52 results about "Vector fitting" patented technology

The invention belongs to the field of filter optimization of coupling matrixes and neural networks, and discloses a filter optimization design method of a neural network based on guidance of a coupling matrix, and physical elements and mechanisms are introduced into the neural network. The method comprises the following steps: firstly, determining filter optimization indexes, randomly generating sample data in a range near the size of the optimization indexes, and screening data samples by adopting a sample screening method to obtain high-quality samples; then, extracting a coupling matrix from the obtained S parameter according to a vector fitting algorithm, and forming new coupling information M'by combining the center frequency f0 and the bandwidth BW; a full-connection neural network is used as a medium of filter device size parameters and output frequency response, the size of a to-be-optimized device of the filter is input, and coupling information M'representing S parameters is output. And the forward simulator is connected with a genetic optimization algorithm GA to realize optimization design of a specific index parameter filter. According to the method, the optimization design time is effectively saved, the optimization efficiency is improved, and the optimization dimension is reduced.

The invention discloses a new energy equipment sequence impedance multi-level frequency domain identification method, and relates to the field of power electronic equipment parameter and model frequency domain identification. Complex vector fitting is carried out on frequency domain response data of impedance measurement, an analytical model or an equivalent zero-pole model of power electronic equipment is obtained, and a basis is provided for grid-connected resonance stability analysis and impedance remodeling; comprising the following steps: S1, carrying out sequence impedance frequency scanning of single harmonic disturbance injection to obtain impedance frequency domain response data; s2, performing complex vector fitting to obtain a pole-residue type port frequency domain model; s3, evaluating'grey / black box 'characteristics according to the transparency of the control architecture information; s4, deducing a theoretical sequence impedance model of the grey box equipment, and identifying parameters based on a pole-residue expression; deducing an equivalent zero pole model for the black box equipment or the new energy cluster; compared with traditional qualitative analysis, accurate quantitative analysis and impedance remodeling are achieved, and analysis accuracy and engineering applicability are improved.

The invention discloses a high-voltage direct-current cable electromagnetic transient frequency domain modeling and time domain solving method, system, medium and equipment, and the method comprises the steps: obtaining an initialization parameter of a high-voltage direct-current cable, calculating an internal impedance matrix [Zi] of a multi-layer cross-section cable body during frequency calculation, and calculating an admittance coefficient matrix of the multi-layer cross-section cable body during frequency calculation; calculating ground loop impedance [] and potential coefficient [] of the air laying cable or the direct burial laying cable, and obtaining a total impedance matrix [] and a total admittance matrix [] of the cable at the time of frequency: importing obtained multi-frequency [] and [] calculation results into a PSCAD TLine module, and realizing time domain solution of frequency domain calculation by using a vector fitting model.

The invention relates to a power transformer partial discharge source positioning method based on a learning vector quantization network model, and belongs to the technical field of power equipment detection, and the method comprises the following steps: S1, measuring voltage frequency responses of a transformer winding at different tapping positions; s2, obtaining an approximate fitting voltage transfer function through a vector fitting VF technology; s3, establishing a transformer segment winding transfer function SWTF library; s4, calculating a partial discharge PD reference signal; s5, constructing a learning vector quantization LVQ network model, and taking the marked partial discharge reference signals as a training set to train the LVQ network model; s6, injecting a partial discharge test pulse into the randomly selected winding tap connection, and taking the obtained response as a partial discharge test signal; and S7, inputting a partial discharge test signal into the trained LVQ network model, and testing the classification performance of the model.

The invention discloses a new energy system frequency domain modeling and equivalent circuit order reduction analysis method, and relates to the technical field of new energy power system modeling and stability analysis. The method is used for solving the problems that in an existing frequency domain modeling means, the automation degree of data acquisition is low, fitting precision and stability are poor, equivalent circuit universality is poor, and order reduction analysis is not systematic. The method comprises the following steps: acquiring system frequency domain response data based on disturbance injection; constructing a system rational function model by adopting an improved vector fitting algorithm; and then model order reduction is realized through state space modeling and an SVD-based balanced truncation method, and response consistency and stability analysis is carried out in a frequency domain and a time domain. The method can be widely applied to modeling analysis of new energy equipment such as wind power equipment, voltage source converters and high-voltage direct-current power transmission equipment, and has the advantages of being high in automation degree, high in modeling precision, good in visualization, clear in structure, suitable for simulation and control design and the like.

A calculation method for a transformer winding considering a frequency-dependent loss, comprising: according to simulation requirements, defining key parameters calculated by using a finite-difference time-domain (FDTD) algorithm; constructing an equivalent flat wire model in an FDTD grid; calculating frequency-dependent internal impedance per unit length of a flat wire; and on the basis of a vector fitting method and a time-domain convolution calculation method, substituting the frequency-dependent internal impedance into a time-domain equation of the FDTD algorithm for updating and iterative solving, so as to obtain an electromagnetic field calculation result. In the present invention, on the basis of the vector fitting method and time-domain convolution calculation, the impact of the frequency-dependent characteristics of an electromagnetic field distributed in a conductor on an electromagnetic transient process is accurately considered during the FDTD time-domain solving, so that the calculation accuracy and efficiency of conventional FDTD algorithms and even conventional cross-scale modeling techniques can be significantly improved, thereby carrying out the targeted electromagnetic transient analysis on the reliability, safety, and predictability of transformers, and achieving the early detection of potential faults and defects, or formulating targeted maintenance plans and repair schemes.

The invention discloses a grid-connected stability analysis method for a system containing a black box converter, and belongs to the technical field of power system stability analysis. The method comprises the following steps: injecting multi-frequency-point voltage disturbance during grid-connected operation of a black box converter, and measuring current response to obtain admittance data; fitting into a frequency domain transfer function by using a vector fitting algorithm; then constructing a system admittance matrix without a black box converter based on a power system network topology, and adding a transfer function to obtain a whole system admittance matrix; and finally, solving the characteristic root by adopting an s-domain modal analysis method to judge the stability of the system, and if the system is not stable, analyzing the oscillation frequency, the participation factor and the element sensitivity, and determining a resonance suppression strategy. According to the method, internal parameters of equipment are not needed, a complete stability analysis process is provided, the resonance source can be accurately positioned, an optimization strategy is provided, and the method can be used for rapid stability evaluation of various new energy systems.

The invention discloses an unmanned aerial vehicle group scheduling method and system based on a multi-layer recurrent neural network, and relates to the technical field of unmanned aerial vehicle scheduling, and the method comprises the following steps: obtaining a state change sequence of offline unmanned aerial vehicles in a preset time window before offline, and obtaining an initial state vector; when re-online is detected, extracting a first state vector and performing vector fitting matching on the first state vector and the initial state vector; constructing a state fluctuation trajectory, evaluating changes of the state fluctuation trajectory under different task time scales, obtaining state fluctuation gradient distribution, converting the state fluctuation gradient distribution into a continuous state vector sequence, performing vector fitting truth compensation on the continuous state vector sequence and a preset reference state vector cluster, and identifying a weak oscillation region according to a truth compensation result; the state vector of the weak oscillation area is converted into a correction mask, the offline unmanned aerial vehicle input state vector of the multi-layer recurrent neural network is corrected and used for unmanned aerial vehicle group scheduling prediction, and the problem that state information is lagged or abnormal due to the fact that the unmanned aerial vehicles are online again after offline is solved.

The invention relates to the technical field of artificial intelligence analysis and optimization of a power system, and discloses a line loss variable separation method, system and device based on dynamic time sequence vector fitting, and a medium, and the method comprises the steps: fusing multi-source heterogeneous data, and constructing an enhanced feature vector sequence through an attention mechanism and a physical equation constraint; establishing a time-varying causal graph model based on a dynamic Bayesian network, introducing an attention mechanism to calculate time sequence feature similarity, calculating time-varying causal strength in combination with physical constraint loss and topological distance, and deducing dynamic causal; a theoretical line loss curve is dynamically fitted in a physical-data double-track modeling mode, and key nodes in a causal graph model are used as correction factors; the contribution degree of each factor to the line loss difference is quantified based on a Shapley value algorithm, and accurate separation and attribution of line loss components are realized.

The invention relates to a rail train part anomaly detection method based on a point-surface distance, and relates to the technical field of rail part anomaly detection, and the method comprises the steps: a surface recognition step: recognizing an actually measured special surface in a target image through a surface recognition strategy, and generating a corresponding special recognition feature; a model indexing step: calling a target part model from the part type database according to the special identification features and importing the target part model into a space coordinate system; a vector fitting step: determining a presentation vector of the target part model in the space coordinate system through a model presentation strategy, and constructing a coordinate mapping relation between the space coordinate system and the target image; a point confirmation step: determining the graphic plane coordinates of the reference special point in the target image according to the mapping relation; and a deviation calculation step: calculating a difference value between the coordinates of the reference special point and the actually measured special point to obtain a special point deviation, and solving a reference distance deviation through a preset distance solving algorithm. The method has the effect of conveniently detecting the abnormal vector change generated by the rail train parts.

The invention relates to the technical field of network data processing, in particular to a self-evolution online learning personalized recommendation method, which comprises the following steps of: acquiring behavior data of a learner during online learning in real time; constructing a simplified knowledge graph of the knowledge system of each subject by adopting a hierarchical decoupling method; a behavior continuity merging algorithm is adopted to obtain learning behaviors of the learner at each knowledge point based on behavior data mapping; fitting by adopting a time sequence capability vector fitting algorithm based on the behavior data to obtain the learning capability of the learner at each knowledge point; periodically determining a first recommended learning mode of each knowledge point based on the learning behavior and learning ability of the group learner at each knowledge point; the first recommendation learning mode of each knowledge point in the current period is corrected based on the learning behavior and learning ability of an individual learner to serve as a personalized recommendation result, so that'group efficiency + individual adaptation 'dual-drive recommendation is realized, meanwhile, dynamic standard correction and real-time response can be realized, and the consistency and accuracy of recommendation logic are improved.

The application discloses a macro modeling method for ensuring direct current precision, and the method comprises the following steps: reading n-port network parameters; initializing sampling points for constructing a macro model; initializing the order of the macro model; initializing the pole of the macro model; establishing a weighted linear system according to the current order, the sampling points and the pole; solving a linear equation set, judging convergence according to an error, and obtaining the macro model; and outputting the optimized macro model by applying a residue matrix disturbance to optimize the macro model. The macro modeling method for ensuring direct current precision provided by the application controls the direct current precision in two stages. In the first stage, a weighted least square method is adopted on the basis of a linear system established by a traditional vector fitting algorithm to preliminarily control the direct current precision. In the second stage, a small disturbance is applied to the residue matrix of the macro model, so that the disturbed macro model has accurate direct current characteristics.

The invention provides a parametric modeling method and device for a silicon-based millimeter wave on-chip component, and belongs to the technical field of integrated circuit modeling and millimeter wave circuit design. The method comprises the following steps: under a preset key parameter interpolation node, acquiring a transmission parameter of a de-embedding test structural member comprising a de-embedded component at each test frequency point; removing a parasitic effect embedded in the test structural member by using transmission matrix operation so as to obtain a transmission matrix of the de-embedded component and obtain a scattering parameter of the de-embedded component; then, through a vector fitting algorithm, obtaining a pole-residue model of the de-embedded component under the parameter interpolation node; and based on the pole-residue model under each parameter interpolation node, establishing a parameterized de-embedding model of the de-embedded component through a gravity center interpolation algorithm. According to the method, rapid and accurate modeling of silicon-based millimeter wave on-chip components of any size can be realized, and the reliability and efficiency of millimeter wave integrated circuit design are improved.

The invention relates to a refined frequency-dependent mutual impedance three-core power cable high-frequency transient model, and belongs to the field of power disturbance analysis of power systems. The model is composed of a plurality of distributed cable units (ReCABLE units), each ReCABLE unit represents a three-core power cable section with the unit length of 10, a plurality of three-core power cable sections with the unit length of 10 are built into a complete three-core power cable, and the distributed cable units with the corresponding number are selected for the complete three-core power cable according to the total length of the complete three-core power cable; and each ReCABLE unit comprises three refined frequency-varying mutual impedance modules Mcs (f), three frequency-varying admittance modules Yin (f), three frequency-varying wire core impedance modules Zc (f) and one frequency-varying metal sheath impedance module Zs (f). According to the method, a high-precision model is established through a vector fitting-circuit synthesis-passive optimization method based on mutual impedance frequency change data calculated through experimental measurement, so that the defect of an existing cable model in mutual impedance parameter simulation precision is effectively overcome; and a more reliable model basis is provided for transient analysis research and engineering application of the three-core power cable.

The present invention relates to a purifier turbulence fitting system, an intelligent state assessment system, and an intelligent control system based on an adversarial neural network. The intelligent control system includes a cloud-based purifier turbulence vector pre-training model, a purifier multi-point flow velocity sensor, a neural network turbulence vector correction module, a turbulence vector pre-processing model, a neural network turbulence vector single time slice generation module, a discriminative neural network turbulence vector single time slice assessment module, a neural network turbulence vector multi-time slice generation module, a discriminative neural network turbulence vector multi-time slice assessment module, a neural network turbulence vector fitting flow direction module, a neural network purifier operation status assessment module, and an intelligent control system. The present invention uses joint detection of multiple detection points and the purifier pre-training model to fit and iteratively correct the purifier turbulence output, thereby achieving accurate control of the purifier.

The invention discloses a neural network electromagnetic performance prediction method with causality and passivity constraints, and the method comprises the steps: obtaining geometric design variables of a TSV structure to be detected, inputting the geometric design variables into a trained electromagnetic performance prediction model, and obtaining the prediction values of a plurality of scattering coefficients of the TSV structure to be detected; wherein the electromagnetic performance prediction model comprises a polar point sub-network, a residue sub-network and a prediction sub-network, and the polar point sub-network and the residue sub-network are respectively used for determining a to-be-detected polar point vector and a to-be-detected residue vector corresponding to each scattering coefficient according to geometric design variables of a to-be-detected TSV structure; the prediction sub-network is used for determining a prediction value of a scattering coefficient of a corresponding type according to a to-be-measured pole vector and a to-be-measured residue vector; the electromagnetic performance prediction model is trained through a sample pole vector and a sample residue vector, and the sample pole vector and the sample residue vector are obtained by processing a sample scattering coefficient based on a vector fitting technology. The method is low in implementation cost and good in adaptability.

The invention provides a multi-port system self-adaptive macro modeling method based on a Roca matrix and vector fitting, and the method comprises the following steps: obtaining S parameters of a multi-port system under each frequency point, and selecting an initial frequency point and a corresponding S parameter matrix; after the frequency is converted to an S domain, initial vector fitting is carried out based on a Loa matrix, and a system initial rational function expression is generated; adopting a dichotomy for interpolation in the frequency band with the maximum change of the rational function, inserting a new frequency point in the midpoint of the frequency band, and carrying out the next vector fitting based on the Roca matrix; and quantizing the sum of the port errors of each frequency band by using Gaussian integral, sorting according to the size, carrying out dichotomy interpolation again in the frequency band with the larger error, and iteratively executing the process until the fitting result reaches the convergence condition, thereby finally obtaining the macro modeling of the multi-port system. On the basis of the interpolation method and the Rocner vector fitting algorithm, the calculation memory requirement is remarkably reduced and the solving efficiency is improved while the solving precision is ensured.

The invention discloses a macro modeling method for ensuring direct current precision. The method comprises the following steps: reading n-port network parameters; initializing sampling points used for constructing a macro model; initializing the order of the macro model; initializing a pole of the macro model; establishing a weighted linear system according to the current order, the sampling point and the pole; solving the system of linear equations, judging convergence according to errors, and obtaining a macro model; the macro model is optimized by applying residue matrix disturbance, and the optimized macro model is output. According to the macro modeling method for guaranteeing the direct current precision, the direct current precision is controlled in two stages, in the first stage, a weighted least square method is adopted on the basis of establishing a linear system through a traditional vector fitting algorithm, and the direct current precision is preliminarily controlled; and in the second stage, tiny disturbance is applied to the residue matrix of the macro model, so that the disturbed macro model has accurate direct current characteristics.

The present application relates to the technical field of network data processing, in particular to a self-evolving online learning personalized recommendation method, comprising: collecting behavior data of learners in real time during online learning; constructing a simplified knowledge graph of each subject knowledge system by using a hierarchical decoupling method; obtaining the learning behavior of learners at each knowledge point based on behavior data mapping by using a behavior continuity merging algorithm; fitting the learning ability of learners at each knowledge point based on behavior data by using a time sequence ability vector fitting algorithm; periodically determining the first recommended learning mode of each knowledge point based on the learning behavior and learning ability of group learners at each knowledge point; and correcting the first recommended learning mode of each knowledge point in the current period based on the learning behavior and learning ability of individual learners to serve as a personalized recommendation result, realizing double-driven recommendation of "group effectiveness + individual adaptation", and simultaneously realizing dynamic standard correction and real-time response, improving the logical coherence and accuracy of the recommendation.

A method for zero-point correction of an S-parameter model includes: calculating the S-parameters at zero frequency based on the constructed S-parameter model; calculating the error matrix between the S-parameter model and the original S-parameters at zero frequency; calculating the residues corresponding to the preset real poles based on preset real poles and the error matrix; and adding a term to the S-parameter model using the preset real poles and their corresponding residues to obtain the zero-point corrected S-parameter model. This method further corrects the S-parameters after vector fitting and passivity correction, ensuring that the values ​​of the S-parameters at zero frequency remain consistent with the original S-parameters while minimizing disruption of the original model's passivity and maintaining accuracy. This corrects the error of the original model's S-parameters at zero frequency, eliminating errors in subsequent simulations caused by the inaccuracy of the S-parameters at zero frequency.

The invention discloses an S parameter time domain modeling method, device, medium and system based on vector fitting, and belongs to the field of microwave radio frequency device modeling and emulation.The method comprises the steps that a frequency domain response function is constructed based on a device S parameter, a residue and a pole value are approximated step by step through iterative optimization and a least square solution method, and the frequency domain response function is obtained; and finally, carrying out time domain convolution on the input signal based on the transfer function, and carrying out conversion of the output signal through the time domain convolution. According to the method, the simulation efficiency of the broadband microwave array system link time domain signal flow is improved.

The invention provides an MEMS modeling method combining vector fitting with a neural network transfer function, and relates to the technical field of micro electro mechanical systems. According to the embodiment of the invention, the fitting parameters corresponding to the performance curves of the MEMS devices with different geometric sizes and material characteristics can be predicted on the basis of vector fitting in combination with an artificial neural network transfer function algorithm, so that the performance curves of the MEMS devices with different geometric sizes and material characteristics are obtained; the modeling process of the MEMS device is quickly and accurately simulated and analyzed, and an MEMS designer can be guided to optimize the device.

The invention relates to a mechanism-data fusion new energy grid-connected system small interference stability analysis method. The method comprises the following steps: firstly, obtaining operation data of a new energy unit, and carrying out normalization preprocessing; then, a new energy unit impedance characteristic neural network model is constructed and trained, and a new energy field station impedance identification model is synthesized based on the trained unit impedance model in combination with the field station topology and the line parameters; then, on the basis of the new energy station impedance identification model, a vector fitting method and a PySR algorithm are adopted to construct an optimal fitting transfer function of the new energy station; then, determining a discrete state space equation based on the optimal fitting transfer function of the new energy station; and finally, calculating a discrete eigenvalue and a continuous eigenvalue of the system based on the discrete state space equation, and analyzing the stability of the system based on the discrete eigenvalue and the continuous eigenvalue. The online, continuous and accurate dynamic description of the impedance of the new energy unit and the dynamic stability prediction and analysis of a complex system are realized.

The invention discloses a new energy station broadband oscillation online early warning method based on a virtual split ratio, and belongs to the technical field of power system stability analysis and control. The method comprises the following steps: injecting a harmonic wave small disturbance signal into a current or voltage sampling value of a new energy converter, synchronously recording a static var generator (SVG) branch current waveform and extracting a disturbance component, and calculating a virtual shunt ratio; obtaining wide frequency band data through frequency sweeping, and obtaining a domain expression of a virtual split ratio through vector fitting; and solving the zero point of the expression to obtain a station characteristic root, and carrying out oscillation risk judgment and early warning according to the position of the station characteristic root on the complex plane. According to the method, control parameters and line parameters of the new energy converter in the station do not need to be obtained, a high-power disturbance source is not needed, the problems that an existing method is high in parameter dependence degree, modeling is difficult, and the calculated amount is large are solved, and rapid online early warning of the broadband oscillation risk in the full black box state of the new energy station is achieved.

The invention discloses an improved vector fitting algorithm based on hierarchical residual optimization and a storage medium, and belongs to the technical field of integrated circuit signal integrity analysis. In order to solve the problems of low modeling precision and poor convergence when the existing vector fitting algorithm is used for processing a multi-delay, high-frequency noise and multi-port network, the invention provides a layered residual optimization mechanism which comprises the following steps of: firstly, extracting a main delay component through time domain delay estimation and generating a de-delay smooth signal; dynamically adjusting poles by adopting an adaptive vector fitting algorithm to finish local modeling; reconstructing the total model through multi-stage iteration and updating the residual error until the error reaches the standard; and finally, mapping the pole parameters into a hybrid topology SPICE circuit comprising a transmission line and a lumped network. According to the method, the high-frequency signal fitting precision is remarkably improved through hierarchical decomposition, the multi-delay scene calculation complexity is effectively reduced, the generated equivalent circuit is compatible with multi-port S parameters and keeps passivity, and the simulation efficiency of a high-speed circuit can be improved.

The application relates to a passive property analysis method, device, equipment and medium in an S parameter modeling process. The method comprises the following steps: acquiring pole points and residue matrixes of S parameter vector fitting, each pole point corresponding to a residue matrix; constructing a new S parameter rational fraction based on the pole points and the residue matrixes; constructing a judgment matrix based on the pole points and the residue matrixes; constructing a plurality of detection frequency points according to the judgment matrix; and determining a passive abnormal frequency point based on the new S parameter rational fraction and the plurality of detection frequency points. When the passive property analysis is performed in the S parameter modeling process, the pole points and the residue matrixes of the S parameter vector fitting are considered to determine the passive abnormal frequency point, the frequency at which the passive property of the S parameter will be abnormal can be quickly determined, the violation degree of the frequency point in which the passive property is seriously violated is given, the high-cost calculation process of an original Hamiltonian is avoided, the calculation efficiency is improved, and the pressure of subsequent passive property correction is reduced.

The application discloses a new energy station wide frequency oscillation online early warning method based on a virtual shunt ratio, and belongs to the technical field of power system stability analysis and control. The method comprises the following steps: injecting a harmonic small disturbance signal into a current or voltage sampling value of a new energy converter, synchronously recording a static var generator (SVG) branch current waveform and extracting a disturbance component, and calculating a virtual shunt ratio; obtaining wide frequency band data through frequency sweeping and obtaining a domain expression of the virtual shunt ratio by using vector fitting; solving the zero point of the expression to obtain station characteristic roots, and judging and warning an oscillation risk according to the position of the characteristic roots on a complex plane. The application does not need to obtain control parameters and line parameters of the new energy converter in the station, and does not need a high-power disturbance source, thereby solving the problems of high parameter dependency, difficult modeling and large calculation amount of the existing method, and realizing rapid online early warning of the wide frequency oscillation risk in the full "black box" state of the new energy station.

The invention discloses an AC / DC power transmission line broadband equivalent analysis modeling method and system, and belongs to the technical field of power system AC / DC power transmission, and the method comprises the steps: obtaining the self-impedance and mutual impedance frequency change parameters of each phase based on a phase domain frequency change model, and converting the self-impedance and mutual impedance into differential mode impedance and common mode impedance; vector fitting is carried out on the impedance data to construct an RL branch parameter solving equation, the cascade number is determined, and an RLC circuit equivalent model corresponding to differential mode impedance and common mode impedance is obtained; a topological form of a differential mode RLC equivalent circuit corresponding to differential mode impedance and a compensation current source is adopted to construct a power transmission line equivalent circuit considering mutual inductance. According to the long-distance AC / DC submarine cable RLC equivalent modeling method, an AC / DC cable port voltage and current relation considering mutual inductance characteristics is deduced based on phase-mode transformation, a cable mutual inductance equivalent circuit based on a differential / common mode RLC circuit and a compensation current source is constructed, and the long-distance AC / DC submarine cable RLC equivalent modeling method capable of accurately reflecting frequency change, time delay, distribution characteristics and mutual inductance effects is provided.

The present invention discloses a time prediction model training method, a time prediction method and a device. After extracting the first route feature vector of a route feature set, the first route feature vector is dimensionally converted to obtain a second route feature vector. Time prediction is performed based on the first route feature vector and the second route feature vector to obtain an estimated arrival time, and the second route feature vector is vector-fitted to obtain a route reconstruction feature vector. Then, a time prediction loss value is obtained based on the estimated arrival time and the actual arrival time, and a reconstruction loss value is obtained based on the route reconstruction feature vector and all road section feature information. Then, the parameters of the time prediction model are corrected based on the time prediction loss value and the reconstruction loss value. The embodiment of the present invention can improve the prediction accuracy of the estimated arrival time. The present invention can be widely used in information processing technologies in various scenarios such as cloud technology, artificial intelligence, smart transportation, and assisted driving.

The application provides a network parameter representation method based on a multi-path convolutional autoencoder. Network parameters are divided into network parameter real parts and network parameter imaginary parts. Each part is processed through N+1 paths, the first of which is a low-frequency path, and the remaining N paths are high-frequency compensation paths, each path processing different information of the network parameters. The network parameter representation method can avoid the residual-pole mismatch problem existing in the vector fitting method, has higher compression efficiency and higher accuracy, and can be used to establish a neural network proxy model and accelerate the design of microwave devices.