Adaptive optimization method for multi-scale control variable collaborative inversion

Abstract

The invention provides a self-adaptive optimization method for multi-scale control variable collaborative inversion, and belongs to the technical field of ocean internal tide, and the method comprises the steps: taking a preset open boundary condition and a bottom friction coefficient as model input parameters, and driving a forward internal tide model to operate; optimizing a control variable at an independent point by assimilating an observation data set; performing nonlinear fitting on spatial distribution of an open boundary condition by adopting a cubic spline interpolation technology, and performing multi-dimensional smooth optimization on a field structure of a bottom friction coefficient by adopting a curved surface spline interpolation method; and continuously optimizing the independent point values of the open boundary condition and the bottom friction coefficient until the deviation index between the simulation result and the observation data set falls into a preset error threshold range. According to the method, the problems that the nonlinear physical characteristics of the actual ocean cannot be described and the high-precision simulation of the internal tide cannot be realized during the optimization of the control variables of the internal tide model at present are solved.

Description

Technical Field

[0001] The present invention belongs to the technical field of internal tides in the ocean, and particularly relates to an adaptive optimization method for collaborative inversion of multi-scale control variables. Background Art

[0002] In the dynamics of ocean internal tides, a high-precision numerical model is a key tool for revealing the mechanisms of internal tide generation, propagation, and energy dissipation. The simulation accuracy of the internal tide model highly depends on the accuracy of the control variables. Among them, the open boundary conditions are the boundary inputs based on tidal levels and flow velocities, and the bottom friction coefficient is based on reflecting the resistance of the seabed topography to tidal flow. These are the core parameters affecting the model performance. However, the existing internal tide models face the following problems in optimizing the control variables: Most of the models in the related technologies use linear interpolation or empirical formulas to set the control variables, which are difficult to accurately describe the tidal wave deformation and energy dispersion of internal tide fluctuations in the actual ocean, resulting in a significant deviation between the simulation results and the measured data.

[0003] The optimization of the open boundary conditions and the bottom friction coefficient is usually processed independently, lacking the collaborative modeling of their spatial correlation and scale effect.

[0004] When the inversion algorithms in the related technologies process high-dimensional control variables, they are prone to falling into local optimal solutions, and there is a lack of constraints on spatial continuity during the optimization process, resulting in a slow convergence of the simulation error. Summary of the Invention

[0005] The present invention provides an adaptive optimization method for collaborative inversion of multi-scale control variables, aiming to solve the problems existing in the optimization of the control variables of the current internal tide model, such as the inability to depict the non-linear physical characteristics actually existing in the real ocean and the inability to achieve high-precision simulation of internal tides.

[0006] The method includes: S101. Based on the preset open boundary conditions and bottom friction coefficient as the model input parameters, drive the forward internal tide model to run, and extract the surface flow velocity data from the preset observation points as the observation data set; S102. Set the initial parameter values based on the open boundary conditions and the bottom friction coefficient, and optimize the control variables at the independent points by assimilating the observation data set: use the cubic spline interpolation technique to perform non-linear fitting on the spatial distribution of the open boundary conditions, and use the surface spline interpolation method to perform multi-dimensional smoothing optimization on the field structure of the bottom friction coefficient; S103. Loop through S101 and S102, adjust the model input parameters, and continuously optimize the independent point values of the open boundary conditions and the bottom friction coefficient until the deviation index between the simulation result and the observation data set falls within the preset error threshold range.

[0007] Further, it should be noted that the specific steps of S101 are as follows: perform forward simulation based on a preset internal tide adjoint assimilation model, use the given open boundary conditions and bottom friction coefficient as model input parameters, run the forward model to obtain simulation results, and extract surface velocity data from preset observation points as the observation data set.

[0008] Further, it should be noted that the specific steps of S102 are as follows: set the initial values of the control variables, the initial value of the open boundary condition is 0, and the initial value of the bottom friction coefficient is 0.002. Based on the observation data set obtained in S101, optimize the control variables at the independent points: Optimize the spatial distribution of the open boundary conditions using the cubic spline interpolation method; optimize the multi-dimensional field structure of the bottom friction coefficient using the surface spline interpolation method.

[0009] Further, it should be noted that the specific steps of optimizing and inverting the open boundary conditions using cubic spline interpolation are as follows: select preset points as independent points among all the open boundary points in the calculation area, and use the optimization algorithm to evaluate the values; The values of the remaining grid points are obtained by cubic spline interpolation of the values of the independent points. Let N be the total number of open boundary points, be the value of the open boundary condition at the independent point, be the number of independent points, be the result obtained by through cubic spline interpolation, that is (1) where are the cubic spline interpolation coefficients.

[0010] Further, it should be noted that in the method, define the cubic spline interpolation function existing on the interval and use to represent the derivative value of at the point Then, on each small interval (2).

[0011] Further, it should be noted that the expression of on the interval (3) where , the second derivative of can be obtained as The expression is as follows: (4).

[0012] It should be further noted that the open boundary conditions are as follows: The first type of open boundary condition is that the tangent slopes of the curve at the two endpoints and are known, that is, and are known; The second type of open boundary condition: the second derivative of the function at the two endpoints and is known, that is, and are known; Periodic boundary condition: the function is a periodic function, and the spline function should also be a periodic function, that is, it satisfies the conditions and at the endpoints.

[0013] It should be further noted that the optimization inversion of the bottom friction coefficient by using surface spline interpolation in S102 specifically includes: Select the values of some grid points in the bottom friction coefficient field as independent points, and let be the value of the bottom friction coefficient at the l th independent point. The values of each grid point in the bottom friction coefficient field are obtained through surface spline interpolation, and satisfy the following relationship: (13) where is the surface spline interpolation coefficient.

[0014] It should be further noted that in the method, the expression form of the surface spline interpolation coefficient is extracted: (18) The cost function J is obtained with respect to the gradient of the bottom friction coefficient at the independent points: (19); By transferring the gradient of the cost function to the bottom friction coefficient field to the independent points, clarifying the influence of interpolation and distance on the gradient, and updating the parameters by the gradient, combined with the smoothness of the surface spline, reducing the simulation and observation errors, and finally achieving the goal of data assimilation.

[0015] It can be seen from the above technical solutions that the present invention has the following advantages: The adaptive optimization method for multi-scale control variable collaborative inversion provided by this application uses open boundary conditions and bottom friction coefficients as the key parameters to drive the model. Aiming at the defects of independent processing and insufficient coupling of control variables in the prior art, the overall characterization ability of the model for the physical process of internal tides is improved through multi-variable collaborative driving. Taking the simulated surface velocity as the observed data, a direct correlation between the model output and the actual ocean observation is established to solve the problem of large deviation between the simulation result and the measured value caused by the broken data assimilation link of the traditional model. Through piecewise smooth functions and second derivative continuity constraints, the curve mutation and high-frequency noise caused by traditional linear interpolation are avoided, and the non-linear spatial distribution of tidal fluctuations is more realistically reflected. By cyclically assimilating the observed data, the open boundary conditions and the bottom friction coefficients are synchronously adjusted to realize the coupling of their spatial distributions and physical effects. The optimized control variable field can be directly used for ocean engineering design to avoid the design risks caused by traditional empirical parameters. Description of the Drawings

[0016] In order to more clearly illustrate the technical solutions of the present invention, the drawings required for description will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0017] Figure 1 It is a flowchart of the adaptive optimization method for multi-scale control variable collaborative inversion; Figure 2 It is a spatial distribution diagram of the given open boundary conditions and the inversion results of two interpolation schemes. Detailed Embodiments

[0018] The adaptive optimization method for multi-scale control variable collaborative inversion provided by the present invention optimizes the model parameters using the independent point scheme. Some specific points are selected as independent points in the calculation area, and the spatial distribution of the control variables is obtained by spline interpolation from the values of these independent points. Different methods are used for different parameters in the model. For example, the cubic spline interpolation method is selected for the optimization of the open boundary conditions, and the surface spline interpolation method is selected for the optimization of the bottom friction coefficient. It solves the problems existing in the optimization of the control variables of the current internal tide model, such as the inability to characterize the non-linear physical characteristics existing in the actual ocean and the inability to achieve high-precision simulation of internal tides.

[0019] The following will detail the specific steps of the adaptive optimization method for multi-scale control variable collaborative inversion involved in this application. For the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are proposed to thoroughly understand the embodiments of this application. However, those skilled in the art should clearly understand that this application can also be implemented in other embodiments without these specific details.

[0020] It should be understood that when used in the specification of the present application, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their combinations. The terms "comprising", "including", "having" and their variants all mean "including but not limited to", unless otherwise specifically emphasized in other ways.

[0021] Statements such as "an embodiment" or "some embodiments" described in the present application mean that the specific features, structures or characteristics described in the embodiment are included in one or more embodiments of the present application. Thus, statements such as "in one embodiment", "in some embodiments", "in other some embodiments", "in still other embodiments" and the like that appear in different places in the present application do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways.

[0022] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0023] Please refer to Figure 1 The figure shows a flowchart of an adaptive optimization method for multi-scale control variable collaborative inversion in a specific embodiment. The method includes: S101: Drive the forward model through the given control variables. The control variables are the open boundary conditions and the bottom friction coefficient, and the surface velocity generated by the simulation results at the given observation points is used as the observation data.

[0024] In this embodiment, the forward model uses the control variables as the driving conditions. The open boundary conditions specify the input conditions of physical quantities such as tides and water flows at the edge of the model area. For example, the tidal level changes, flow velocity magnitude and direction at the boundary of the model sea area are set; the bottom friction coefficient reflects the frictional resistance of the seabed to the water flow, and its value affects the degree of velocity attenuation of the water flow near the seabed. After the forward model runs, the surface velocity data is recorded at the pre-set observation point positions.

[0025] It is understandable that a forward model is constructed based on relevant theories such as hydrodynamics and ocean dynamics, and the Navier-Stokes equation, continuity equation, etc. are solved by the finite difference method or the finite element method. Under the given control variable conditions, the internal tidal movement process of the ocean is simulated, and then the surface velocity at the observation point is obtained. In this way, the input-output relationship of the simulation process is clarified, and the observed data provides a reliable comparison basis for subsequent optimization, ensuring that the optimization direction is centered around the actual ocean phenomena and making the model simulation results closer to the real situation.

[0026] S102: Assign initial values to the control variables, optimize the control variables at the independent points by assimilating the observed data, use cubic spline interpolation to optimize and invert the open boundary conditions, and use surface spline interpolation to optimize and invert the bottom friction coefficient.

[0027] In this embodiment, initial values are assigned to the control variables, which can be determined according to existing research experience, historical data or simple estimation. The data assimilation technology is used to fuse the observed data obtained in step S101 with the model simulation results to optimize the control variables at the independent points.

[0028] In this embodiment, for the open boundary conditions, the cubic spline interpolation method is adopted. This method constructs a piecewise cubic polynomial to ensure that the interpolation function has continuous first and second derivatives at the nodes, thereby smoothly fitting the change curve of the open boundary conditions; for the bottom friction coefficient, surface spline interpolation is used. By constructing a smooth surface in space, the bottom friction coefficient values at unknown points are estimated according to the known data points, which is suitable for parameter optimization in two-dimensional space.

[0029] The data assimilation in this embodiment uses the effective information in the observed data to correct the model parameters to reduce the error between the simulation results and the actual observations. The cubic spline interpolation and surface spline interpolation, through mathematical fitting means, reasonably infer the control variable values at other positions according to the limited known data points, making the distribution of the control variables more in line with the characteristics of the actual ocean environment. In this way, by combining data assimilation with the interpolation method, the value of the observed data is fully exploited, and the control variables are optimized by mathematical means, improving the accuracy and rationality of the model parameters.

[0030] S103: Repeat steps S101 to S102 continuously to optimize the model control variables until the error between the simulation results and the observed data reaches within a preset threshold.

[0031] In this embodiment, steps S101 and S102 are continuously repeated. In each iteration, the forward model is re-run based on the control variables optimized in the previous round to obtain new simulation results and compare them with the observed data to calculate the error. When the error between the simulation results and the observed data reaches within the pre-set threshold range, it is considered that the model control variables have been optimized to an appropriate degree, and the iteration is stopped.

[0032] It can be seen that the method of this embodiment gradually adjusts the control variables through iterative optimization, making the model simulation results continuously approach the actual observed data. By using the error feedback mechanism, the preset error threshold is used as the convergence condition to achieve the optimal solution of the control variables, ensuring that the model control variables are fully optimized to meet the accuracy requirements for internal tide simulation and prediction in practical applications.

[0033] Based on the above embodiments, in order to further improve the reliability of the adaptive optimization method for collaborative inversion of multi-scale control variables provided in the above embodiments, the following is a more specific implementation manner. In the following embodiments, the specific steps are as follows: S201: Drive the forward model with the given control variables. The control variables are the open boundary conditions and the bottom friction coefficient, and the surface current velocity generated at the given observation points of the simulation results is used as the observed data.

[0034] In this embodiment, an isopycnal internal tide adjoint assimilation model is used for internal tide simulation. The forward model is driven with the given control variables (open boundary conditions, bottom friction coefficient), and the surface current velocity generated at the given observation points of the simulation results is used as the "observed data".

[0035] S202: Assign initial values to the control variables, and optimize the control variables at the independent points by assimilating the observed data. Use cubic spline interpolation to optimize and invert the open boundary conditions, and use surface spline interpolation to optimize and invert the bottom friction coefficient.

[0036] In some embodiments, initial values are assigned to the control variables. The initial value of the open boundary condition is 0, and the initial value of the bottom friction coefficient is 0.002. The control variables at the independent points are optimized by assimilating the "observed data" in S101. The cubic spline interpolation method is selected for the optimization of the open boundary conditions, and the surface spline interpolation method is selected for the optimization of the bottom friction coefficient.

[0037] The method for optimizing one of the model parameters, the open boundary conditions, using the cubic spline interpolation method in this embodiment is as follows: Some specific points are selected as independent points among all the open boundary points in the computational domain. Their values are calculated using the optimization algorithm, and the values at other grid points are obtained by cubic spline interpolation of the values of the independent points. Specifically, let N be the total number of open boundary points, be the values of the open boundary conditions of the independent points, be the number of independent points, be the result obtained by cubic spline interpolation, that is, (1) where is the cubic spline interpolation coefficient, and its specific derivation process is as follows: Assume that on the interval the cubic spline interpolation function exists, and use to represent the derivative value of at the point i = 0, 1, 2, …, n , where n + 1 = N P , then on each small interval it satisfies: (2) The expression of on the interval is as follows: Among them, , , next, take the second derivative of to obtain the expression as follows: (4) In order to ensure that has a continuous second derivative at the node , there should be , that is (5) Simplify and organize the above formula to get: (6) Among them:

[0038] The system of equations (6) is a n + 1 - unknown n - 1 linear system of equations, so the solution of this system of equations has infinitely many. In the application of practical problems, usually only a specific solution can be selected. At this time, boundary conditions need to be given according to specific situations. Common boundary conditions are: The first type of boundary condition: The tangent slopes of the curve at the two endpoints and are known, that is and are known (optimized to 1 for open boundary conditions); The second type of boundary condition: The second derivatives of the function at the two endpoints and are known, that is and It is known that the optimization for open boundary conditions is given as 0; Periodic boundary conditions: The function is a periodic function. Correspondingly, the spline function should also be a periodic function, that is, it satisfies the conditions and .

[0039] After giving the three types of boundary conditions, continue to derive the cubic spline interpolation coefficients. Since the derivation processes of these three types of boundary conditions are similar, only the periodic boundary conditions are taken as an example for derivation. To simply apply the cubic spline interpolation method to the optimization inversion of open boundary conditions, the independent points are selected on the open boundary in an equally spaced manner (the distance is set as h ), that is, assume , then there is . According to the characteristics of the periodic boundary conditions, there is (7) Combining equations (6) and (7) can obtain: (8) Where: (9) Combining equations (8) and (9), the values of the points on the open boundary can be obtained through calculation: (10) The above formula is the specific manifestation form of formula (1). Then, it is necessary to extract the specific expressions of the cubic spline interpolation coefficients from formula (10).

[0040] Let . If j is on , then the cubic spline interpolation coefficient of the l-th independent point corresponding to the i-th grid point is: j (11) After derivation, the gradient of the cost function J with respect to the open boundary conditions at the independent points can be obtained: (12) For the optimization inversion of open boundary conditions, such as Figure 2As shown in the figure, taking the periodic boundary condition as an example, the Fourier coefficient a in the open boundary condition is inverted. The red solid line in the figure represents the given open boundary condition, that is, the spatial distribution of the Fourier coefficient a. The black solid line and the black dashed line respectively represent the results after inversion by cubic spline interpolation and linear interpolation. Compared with the independent point scheme based on linear interpolation given in the figure, the open boundary curve obtained by using the cubic spline interpolation scheme is smoother, more conforms to the given spatial distribution, has a smaller error, and has a stronger physical meaning.

[0041] S203: Repeat steps S201 to S202, continuously optimize the model control variables, so that the error between the simulation result and the observed data reaches within the preset threshold.

[0042] In step S203 of this embodiment, the method for optimizing one of the model parameters, the bottom friction coefficient, by using the surface spline interpolation method is as follows: Select the values of some grid points (a total of N ) in the bottom friction coefficient field as independent points. Let be the value of the bottom friction coefficient at the l th independent point. The value of each grid point in the bottom friction coefficient field is obtained through surface spline interpolation. and satisfy the following relationship: (13) Among them, is the surface spline interpolation coefficient, and its specific derivation process is as follows: According to the expression of the surface spline interpolation function, equation (13) can be rewritten as: (14) Among them, R is the influence radius, represents the distance between the grid point in the bottom friction coefficient field and the l th independent point. In order to determine the specific expression form of the coefficient matrix , equation (14) is rewritten in matrix form: (15) Specifically,

[0043] Among them,

[0044] Let , then equation (15) can be written as , and then we can get (16) Substituting equation (16) into equation (14), we can obtain (17) Therefore, the expression form of the surface spline interpolation coefficient is extracted from equation (17): (18) Through derivation, the cost function can be obtained J The gradient with respect to the bottom friction coefficient at the independent points: (19).

[0045] The gradient involved in this embodiment indicates the cost function J The rate of change with respect to the bottom friction coefficient at the independent points. In the optimization algorithm, the parameter update direction is the negative gradient direction, and the step size is controlled by the learning rate. The magnitude of the gradient reflects the contribution degree of adjusting the bottom friction coefficient at a certain independent point to reducing the simulation and observation errors. For example, points with larger gradients have a more significant impact on the errors and need to be adjusted preferentially. By iteratively calculating the gradient and updating the bottom friction coefficient, the control variables gradually approach the optimal value, reducing the error between the simulated flow velocity and the observed data. The smoothness of the surface spline interpolation ensures the spatial continuity of the gradient propagation, avoiding local overfitting, so that the global error converges more uniformly. When the gradient norm is lower than the preset threshold, it indicates that the error has been sufficiently reduced and the optimization process can be terminated.

[0046] It should be understood that the magnitudes of the sequence numbers of the steps in the above embodiments do not mean the order of execution. The execution order of each process should be determined according to its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present invention.

[0047] As an embodiment of this application, for a certain port, hereinafter referred to as Port X, due to the need for channel expansion, it is necessary to improve the prediction accuracy of the surface tide to ensure the safety of ships entering and leaving the port. Previously, the tidal model used by the port had obvious deviations between the simulated surface flow velocity and the measured values (average error of about 0.4 m / s) due to inaccurate input of the sea boundary tide level and setting of the seabed resistance parameters, affecting ship berthing scheduling.

[0048] Therefore, the multi-scale control variable collaborative inversion method proposed by the present invention is used to optimize the model control variables, with the goal of reducing the error between the simulated flow velocity and the measured values to within 0.15 m / s.

[0049] Step S501: Construct a forward model and generate observation data.

[0050] The sea area of X Port is about 50 km², and the average water depth is 8 m. The forward model takes the "open boundary condition" and "bottom friction coefficient" as the core control variables to drive the model to simulate tidal movements.

[0051] According to the historical tide level data of the sea area where the port is located, the tide level time series of the model boundary is set with the same period average value. For example, the tide level at point A at the bay mouth follows the semi-diurnal tide law, rising and falling twice a day, with an average tide level of 2.0 m and a maximum tidal range of 3.5 m; at point B at the bay head, due to the shallowing of the water depth, the tidal amplitude decays to 2.5 m, and the phase lags by about 2 hours. Referring to the historical research experience of this sea area, the initial bottom friction coefficient is uniformly taken as 0.0025 without spatial differences, assuming uniform seabed resistance.

[0052] Five key observation points are selected in X Port (C: water depth of 6 m near the bay mouth, D: water depth of 10 m at the center of the channel, E: water depth of 4 m at the shoal at the bay head, F: water depth of 7 m outside the breakwater, G: water depth of 5 m at the port entrance). The flow velocities at the surface layer and 1 m below the water surface are continuously measured in real time for 30 days by an acoustic Doppler current profiler (ADCP) as the real data for subsequent optimization.

[0053] Step S502: Initialize the control variables and carry out collaborative optimization.

[0054] Set the open boundary condition, initially using the historical average tide level; the initial value of the whole field is uniformly 0.0025. Taking the surface flow velocity at the observation points as the target, the control variables are adjusted by comparing the model simulation values with the measured values. After the model runs, it is found that the simulated flow velocity is generally 0.2 - 0.3 m / s lower than the measured value at the initial stage of the flood tide. It is speculated that the initial phase of the tide level at the bay mouth is slower. The open boundary tide level time series is adjusted by cubic spline interpolation - the tide level time curve at point A at the bay mouth is "advanced" by 1 hour. For example, the tide level of -1.5 m at the original t = 0 is adjusted to -1.3 m, and the high tide level of 1.5 m at t = 6 is adjusted to 1.7 m, so that the interpolated tide level sequence is more in line with the actual ebb and flow rhythm.

[0055] Further analysis shows that at the observation point E in the nearshore shoal area with a water depth of 4 m, the simulated flow velocity is significantly lower than the measured value, while at the observation point D in the deep water area with a water depth of 10 m, the error is only 0.1 m / s.

[0056] It is speculated that the seabed friction in the shoal area is stronger, and the bottom friction coefficient needs to be increased. Through surface spline interpolation, the spatial distribution of the bottom friction coefficient is constructed in the two-dimensional sea area - the bottom friction coefficient around point E with a nearshore water depth < 5 m is increased from 0.0025 to 0.0035, the water depth > 8 m, and point D remains unchanged at 0.002. In the transition area, the water depth of 5 - 8 m is smoothly connected to avoid parameter mutations.

[0057] Step S503: Iteratively optimize until the error meets the standard.

[0058] First iteration: The open boundary tidal level is advanced by 1 hour, and the nearshore bottom friction coefficient is increased. The average error between the simulated and measured surface flow velocities at 5 observation points is 0.28 m / s.

[0059] It is found that the simulated flow velocity at point B at the bay head is still low, with an error of 0.25 m / s. The tidal level amplitude at the bay mouth is fine-tuned again through cubic spline interpolation, increased from 3.5 m to 3.7 m, so that the lag time of the tidal level at the bay head is shortened to 1.5 hours.

[0060] The simulated flow velocity error near point E still reaches 0.2 m / s. The bottom friction coefficient in this area is further increased to 0.0038 through surface spline interpolation, which is closer to the actual shoal resistance.

[0061] Second iteration: The model is re-run, and the average error is reduced to 0.18 m / s. The simulated flow velocity error at the center of the channel is 0.12 m / s, approaching the threshold. The tidal level phase at the bay mouth is fine-tuned, advanced by another 0.5 hour, so that the tidal level at point D is synchronized with the measurement; at point F outside the breakwater, the water depth is 7 m. Due to the flow being blocked by the breakwater, the simulated flow velocity is 0.1 m / s higher. The bottom friction coefficient in this area is slightly reduced to 0.0022 through surface spline interpolation.

[0062] Third iteration: After running, the average error at 5 observation points is reduced to 0.12 m / s, lower than the 0.15 m / s threshold, meeting the accuracy requirements, and the optimization is stopped.

[0063] Final result: The tidal level time series corrected by cubic spline interpolation is adopted. The tidal level amplitude at the bay mouth is 3.7 m, the period is 12.4 hours, and the phase is completely synchronized with the measurement; the spatial distribution field obtained through surface spline interpolation is 0.0035 - 0.0038 in the area with a nearshore water depth < 5 m, 0.0022 in the area with a water depth of 7 m outside the breakwater, and 0.002 in the deep water area, with a smooth change in the transition zone. The average error between the simulated surface flow velocity of the optimized model and the measured data is only 0.12 m / s. The port dispatching personnel feedback that the prediction accuracy of the model for the high tidal level and strong flow velocity periods during ship entry into the port has been significantly improved, effectively reducing the berthing delays caused by tidal misjudgment, and the practical application value is prominent.

[0064] The adaptive optimization method for multi-scale control variable collaborative inversion of the present application combines the units and algorithm steps of each example described in the embodiments disclosed herein, and can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0065] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. An adaptive optimization method for collaborative inversion of multi-scale control variables, characterized in that, The method includes: S101. Driving the forward internal tide model to run based on the preset open boundary conditions and bottom friction coefficient as the model input parameters, and extracting the surface velocity data from the preset observation points as the observation data set; S102. Setting the initial parameter values based on the open boundary conditions and bottom friction coefficient, and optimizing the control variables at the independent points by assimilating the observation data set: using the cubic spline interpolation technique to perform nonlinear fitting on the spatial distribution of the open boundary conditions, and using the surface spline interpolation method to perform multi-dimensional smoothing optimization on the field structure of the bottom friction coefficient; S103. Repeatedly execute S101 and S102, adjust the model input parameters, and continuously optimize the independent point values of the open boundary conditions and bottom friction coefficient until the deviation index between the simulation result and the observation data set falls within the preset error threshold range.

2. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 1, wherein The specific steps of S101 are: performing forward simulation based on the preset internal tide adjoint assimilation model, using the given open boundary conditions and bottom friction coefficient as the model input parameters, running the forward model to obtain the simulation result, and extracting the surface velocity data from the preset observation points as the observation data set.

3. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 1, characterized in that The specific steps of S102 are: setting the initial values of the control variables, the initial value of the open boundary condition is 0, and the initial value of the bottom friction coefficient is 0.002, and optimizing the control variables at the independent points based on the observation data set obtained in S101: Performing spatial distribution optimization on the open boundary conditions using the cubic spline interpolation method; performing multi-dimensional field structure optimization on the bottom friction coefficient using the surface spline interpolation method.

4. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 1, wherein The specific steps of optimizing and inverting the open boundary conditions using the cubic spline interpolation are: selecting preset points as independent points among all the open boundary points in the calculation area, and using the optimization algorithm to evaluate the values; The values of the remaining grid points are obtained by cubic spline interpolation of the values of the independent points. Let N be the total number of open boundary points, be the values of the open boundary conditions of the independent points, be the number of independent points, is obtained by using through cubic spline interpolation, that is （1） Among them, are the cubic spline interpolation coefficients.

5. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 4, wherein In the method, a cubic spline interpolation function defined on the interval exists, and is denoted by . Let represent the derivative value at the point . Then, on each subinterval , it satisfies: ​ （2）。 6. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 5, wherein In the interval The expression is as follows: （3） Among them, , by taking the second derivative of , the obtained expression is as follows: （4）。 7. The adaptive optimization method for multi-scale control variable collaborative inversion according to claim 4, characterized in that The open boundary conditions are as follows: The first type of open boundary condition is that the tangent slopes of the curve at the two endpoints and are known, that is and are known; The second type of open boundary condition: The function at the two endpoints and the second-order derivatives are known, that is and are known; Periodic boundary conditions: The function is a periodic function, and the spline function should also be a periodic function, i.e., satisfying the conditions at the endpoints and .

8. The adaptive optimization method for collaborative inversion of multi-scale control variables according to claim 1, characterized in that In S102, optimizing and inverting the bottom friction coefficient using the surface spline interpolation specifically includes: Select the values of some grid points in the bottom friction coefficient field as independent points. Let be the value of the bottom friction coefficient at the l -th independent point. The values of each grid point in the bottom friction coefficient field are obtained by surface spline interpolation. and satisfy the following relationship: （13） Among them, is the surface spline interpolation coefficient.

9. The adaptive optimization method for collaborative inversion of multi-scale control variables according to claim 8, characterized in that In the method, the expression form of the surface spline interpolation coefficient is extracted: （18） Obtain the cost function J Gradient with respect to the bottom friction coefficient at the independent points: （19）； By transferring the gradient of the cost function with respect to the bottom friction coefficient field to independent points, clarifying the influence of interpolation and distance on the gradient, and driving parameter updates by the gradient, combined with the smoothness of the surface spline, to reduce the simulation and observation errors and ultimately achieve the goal of data assimilation.

Images