physical line and a digital twin line;

[0082] The physical line refers to: the entity reference target of digital twin line modeling; specifically, it includes physical entities (such as transmission lines, equipment, sensors, etc.) and IoT perception data; various physical entities are combined into a whole through logical relationships such as operation plans and process routes to form a physical line with certain operating capabilities, and the operating data of the physical line is obtained through IoT devices and transmitted to the digital twin line via the network, forming a data interaction channel between the physical line and the digital twin line.

[0083] The digital twin circuit refers to: the digital expression of the physical circuit in the virtual space; specifically, it includes the modeling of static properties, dynamic properties and performance properties.

[0084] In this embodiment, static properties, dynamic properties and performance properties are established on the basis of the geometric model, and then the behavior rules of the transmission line construction equipment in the digital twin model and the operation logic of the digital twin line are improved.

[0085] For example: building a digital twin engineering scenario on the Plant Simulation platform.

[0086] Manage model resources for tower models, construction machinery models, tool models, etc. referenced in digital twin engineering scenarios to form a model library;

[0087] Construction planning and construction simulation are linked to form a resource library of different processes and methods such as foundation construction, tower assembly construction, and line stringing construction; the resource library supports viewing safety measures, quality measures, safety accident cases, and major machinery and tools information associated with different processes and methods, thereby providing model data support for twin projects.

[0088] The operation data of the construction process is obtained through the Internet of Things devices and transmitted to the digital twin line via the network to establish the corresponding digital twin model. The key status parameters such as equipment operation status, equipment quantity, equipment performance, etc. are determined as the input and output of the digital twin model.

[0089] S2. Divide the system attributes of the digital twin model of transmission line construction equipment into static attributes, dynamic attributes, and performance attributes;

[0090] Furthermore, real-time data is the foundation for building digital twin models and a prerequisite for parameter correction. This data is collected by collecting basic information about transmission line construction equipment, environmental and climate information, operating status and maintenance records, material and supplier information, testing equipment, and other data, and then transmitted to the digital twin model.

[0091] See also Figure 2 Specifically, the static attributes refer to: structural attributes of transmission line construction equipment; specifically including line length, conductor cross-sectional dimensions, tower parameters, insulator parameters, transformer capacity, supplier information of used materials, and specific information of construction equipment (such as rotary excavators, cement mixers, winches, cranes, etc.); in actual application, a comprehensive method of on-site measurement and electrical testing is used to confirm the static attribute values. Static attributes provide basic data for describing the physical structure and steady-state characteristics of transmission system construction equipment.

[0092] The dynamic attributes are dynamic indicators whose values ​​change with working conditions and reflect the real-time operating status of the transmission line construction system and equipment. They specifically include tower installation progress, material consumption, wind speed and direction, and equipment temperature. In actual use, the measurement and collection of dynamic attributes are achieved through sensor technology, monitoring systems, manual observation and recording, and integration with SCADA system data. Dynamic attributes directly drive the state evolution of the digital twin model.

[0093] The performance attributes refer to: performance indicators that reflect the operational stability of the transmission line construction system and equipment; specifically, they include maintenance records, life parameters, process parameters, work efficiency of workers or equipment, and quality control indicators (such as the quality score of welded joints, cement quality, geometric accuracy of tower structures, etc.); these data are difficult to obtain through direct measurement. In this embodiment, by analyzing historical operating data, parameter prediction of performance attributes is achieved, and it serves as a data source for key performance parameters to drive the optimization and evaluation of the digital twin model.

[0094] S3. In the initial digital twin model of transmission line construction equipment, the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model are integrated to build a time series prediction framework. An adaptive memory decay mechanism is introduced to optimize the output objective function of the time series prediction framework and control the parameter convergence speed of dynamic attributes and performance attributes.

[0095] In this embodiment, based on the seq2seq algorithm, a multi-head attention Transformer model is integrated for time series prediction, thereby improving the accuracy of sequence prediction. The multi-head self-attention Transformer model has been widely used and studied in the field of machine translation. It takes multiple time series data points as input and predicts multiple time series data points as output, forming a multi-input and multi-output time series prediction framework. In this process, the time series prediction framework needs to be modified. The main aspects of the modification are the design of the stacking method and the design of the attention mechanism. The details are as follows:

[0096] See also Figure 3 , the time series prediction framework includes an encoder and a decoder;

[0097] The stack structure of the Transformer in the encoder includes a first encoding sublayer and a second encoding sublayer arranged in sequence; the first encoding sublayer includes a multi-head self-attention mechanism and an accumulation layer normalization mechanism arranged in sequence; the second encoding sublayer includes a feedforward neural network and an accumulation layer normalization mechanism arranged in sequence;

[0098] The multi-head self-attention mechanism is used to capture relevant information in the sequence, and the feedforward neural network is used to perform nonlinear transformation and mapping on the input. Between these two sub-layers, residual connections are used to facilitate information transfer, and the accumulation and layer normalization mechanisms are used to enhance the stability of the model. To ensure consistent dimensionality, all sub-layers in the encoder and the embedding layer will generate output dimensions of d. model feature representation.

[0099] The stack structure of the Transformer in the decoder includes a first decoding sublayer, a second decoding sublayer, and a third decoding sublayer arranged in sequence; the first decoding sublayer includes a masked multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the second decoding sublayer includes a multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the third decoding sublayer includes a feedforward neural network and a cumulative sum layer normalization mechanism arranged in sequence;

[0100] The encoding second sublayer is connected to the decoding second sublayer and is used to perform a multi-head attention mechanism on the encoder output to focus on relevant information of the input sequence.

[0101] Furthermore, for the single-point attention with matrix K scale, the input includes the query and key value d of dimension k and dimension value d v , get each point information by querying the key value, and then divide it by the key value of the input dimension The SoftMax function is used to obtain the weight of the value. Finally, the attention function is calculated by querying at the same time, and they are packaged together into the matrix Q. The key and value are also compressed into the matrices K and V. The calculation formula of the final output matrix is ​​as follows:

[0102]

[0103] Among them: Attention(Q, K, V) is the output matrix of single-point attention, Q, K, V are all matrices, softmax is the activation function, d k is the key value of the input dimension;

[0104] In the single-point attention mechanism, each attention head can perform attention calculations independently and simultaneously to capture the correlation information between the input sequences. The calculation formula is as follows:

[0105] head i =Attention(QW i Q , KW i K , VW i V );

[0106] Where: W i Q 、W i K 、W i V are the weight matrices corresponding to Q, K, and V in the i-th attention head respectively;

[0107] In this embodiment, the results of parallel calculations by multiple attention heads are concatenated through the Concat function, so that the model comprehensively considers the calculation results of multiple attention heads and fully captures the correlation relationship in the input sequence. The expression of the multi-head self-attention mechanism is as follows:

[0108] MutiHead(Q,K,V)=Concat(head1,...,head h )W o ;

[0109] Among them: MultiHead(Q, K, V) is the output matrix of multi-head attention, and Concat is the splicing operation.

[0110] Furthermore, the introduction of the adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework specifically refers to:

[0111] During the collection, processing, and analysis of historical operating data, data saturation may occur. When the data volume reaches a certain level, further increasing the data volume will gradually reduce the contribution to improving the results or obtaining new information. Therefore, in order to solve this problem caused by the continuous input of online data during the calibration process, a memory decay factor is introduced to adapt to the dynamic characteristics of data points. The details are as follows:

[0112] For three different types of data points in historical operation data, namely stable data points, fluctuating data points and mutation data points, the memory decay factor λ(t) is introduced to optimize the output objective function to adapt to the dynamic characteristics of the data points; the details are as follows:

[0113] For stationary data points: keep the memory decay factor at a fixed constant value λ between [0, 1];

[0114] For fluctuating data points and mutation data points: introduce the correction term Δλ(t), the expression of the correction term is as follows:

[0115] λ(t)=λ(t-1)+Δλ(t);

[0116] Then for the fluctuating data points, the correction is: Δλ(t)=a(1-λ(t-1))α>0;

[0117] Then for the mutation data point, the correction is: Δλ(t)=β(1+λ(t-1))β>0;

[0118] Where: α and β are adjustment coefficients, and t is the current time step. By properly selecting the adjustment coefficients, the covariance matrix can be minimized, thereby optimizing the data smoothing process.

[0119] The optimized output objective function is as follows:

[0120]

[0121] Where: J is the output objective function, w1 and w2 are weight coefficients, y p is the predicted output, y s is the actual output, ΔU is the control increment, T is the total number of time steps, and t is the current time step;

[0122] The constraints of the output objective function are as follows:

[0123] V m_min ≤V m (t)≤V m_max

[0124] I m_min ≤I m (t)≤I m_max

[0125] T m_min ≤T m (t)≤T m_max

[0126] ΔU min ≤ΔU(t)≤ΔU max ;

[0127]

[0128] f min ≤f(t)≤f max

[0129] V min ≤V(t)≤V max

[0130] 0 <w1<1,0<w2<1;

[0131] w1+w2=1,

[0132] Where: V m , I m 、T m All are state quantities, V m_min 、V m_max , I m_min , I m_max 、T m_min 、T m_max The upper and lower limits of its state quantity, ΔU is the increment value, ΔU min , ΔU max are the upper and lower limits of the increment value, ε is the rate of change constraint, f is the frequency, V is the voltage, f min 、f max are the upper and lower limits of the frequency, and dt is the time increment.

[0133] S4. Update static attributes in the digital twin model of transmission line construction equipment through event-driven;

[0134] Furthermore, it specifically means: based on the changes in the construction site during the transmission line construction and the changes in the structural attributes corresponding to the construction site changes, the re-identification and re-assignment of the parameters of the static attributes are triggered, and the assignment is updated to the digital twin model of the transmission line construction equipment.

[0135] In this embodiment, the correction of static attribute parameters is event-driven, and the update of static attributes is triggered according to changes in field signals and data. For example, when the system undergoes structural changes, equipment is replaced, or other events that affect static parameters occur, the static parameters will be re-identified and re-assigned to ensure that the digital twin model is consistent with the actual system. When the equipment operation exceeds the life threshold or the model error is too large, the re-identification program is enabled to re-measure the equipment, determine the static attribute parameter value, and assign the updated value to the digital twin model.

[0136] S5. Modify the dynamic and performance attributes of the digital twin model of transmission line construction equipment based on the MPC control strategy.

[0137] In this embodiment, MPC is an advanced and powerful control strategy that uses a system model to predict future dynamic behavior and optimizes control inputs based on these predictions to achieve specific control objectives. The advantage of MPC for parameter correction is that it can automatically adjust parameters based on the system dynamic model and predictive control strategy, with the goal of minimizing performance indicators or satisfying constraints, and re-optimize parameters at each time step, thereby achieving precise control of the system. It has strong adaptability, can handle nonlinear systems, and can also optimize parameter correction for multivariable systems. Among them, the prediction equation is the design basis of the MPC control strategy. By discretizing the linear state space equation and redefining the output variable as the prediction equation, its principle is as follows:

[0138] The expression of the prediction equation is as follows:

[0139]

[0140] y=[P d , P p ];

[0141] Where: x is the system state vector, A is the state transfer matrix, B is the control input matrix, y is the output vector, c is the output matrix, P d is a dynamic attribute parameter, P p It is a static attribute parameter;

[0142] When the prediction time domain is p and the control time domain is m, then at the kth moment, the state variables and corresponding output variables in the prediction time domain are:

[0143]

[0144]

[0145] Where: X L is the extended state vector, A L is the extended state transfer matrix, BL is the extended control input matrix, U L is the extended control input vector, Y L is the extended output vector, x(k|k) is the state quantity predicted at time k+1 at time k;

[0146] In this embodiment, the MPC control strategy minimizes the error between the predicted output and the actual output by solving the optimal parameters, and controls the incremental smoothness constraint to avoid parameter oscillation; the controller performance is adjusted by setting the value of the weight coefficient, and the output objective function and the constraint parameters of the output objective function are defined as shown above.

[0147] The step S5 specifically includes:

[0148] S51. Define error limits within different time steps t; the expression of the error limit is as follows:

[0149]

[0150] Where: θ(t) is the error limit within time step t, y p is the predicted output, y s is the actual output, t is the current time step;

[0151] S52. For dynamic attributes, periodically obtain measured data of the dynamic attributes at each time step t and use it as the actual output; for performance attributes, predict the parameters of the performance attributes based on the time series prediction framework and use the prediction results as the actual output;

[0152] S53. Based on the actual output historical data, a maximum error limit θ′ is set; when θ(t)>θ′, a parameter update is triggered, and the parameters of the dynamic attributes and performance attributes are updated respectively, and then step S53 is performed;

[0153] S54. Iterate the digital twin model of the transmission line construction equipment and set a value strategy for the output objective functions w1 and w2 in the model to balance the values ​​of the weight coefficients w1 and w2 and modify the performance attributes; the value strategy is as follows:

[0154]

[0155]

[0156] Wherein: w1 and w2 are weight coefficients; lter is the number of iterations of the MPC control strategy; N1 is the iteration threshold; N is the maximum number of iterations; A1 is the value of w1 when the number of iterations is less than N1; A2 is the value of w2 when the number of iterations is less than N1; B1 is the value of w1 when the number of iterations is between N1 and N; B2 is the value of w2 when the number of iterations is between N1 and N.

[0157] In this embodiment, dynamic attribute parameters are regularly acquired through sensors. A periodic drive method is adopted to synchronously collect real-time measurement values ​​of dynamic parameters at a certain period. Based on the MPC control strategy, the error limit is used to determine whether to update the original parameter values ​​in the digital model.

[0158] As for performance parameters, the prediction of performance parameters is achieved through a time series prediction framework that integrates the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model. In this process, the error limit is used as the driving factor. When the error between the simulation results of the virtual entity and the measured data of the actual physical system exceeds the set limit, the performance parameters are corrected based on the MPC control strategy.

[0159] For example, the number of model iterations is set to 1000, and the values ​​of the adaptive weights w1 and w2 are adjusted to achieve the optimal combination of error minimization and incremental control smoothing.

[0160] The output objective function is divided into two parts, where: w1||λ(t)(y p (t)-y s (t))|| 2 This item is used to reflect the error between the predicted output and the actual output, and This item controls incremental smoothing.

[0161] Furthermore, the value selection strategies for w1 and w2 are defined as follows:

[0162]

[0163]

[0164] The above value selection strategy is understood as follows: when the number of iterations is less than 700, the value of w1 is kept greater than 0.85, thereby giving priority to minimizing the error, and the corresponding value of w2 is less than 0.15. The sum of the two is 1, which meets the constraint conditions. In the specific correction, when the error value is greater than the error limit θ, the value of w1 is increased to reduce the error.

[0165] Similarly, when the number of iterations is greater than 700 and less than 1000, the value of w1 is gradually reduced while the value of w2 is increased to make the control increment smoother.

[0166] See also Figure 4 In this embodiment, a construction progress simulation comparison chart is shown for a randomly selected thirty-six-hour period during the construction cycle. The chart visually compares the construction progress output by the digital twin model with that of the physical model before and after parameter correction. While the output of the digital twin model before parameter correction remains within a certain range of actual values, it exhibits significant oscillation and is of limited reference value. In the first ten hours or so after parameter correction, the digital twin model still exhibits some error compared to the actual values ​​due to the model adaptation period, but this error is significantly reduced compared to the pre-correction period. Ten hours later, the model adaptation is complete, and the error between the output and actual values ​​is within 0.02, indicating that the two values ​​are almost identical, demonstrating excellent correction results and good effectiveness.

[0167] See also Figure 5 In this embodiment, the errors of the MPC correction method, the RB unscented Kalman filtering technology, and the Bayesian method are compared. As can be seen from the figure, within 24 hours, the error function value of the RB correction method decreases from about 0.23 to about 0.14, and the error function value of the Bayesian correction method decreases from about 0.21 to about 0.13, and then the two oscillate irregularly around 0.14 and 0.13 respectively, and both show poor convergence. Compared with the former two, the error function value of the MPC-based correction method of the present invention decreases from about 0.16 to about 0.10 within 24 hours, and the convergence is significantly enhanced, and the curve converges around the 6th hour, and the convergence speed is significantly increased, due to It plays a role in controlling the incremental smoothness constraint and avoiding parameter oscillation. This is fully taken into consideration in the present invention, so the error function curve is smoother.

[0168] Example 2:

[0169] See also Figure 6 A dynamic correction system for parameters of a line equipment digital twin model based on MPC is provided, and the system is applied to the method described in Example 1. The system includes:

[0170] Digital twin model construction module 1 is used to collect historical operation data during transmission line construction and build a digital twin model of transmission line construction equipment;

[0171] Furthermore, the digital twin model of the transmission line construction equipment constructed by the digital twin model construction module 1 includes: a physical line and a digital twin line;

[0172] The physical circuit refers to: the entity reference target of digital twin circuit modeling; specifically including physical entities and IoT perception data;

[0173] The digital twin circuit refers to: the digital expression of the physical circuit in the virtual space; specifically, it includes the modeling of static properties, dynamic properties and performance properties.

[0174] System attribute division module 2, used to divide the system attributes of the digital twin model of the transmission line construction equipment into static attributes, dynamic attributes and performance attributes;

[0175] Furthermore, the attributes divided by the system attribute division module 2 are as follows:

[0176] The static attributes refer to: structural attributes of transmission line construction equipment; specifically, line length, conductor cross-sectional dimensions, tower parameters, insulator parameters, transformer capacity, supplier information of used materials, and specific information of construction equipment;

[0177] Dynamic attributes refer to dynamic indicators whose values ​​change with working conditions and reflect the real-time operating status of transmission line construction systems and equipment. Specifically, they include tower installation progress, material consumption, wind speed and direction, and equipment temperature.

[0178] The performance attributes refer to: performance indicators that reflect the operational stability of transmission line construction systems and equipment; specifically, they include maintenance records, life parameters, process parameters, work efficiency of workers or equipment, and quality control indicators.

[0179] Digital twin model optimization module 3 is used to integrate the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model in the initial digital twin model of transmission line construction equipment to build a time series prediction framework and introduce an adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework;

[0180] Furthermore, the digital twin model optimization module 3 is used to construct a framework and perform optimization according to the following method:

[0181] The introduction of the adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework specifically refers to:

[0182] For three different types of data points in historical operation data, namely stable data points, fluctuating data points and mutation data points, the memory decay factor λ(t) is introduced to optimize the output objective function to adapt to the dynamic characteristics of the data points; the details are as follows:

[0183] For stationary data points: keep the memory decay factor at a fixed constant value λ between [0, 1];

[0184] For fluctuating data points and mutation data points: introduce the correction term Δλ(t), the expression of the correction term is as follows:

[0185] λ(t)=λ(t-1)+Δλ(t);

[0186] Then for the fluctuating data points, the correction is: Δλ(t)=α(1-λ(t-1))α>0;

[0187] Then for the mutation data point, the correction is: Δλ(t)=β(1+λ(t-1))β>0;

[0188] Among them: α and β are adjustment coefficients, t is the current time step;

[0189] The optimized output objective function is as follows:

[0190]

[0191] Where: J is the output objective function, w1 and w2 are weight coefficients, y p is the predicted output, y s is the actual output, ΔU is the control increment, T is the total number of time steps, and t is the current time step;

[0192] The constraints of the output objective function are as follows:

[0193] V m_min ≤V m (t)≤V m_max

[0194] I m_min ≤I m (t)≤I m_max

[0195] T m_min ≤T m (t)≤T m_max

[0196] ΔU min ≤ΔU(t)≤ΔU max ;

[0197]

[0198] f min ≤f(t)≤f max

[0199] V min ≤V(t)≤V max

[0200] 0 <w1<1,0<w2<1;

[0201] w1+w2=1;

[0202] Where: V m , I m 、T m All are state quantities, Vm_min 、V m_max , I m_min , I m_max 、T m_min 、T m_max The upper and lower limits of its state quantity, ΔU is the increment value, ΔU min , ΔU max are the upper and lower limits of the increment value, ε is the rate of change constraint, f is the frequency, V is the voltage, f min 、f max are the upper and lower limits of the frequency, and dt is the time increment.

[0203] The expression of the multi-head self-attention mechanism is as follows:

[0204]

[0205] head i =Attention(QW i Q , KW i K , VW i V );

[0206] MultiHead(Q,K,V)=Concat(head1,...,head h )W ° ;

[0207] Among them: Attention(Q, K, V) is the output matrix of single-point attention, Q, K, V are all matrices, softmax is the activation function, d k is the key value of the input dimension, W i Q 、W i K 、W i V are the weight matrices corresponding to Q, K, and V in the i-th attention head, MultiHead(Q, K, V) is the output matrix of multi-head attention, and Concat is the concatenation operation.

[0208] The time series prediction framework constructed by the digital twin model optimization module 3 includes an encoder and a decoder;

[0209] The stack structure of the Transformer in the encoder includes a first encoding sublayer and a second encoding sublayer arranged in sequence; the first encoding sublayer includes a multi-head self-attention mechanism and an accumulation layer normalization mechanism arranged in sequence; the second encoding sublayer includes a feedforward neural network and an accumulation layer normalization mechanism arranged in sequence;

[0210] The stack structure of the Transformer in the decoder includes a first decoding sublayer, a second decoding sublayer, and a third decoding sublayer arranged in sequence; the first decoding sublayer includes a masked multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the second decoding sublayer includes a multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the third decoding sublayer includes a feedforward neural network and a cumulative sum layer normalization mechanism arranged in sequence;

[0211] The encoding second sublayer is connected to the decoding second sublayer.

[0212] Attribute update module 4, used to update static attributes in the digital twin model of transmission line construction equipment through event-driven;

[0213] Furthermore, the attribute updating module 4 is used to update the static attributes according to the following method:

[0214] Based on the changes in the construction site during transmission line construction and the corresponding changes in structural attributes, the parameters of static attributes are re-identified and re-assigned, and the assignments are updated to the digital twin model of the transmission line construction equipment.

[0215] The attribute correction module 5 is used to correct the dynamic attributes and performance attributes in the digital twin model of the transmission line construction equipment based on the MPC control strategy.

[0216] Furthermore, the attribute correction module 5 is used to perform correction according to the following steps:

[0217] S51. Define error limits within different time steps t; the expression of the error limit is as follows:

[0218]

[0219] Where: θ(t) is the error limit within time step t, y p is the predicted output, y s is the actual output, t is the current time step;

[0220] S52. For dynamic attributes, periodically obtain measured data of the dynamic attributes at each time step t and use it as the actual output; for performance attributes, predict the parameters of the performance attributes based on the time series prediction framework and use the prediction results as the actual output;

[0221] S53. Based on the actual output historical data, a maximum error limit θ′ is set; when θ(t)>θ′, a parameter update is triggered, and the parameters of the dynamic attributes and performance attributes are updated respectively, and then step S54 is performed;

[0222] S54. Iterate the digital twin model of the transmission line construction equipment and set a value strategy for the output objective functions w1 and w2 in the model to balance the values ​​of the weight coefficients w1 and w2 and modify the performance attributes; the value strategy is as follows:

[0223]

[0224] Wherein: w1 and w2 are weight coefficients; lter is the number of iterations of the MPC control strategy; N1 is the iteration threshold; N is the maximum number of iterations; A1 is the value of w1 when the number of iterations is less than N1; A2 is the value of w2 when the number of iterations is less than N1; B1 is the value of w1 when the number of iterations is between N1 and N; B2 is the value of w2 when the number of iterations is between N1 and N.

[0225] Example 3:

[0226] See also Figure 7 , a dynamic correction device for parameters of a line equipment digital twin model based on MPC, the device comprising a processor 6 and a memory 7;

[0227] The memory 7 is used to store computer program code 71 and transmit the computer program code 71 to the processor 6;

[0228] The processor 6 is used to implement the MPC-based dynamic correction method for parameters of a line equipment digital twin model according to the instruction embodiment 1 in the computer program code 71.

[0229] This embodiment also includes a computer-readable storage medium, which stores computer-executable instructions. When the computer-executable instructions are executed on a computer, the MPC-based line equipment digital twin model parameter dynamic correction method described in Example 1 is implemented.

[0230] Generally speaking, computer instructions for implementing the method of the present invention may be carried by any combination of one or more computer-readable storage media. Non-transitory computer-readable storage media may include any computer-readable media except for signals that are temporarily propagating.

[0231] Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or components, or any combination thereof. More specific examples (a non-exhaustive list) of computer-readable storage media include: an electrical connection having one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EKROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device or device.

[0232] Computer program code for performing the operations of the present invention can be written in one or more programming languages ​​or a combination thereof, including object-oriented programming languages ​​such as Java, SMalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages, in particular, Python suitable for neural network computing and platform frameworks based on TensorFlow, PyTorch, etc. can be used. The program code can be executed entirely on the user's computer, partially on the user's computer, as a separate software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or to an external computer (for example, using an Internet service provider to connect through the Internet).

[0233] The above-mentioned devices and non-temporary computer-readable storage media can be found in the detailed description of a method for dynamic parameter correction of a digital twin model of line equipment based on MPC and its beneficial effects, which will not be repeated here.

[0234] Although the embodiments of the present invention have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A method for dynamic correction of parameters of a line equipment digital twin model based on MPC, characterized in that: include: S1. Collect historical operation data during transmission line construction and build a digital twin model of transmission line construction equipment; S2. Divide the system attributes of the digital twin model of transmission line construction equipment into static attributes, dynamic attributes, and performance attributes; S3. In the digital twin model of transmission line construction equipment, the Seq2Seq algorithm is integrated with the Transformer model with a multi-head self-attention mechanism to build a time series prediction framework. An adaptive memory decay mechanism is introduced to optimize the output objective function of the time series prediction framework and control the parameter convergence speed of dynamic and performance attributes. S4. Update static attributes in the digital twin model of transmission line construction equipment through event-driven; S5. Modify the dynamic and performance attributes of the digital twin model of transmission line construction equipment based on the MPC control strategy.

2. The method for dynamic parameter correction of a line equipment digital twin model based on MPC according to claim 1, characterized in that: The step S3 introduces an adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework, specifically: For three different types of data points in historical operation data, namely stable data points, fluctuating data points and mutation data points, the memory decay factor λ(t) is introduced to optimize the output objective function to adapt to the dynamic characteristics of the data points; the details are as follows: For stationary data points: keep the memory decay factor at a fixed constant value λ between [0,1]; For fluctuating data points and mutation data points: introduce the correction term Δλ(t), the expression of the correction term is as follows: λ(t)=λ(t-1)+Δλ(t); Then for the fluctuating data points, the correction is: Δλ(t)=α(1-λ(t-1))α>0; Then for the mutation data point, the correction is: Δλ(t)=β(1+λ(t-1))β>0; Among them: α and β are adjustment coefficients, t is the current time step; The optimized output objective function is as follows: Where: J is the output objective function, w1 and w2 are weight coefficients, y p is the predicted output, y s is the actual output, ΔU is the control increment, T is the total number of time steps, and t is the current time step; The constraints of the output objective function are as follows: V m_min ≤V m (t)≤V m_max I m_min ≤I m (t)≤I m_max T m_min ≤T m (t)≤T m_max D.U. min ≤ΔU(t)≤ΔU max ; f min ≤f(t)≤f max V min ≤V(t)≤V max 0<w1<1,0<w2<1; w1+w2=1; Where: V m , I m 、T m All are state quantities, V m_min 、V m_max , I m_min , I m_max 、T m_min 、T m_max The upper and lower limits of its state quantity, ΔU is the increment value, ΔU min , ΔU max are the upper and lower limits of the increment value, ε is the rate of change constraint, f is the frequency, V is the voltage, f min 、f max are the upper and lower limits of the frequency, and dt is the time increment.

3. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The expression of the multi-head self-attention mechanism in step S3 is as follows: head i =Attention(QW i Q ,KW i K ,VW i V ); MultiHead(Q,K,V)=Concat(head1,…,head h )W o ; Among them: Attention(Q, K, V) is the output matrix of single-point attention, Q, K, V are all matrices, softmax is the activation function, d k is the key value of the input dimension, W i Q 、W i K 、W i V are the weight matrices corresponding to Q, K, and V in the i-th attention head, MultiHead(Q, K, V) is the output matrix of multi-head attention, and Concat is the concatenation operation.

4. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The step S4, updating the static attributes in the digital twin model of the transmission line construction equipment through event-driven means: based on the changes in the construction site during the transmission line construction and the changes in the structural attributes corresponding to the construction site changes, triggering the re-identification and re-assignment of the parameters of the static attributes, and updating the assigned values ​​to the digital twin model of the transmission line construction equipment.

5. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The step S5 specifically includes: S51. Define error limits within different time steps t; the expression of the error limits is as follows: Where: θ(t) is the error limit within time step t, y p is the predicted output, y s is the actual output, t is the current time step; S52. For dynamic attributes, periodically obtain measured data of the dynamic attributes at each time step t and use it as the actual output; for performance attributes, predict the parameters of the performance attributes based on the time series prediction framework and use the prediction results as the actual output; S53. Based on the actual output historical data, a maximum error limit θ′ is set; when θ(t)>θ′, a parameter update is triggered, and the parameters of the dynamic attributes and performance attributes are updated respectively, and then step S54 is performed; S54. Iterate the digital twin model of the transmission line construction equipment and set a value strategy for the output objective functions w1 and w2 in the model to balance the values ​​of the weight coefficients w1 and w2 and modify the performance attributes; the value strategy is as follows: Wherein: w1 and w2 are weight coefficients; lter is the number of iterations of the MPC control strategy; N1 is the iteration threshold; N is the maximum number of iterations; A1 is the value of w1 when the number of iterations is less than N1; A2 is the value of w2 when the number of iterations is less than N1; B1 is the value of w1 when the number of iterations is between N1 and N; B2 is the value of w2 when the number of iterations is between N1 and N.

6. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The time series prediction framework includes an encoder and a decoder; The stack structure of the Transformer in the encoder includes a first encoding sublayer and a second encoding sublayer arranged in sequence; the first encoding sublayer includes a multi-head self-attention mechanism and an accumulation layer normalization mechanism arranged in sequence; the second encoding sublayer includes a feedforward neural network and an accumulation layer normalization mechanism arranged in sequence; The stack structure of the Transformer in the decoder includes a first decoding sublayer, a second decoding sublayer, and a third decoding sublayer arranged in sequence; the first decoding sublayer includes a masked multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the second decoding sublayer includes a multi-head self-attention mechanism and a cumulative sum layer normalization mechanism arranged in sequence; the third decoding sublayer includes a feedforward neural network and a cumulative sum layer normalization mechanism arranged in sequence; The encoding second sublayer is connected to the decoding second sublayer.

7. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The digital twin model of the transmission line construction equipment includes a physical line and a digital twin line; The physical circuit refers to: the entity reference target of digital twin circuit modeling; specifically including physical entities and IoT perception data; The digital twin circuit refers to: the digital expression of the physical circuit in the virtual space; specifically, it includes the modeling of static properties, dynamic properties and performance properties.

8. The method for dynamic parameter correction of a digital twin model of line equipment based on MPC according to claim 1, characterized in that: The static attributes refer to: structural attributes of transmission line construction equipment; specifically, line length, conductor cross-sectional dimensions, tower parameters, insulator parameters, transformer capacity, supplier information of used materials, and specific information of construction equipment; Dynamic attributes refer to dynamic indicators whose values ​​change with working conditions and reflect the real-time operating status of transmission line construction systems and equipment. Specifically, they include tower installation progress, material consumption, wind speed and direction, and equipment temperature. The performance attributes refer to: performance indicators that reflect the operational stability of transmission line construction systems and equipment; specifically, they include maintenance records, life parameters, process parameters, work efficiency of workers or equipment, and quality control indicators.

9. A dynamic correction system for parameters of a digital twin model of line equipment based on MPC, characterized by: The system is applied to the method according to any one of claims 1 to 8, and the system comprises: A digital twin model construction module (1) is used to collect historical operation data during transmission line construction and construct a digital twin model of transmission line construction equipment; A system attribute division module (2) is used to divide the system attributes of the digital twin model of the transmission line construction equipment into static attributes, dynamic attributes and performance attributes; The digital twin model optimization module (3) is used to integrate the Seq2Seq algorithm and the multi-head self-attention mechanism Transformer model in the initial digital twin model of the transmission line construction equipment to build a time series prediction framework, and introduce an adaptive memory decay mechanism to optimize the output objective function of the time series prediction framework; An attribute update module (4) is used to update static attributes in the digital twin model of the transmission line construction equipment through event-driven operation; The attribute correction module (5) is used to correct the dynamic attributes and performance attributes in the digital twin model of the transmission line construction equipment based on the MPC control strategy.

10. A dynamic correction device for parameters of a digital twin model of line equipment based on MPC, characterized by: The device comprises a processor (6) and a memory (7); The memory (7) is used to store computer program code (71) and transmit the computer program code (71) to the processor (6); The processor (6) is used to execute the MPC-based line equipment digital twin model parameter dynamic correction method according to any one of claims 1 to 8 according to the instructions in the computer program code (71).

Images