Ocean multi-dimensional mixing process global analysis optimization method

Abstract

The invention relates to the crossing field of marine geochemistry and computational mathematics, in particular to an optimization method for global analysis of a marine multi-dimensional mixing process, which comprises the following steps: S1, data acquisition; s2, carrying out data initialization processing; s3, calculating data; s4, analyzing a result; s5, outputting data; through collaborative design of multi-stage step length optimization and parallel computing architecture, the key problem that calculation efficiency and precision of a traditional OMPA model in a high-dimensional parameter space are difficult to consider at the same time is effectively solved. The core technology breakthrough is embodied in three aspects: firstly, a three-stage step optimization mechanism (0.05-0.01-0.001) with an adaptive characteristic is developed, and the calculation complexity is greatly reduced on the premise of ensuring a global optimal solution; secondly, a parallel computing framework based on MATLAB parFor is constructed, so that the computing efficiency of the 3-6-dimensional end member hybrid model is improved by an order of magnitude; and finally, a standardized data processing flow (z-score method) is matched with a modular architecture design, so that the expandability and applicability of the system are enhanced.

Description

Technical Field

[0001] This invention relates to the interdisciplinary field of marine geochemistry and computational mathematics, specifically to an optimization method for global analysis of multidimensional marine mixing processes. Background Technology

[0002] Multidimensional mixing processes in the ocean are key mechanisms driving material exchange between different water systems, directly affecting the dynamic balance of biogeochemical cycles in my country's coastal waters. Specifically, the sources of nutrients in coastal waters (including river input, atmospheric deposition, sediment-water interface exchange, and water mass transport) not only provide the necessary material basis for marine primary productivity, but their concentration gradient changes and elemental composition characteristics also regulate the ecological security pattern through nutrient limitation effects (such as the N / P / Si ratio). Taking the Yangtze River Estuary-Zhejiang coast as an example, nutrient structure variation and red tide outbreak frequency show significant spatiotemporal coupling. Current research on the quantification of multidimensional mixing processes still has two major limitations: ① Traditional endmember analysis models are not sufficiently applicable to complex scenarios such as the Yangtze River dilute-Taiwan Warm Current interaction zone; ② The accuracy of material tracing in high-dimensional parameter spaces (≥4 endmembers) is limited by the density of observational data and the convergence efficiency of algorithms.

[0003] Current research on multidimensional mixing processes in the ocean mainly relies on three types of technologies: satellite remote sensing technology achieves macroscopic monitoring of surface waters through indirect means such as machine learning, but it suffers from limitations in both spatiotemporal resolution and quantitative accuracy in nearshore areas where land-sea interactions are significant; numerical models can analyze the spatiotemporal evolution mechanism of nutrients by coupling hydrodynamic and biogeochemical processes, but their accuracy is limited by the boundary constraints provided by remote sensing observations and field data; multi-dimensional tracer technology can provide clear source indications by directly measuring water body characteristic parameters, but it is difficult to achieve precise quantification of multidimensional mixing processes due to the limited types of tracers and insufficient detection accuracy. It is worth noting that with breakthroughs in high-precision marine parameter analysis technologies such as rare earth elements and hydrogen and oxygen isotopes, constructing a multi-tracer collaborative global quantification method based on high-dimensional parameter space has become practically feasible.

[0004] Multivariate ocean parameters (such as tracers) provide theoretical feasibility for analyzing multidimensional ocean mixing processes, but this is accompanied by computational challenges related to nonnegative least squares (NNLS) problems in high-dimensional parameter spaces. Tomczak (1981) pioneered the Multivariate Parameter Optimization Analysis (OMPA) method, based on conventional parameters such as temperature, salinity, and dissolved oxygen, laying the theoretical foundation for research in the Indian and Atlantic Oceans. From the late 20th to the early 21st century, technological limitations restricted ocean tracers to basic hydrological parameters such as temperature and salinity. In recent years, breakthroughs in trace element (such as rare earth elements) and isotope determination technologies have promoted the application of novel tracers such as hydrogen and oxygen isotopes, but this has led to a core contradiction: an imbalance between the increasing dimensionality of tracer parameters and the computational efficiency of high-dimensional mixing. Specifically, the quantization computation time for mixing models with six dimensions or higher increases significantly, and the computation cycle increases exponentially at higher dimensions. This bottleneck severely restricts the practical application of multivariate parameter tracer methods.

[0005] Various optimization methods have been developed for solving the least squares problem in high-dimensional parameter spaces, including algorithms that prioritize computational efficiency such as gradient descent, Newton's method, and quasi-Newton methods. These methods are implemented in mathematical software such as Matlab using overdetermined equation functions like pinv and lsqnonneg. However, these efficient algorithms often only obtain local optima rather than global optima, leading to inherent errors in the computational results. In contrast, while exhaustive search methods using global traversal can guarantee a global optimum without computational errors, their computation time increases geometrically with the dimension of the parameter space (the specific multiple depends on the step size setting). This characteristic severely limits the feasibility of this method in practical applications.

[0006] To address the trade-off between computation time and accuracy, this invention innovatively employs a multi-level refinement architecture combined with a parallel computing framework, significantly shortening the computation cycle and successfully achieving efficient solutions to least squares problems in high-dimensional parameter spaces. Furthermore, it systematically constructs complete computational logic and supporting Matlab programs. Summary of the Invention

[0007] The purpose of this invention is to provide an optimized method for global analysis of multidimensional ocean mixing processes. This method enables the analysis of multidimensional ocean mixing processes, optimizes the computational logic, reduces the computational cycle, and improves computational accuracy, thereby addressing existing technical deficiencies and unmet technical requirements.

[0008] To achieve the above objectives, the present invention provides the following technical solution: an optimization method for global analysis of multidimensional ocean mixing processes, comprising:

[0009] S1. Data Acquisition: Acquire multidimensional parameter data of endmember water masses and seawater samples;

[0010] S2, Data Initialization Processing

[0011] 2.1) Read the collected multidimensional parameter data;

[0012] 2.2) Standardize all collected multidimensional parameter data.

[0013] In this application, this step can be understood as normalizing the parameters based on their range of variation and standard deviation using the zscore algorithm to reduce the error caused by combining the parameters.

[0014] S3. Data Calculation: A multi-level refinement architecture and parallel computing framework are used to calculate the global optimal solution for the standardized data (including the mixing ratio of each endmember and the minimum residual).

[0015] S4. Results Analysis

[0016] 4.1) Evaluate the relative error of each parameter.

[0017] In this application, the above steps can be understood as follows: obtaining the theoretical parameter values ​​of the sample based on the measured parameter values ​​and contribution ratios of each end-member tracer, and comparing them with the measured values ​​of the sample to obtain the relative error.

[0018] 4.2) Assess the extent of the impact of non-conservative processes.

[0019] In this application, the above steps can be understood as follows: obtaining the theoretical non-conservative parameter values ​​of the sample based on the non-conservative measured parameter values ​​and contribution ratios of each endmember, and comparing them with the measured values ​​of the sample to obtain the relative magnitude of the influence of the non-conservative process.

[0020] S5. Data Output: Generate an analysis report containing the mixing ratio of each endmember (obtained from step S3), the minimum residual (obtained from step S3), and the errors of each parameter (obtained from steps 4.1 and 4.2).

[0021] Preferably, the multidimensional parameter data in step S1 includes: tracer parameter data and other non-conservative seawater parameter data, wherein the tracer data includes multiple parameters such as the temperature and salinity of the seawater sample, eleven rare earth elements and two hydrogen and oxygen isotopes in the seawater sample, and the non-conservative seawater parameter data includes multiple parameters such as nutrients and dissolved oxygen.

[0022] Preferably, 2.1.1), read the multidimensional parameter data of the endmember water mass and the measured multidimensional parameter data of the seawater sample, and assign the endmember data and sample data to the arrays endmember_data and sample_data;

[0023] 2.1.2) Validate the validity of the read data;

[0024] The steps described above in this application can be understood as follows: determine whether the two arrays are empty; if not, proceed to the next step; if both arrays are empty, report "Data read failed, please check if the data.xlsx file exists and is in the correct format".

[0025] 2.1.3) Read the number of endmembers, the number of parameters, the number of samples, and the weight of each tracer parameter as input by the user.

[0026] Preferably, in step 2.2), the zscore method is used for data standardization, specifically as follows:

[0027] 2.2.1) Assign the data input by the user to variables e, n, m and W respectively, where e is the number of endmembers, n is the number of parameters, m is the number of samples and W is the weight array of the parameters;

[0028] 2.2.2) Substitute the multi-dimensional parameter data into the following formula;

[0029] endmember_norm=(endmember_data-endmember_mean) / endmember_std

[0030] Where, endmember_mean represents the mean of each parameter, endmember_std represents the standard deviation of each parameter, endmember_norm represents the standardized endmember parameters, and endmember_data represents the original endmember parameters;

[0031] 2.2.3) Using the same algorithm as in step 2.2.2), calculate sample_norm and sample_norm, that is, substitute the multi-dimensional parameter data into the following formula:

[0032] sample_norm=(sample_data-endmember_mean) / endmember_std

[0033] Here, sample_norm represents the standardized sample parameters, and sample_data represents the original sample parameters.

[0034] Preferably, in step S3, the multi-level refinement architecture adopts a three-level step size optimization mechanism, with step sizes of 0.05, 0.01 and 0.001 respectively.

[0035] Preferably, the parallel computing framework adopts MATLAB's parfor parallel computing algorithm to make full use of the CPU's multi-core computing capabilities.

[0036] Preferably, in step S3, the specific content of calculating the global optimal solution is as follows:

[0037] 3.1) Dynamically select the calculation function based on the number of endmembers input by the user, and calculate different sample data based on the selected function;

[0038] 3.2) Substitute the endmember_norm, sample_norm, and W arrays into the selected function;

[0039] 3.3) The optimized exhaustive search method and multi-level refined search are used to calculate the global optimal solution;

[0040] 3.4) Output the global optimal solution set (including the contribution ratio of each endmember and the minimum residual).

[0041] In this application, the above steps can be understood as follows: based on the number of endmembers input by the user, the corresponding calculation function is automatically selected, and the data of multiple samples are calculated in parallel using the parfor function; based on the number of endmembers, the number of parameters, and the number of samples input by the user, the residuals of the equation system under different combinations of the contribution (0-100%) of each endmember are calculated sequentially, and the calculation is refined in three steps (5%, 1%, 0.1%) to find the solution set that minimizes the residuals of the least squares equation system (i.e., the contribution ratio of each endmember); the contribution ratio of each endmember and the minimum residual data are output.

[0042] Preferably, in step 4.1), the specific content of evaluating the relative error of each parameter is as follows:

[0043] 4.1.1) Obtain the measured data of the tracer parameters in step 2.1), and obtain the contribution ratio of each endmember in step 3.4);

[0044] 4.1.2) The theoretical calculated values ​​of the sample tracer parameters are obtained by multiplying the contribution ratio of each endmember by the measured data of the endmember tracer.

[0045] 4.1.3) Extract the actual measured values ​​of the sample tracer parameters from the sample_data array;

[0046] 4.1.4) Compare and analyze the theoretically calculated values ​​of the sample tracer parameters with the actual measured values ​​of the sample tracer parameters to quantitatively assess the degree of influence of non-conservative processes on each tracer parameter during the multidimensional mixing process.

[0047] Preferably, in step 4.2), the specific content of assessing the degree of influence of the non-conservative process includes:

[0048] 4.2.1) Obtain the measured data of the non-conservative parameters in step 2.1), and obtain the contribution ratio of each endmember in step 3.4);

[0049] 4.2.2) Obtain the theoretically calculated value of the sample's non-conservative parameters by multiplying the contribution ratio of each endmember by the measured data of the endmember non-conservative parameters;

[0050] 4.2.3) Extract the actual measured values ​​of non-conservative parameters of the sample from the sample_data array;

[0051] 4.2.4) Compare and analyze the theoretically calculated values ​​of the non-conservative parameters of the sample with the actual measured values ​​of the non-conservative parameters of the sample to quantitatively evaluate the degree of influence of the non-conservative process on each non-conservative parameter in the multidimensional mixing process.

[0052] In this application, it should be explained that the parameter data in steps 4.1.1) and 4.2.1) should be the measured data before standardization. The purpose of standardization in step 2.2) is to substitute each parameter into the subsequent least squares calculation formula (i.e., the Function calculate_xendmember function) to ensure parameter consistency in the calculation process and ultimately obtain the contribution ratio of each endmember. In steps 4.1) and 4.2), error evaluation will be performed by combining the measured parameter data (obtained in step 2.1) and the endmember contribution ratio (in step 2.2) with the subsequent step S3.

[0053] Preferably, an optimization device for global analysis of multidimensional ocean mixing processes includes:

[0054] An input device, the input device being used to acquire multidimensional parameter data of endogenous water masses and seawater samples;

[0055] A data input unit, electrically connected to both the input device and the data storage unit, is used to input multidimensional parameter data of endogenous water masses and seawater samples; (corresponding to S1)

[0056] A data storage unit, electrically connected to the data input unit, is used to read and store multidimensional parameter data of endogenous water masses and seawater samples, and to store analysis result report data; (corresponding to S2 and S5)

[0057] The data processing unit, which is an integrated unit of multiple parallel processor cores, is electrically connected to the data storage unit. It is used to perform data standardization and multi-level data refinement calculations, accelerate multi-dimensional hybrid process calculations, and generate analysis reports; (corresponding to S2, S3, and S4)

[0058] The result output unit is electrically connected to both the data storage unit and the display, and is used to output an analysis report. (Corresponding to S5)

[0059] Preferably, the device runs on MATLAB R2017b or a higher version.

[0060] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described above.

[0061] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0062] The overall logic of this application can be understood as follows:

[0063] 1. Overall Functionality: Used to calculate 3-6 dimensional endmember hybrid models and parameter deviations, and reserves more space for expansion of endmember process calculations.

[0064] 2. Initialization settings: Clear the command window, workspace, and graphics window, and set the output data format to short floating-point numbers.

[0065] 3. Data preparation: Read the end metadata and sample data from the data.xls file and perform validity verification.

[0066] 4. User interaction configuration: Obtain the number of endmembers, number of parameters, number of samples, and weights of each tracer parameter through command-line interaction.

[0067] 5. Data preprocessing: The zscore method is used to standardize the end metadata and sample data.

[0068] 6. Core Calculation Module: Dynamically selects the calculation function based on the number of endmembers input by the user, performs parallel calculations on different sample data, and uses an optimized exhaustive search method and multi-level refinement search to calculate the global optimal solution, thereby improving calculation accuracy and efficiency.

[0069] 7. Results Analysis and Output: Calculate the relative error of the samples and assess the accuracy of the calculation; based on the calculated proportions of each endmember, further calculate the deviations of other parameters to expand the analytical potential; write the results into a new Excel file.

[0070] Compared with the prior art, the beneficial effects of the present invention are:

[0071] 1. This application effectively solves the key problem of balancing computational efficiency and accuracy in high-dimensional parameter spaces in traditional OMPA models through the collaborative design of multi-level step-size optimization and parallel computing architecture. The core technological breakthroughs are reflected in three aspects: First, an adaptive three-level step-size optimization mechanism (0.05→0.01→0.001) was developed, significantly reducing computational complexity while ensuring the global optimal solution; second, a parallel computing framework based on MATLAB parfor was constructed, resulting in an order-of-magnitude improvement in the computational efficiency of the 3-6 dimensional endmember hybrid model; finally, a standardized data processing flow (z-score method) combined with a modular architecture design enhances the system's scalability and applicability.

[0072] 2. This application adopts a multi-level refinement architecture combined with a parallel computing framework, which significantly shortens the computation cycle, successfully achieves efficient solution of the least squares problem in high-dimensional parameter space, and systematically constructs complete computational logic and supporting Matlab programs. Attached Figure Description

[0073] Figure 1 This is a schematic diagram of the overall logic in this invention;

[0074] Figure 2 This is a schematic diagram of the overall calculation process in this invention;

[0075] Figure 3 This is a schematic diagram of the multi-level refined OMPA calculation logic in this invention;

[0076] Figure 4 This is a schematic diagram of the structural composition of the computer device in this invention; Detailed Implementation

[0077] The following will refer to the appendices in the embodiments of the present invention. Figure 1-4 The technical solutions in the embodiments of the present invention are clearly and completely described herein. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of the present invention.

[0078] In the description of this invention, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined with "first" and "second" may explicitly or implicitly include one or more of that feature.

[0079] Please see Figure 1-4 Embodiments of the present invention:

[0080] Example 1:

[0081] As shown in the figure: An optimization method for global analysis of multidimensional mixing processes in the ocean, characterized by comprising:

[0082] S1. Data Acquisition: Acquire multidimensional parameter data of endmember water masses and seawater samples;

[0083] Specifically: Based on the mixing process of water masses in the Yangtze River estuary, two types of core data are collected for multidimensional mixing global analysis: tracer data (temperature θ, salinity S, and hydrogen and oxygen isotope δ). 18 O and δD, concentrations of 11 rare earth elements (Y / La / Pr / Nd / Sm / Tb / Dy / Ho / Er / Tm / Yb), and other non-conservative seawater parameters (nutrients, dissolved oxygen, etc.) were collected. The study covered four major end-member water masses: Yangtze River water, Yellow Sea cold water, Taiwan warm current water, and Kuroshio Current water. Complete parameter datasets of seawater samples from each end-member water mass and the adjacent sea area of ​​the Yangtze River estuary were required. All data were standardized to ensure consistency in subsequent calculations. Data was stored in a structured Excel file (Data.xls), specifically divided into: Sheet1 (end-member data): recording end-member water mass parameters; Sheet2 (sample data): storing measured seawater sample parameters. Both sheets used the same data structure.

[0084] Line records: arranged in order of endmember / sample

[0085] Column definitions (columns 1-15): fixed as tracer parameters (temperature, salinity, 11 rare earth elements, 2 hydrogen and oxygen isotopes); (columns 16-18): fixed as other seawater parameters (nutrients, dissolved oxygen, etc.).

[0086] Technical advantages: The modular data architecture supports user-defined configurations, significantly improving computing efficiency.

[0087] S2, Data Initialization Processing

[0088] 2.1) Read the collected multidimensional parameter data;

[0089] 2.1.1) Read the multidimensional parameter data of the endmember water mass and the multidimensional parameter data of the seawater sample;

[0090] 2.1.2) Validate the validity of the read data;

[0091] 2.1.3) Number of input terminal elements, number of parameters, number of samples, and weight of each tracer parameter;

[0092] In this embodiment, specifically, a workspace cleanup operation is first performed (code clc; clear; close all;) to eliminate potential interference from historical data variables. Then, a standardized output format is set (format short g;) to ensure data format consistency, thereby improving processing efficiency. The system reads end-member metadata and sample data from the Data.xlsx file and stores them in the endmember_data and sample_data arrays, respectively. Tracer parameters are stored in columns 1-15 of the endmember_data and sample_data arrays, while non-conservative parameters such as nutrients and dissolved oxygen are stored in columns 16-18 for subsequent analysis in S3-S4.

[0093] If data reading fails (array is empty), the error message "Data reading failed. Please check if the data.xlsx file exists and is in the correct format" will be displayed.

[0094] The system guides the user to input key parameters: the number of endmembers *e*, the number of parameters *n*, the number of samples *m*, and the weight array *W*. The weight array *W* can be defined using the method described by Karstensen and Tomczak (1999) in the OMP Analysis Package for MATLAB Version 2.0, or it can be customized by the user. This design enables configurable model parameters, facilitating the application of the technical solution in various fields.

[0095] 2.3) Standardize all collected multidimensional parameter data;

[0096] In this embodiment, specifically, after the parameter input is completed, the system performs z-score standardization on the end-member metadata. The calculation formula is endmember_norm = (endmember_data - endmember_mean) / endmember_std, where endmember_mean and endmember_std represent the mean and standard deviation of the n parameters, respectively. Sample data will be standardized using the same statistical measures. This process ensures the comparability of parameters between different tracers and effectively eliminates the influence of dimensional differences on the calculation of multidimensional mixing processes.

[0097] S3. Data computation: A multi-level refined architecture and parallel computing framework are used to calculate the global optimal solution for the standardized data.

[0098] In this embodiment, specifically, the system is responsible for allocating multi-core tasks using the parfor algorithm based on the number of samples *m* obtained in the preceding step (2.1.3), selecting the core calculation function (in this embodiment, Function calculate_4endmembers, i.e., four-endmember calculation) based on the number of endmembers *e* obtained in the preceding step (2.1.3), and finally importing the standardized endmember data array *endmember_norm*, the standardized sample data array *sample_norm*, and the weight array *W* obtained in the preceding step (2.3) into the core calculation function Function calculate_4endmembers. The core calculation function outputs the contribution ratio array *opt_f* of each endmember and the minimum residual array *min_R*. This innovative approach utilizes MATLAB's parfor parallel computing technology, significantly improving computational efficiency by fully leveraging the multi-core processing power of the CPU. Since the calculation processes for each sample are independent, this parallel algorithm is particularly suitable for batch processing of large-scale sample data, effectively shortening the computation cycle.

[0099] The four typical endmembers involved in this embodiment include: Yangtze River water, Yellow Sea cold water, Taiwan Warm Current water, and Kuroshio Current water. The system will automatically match the corresponding calculation function (Function calculate_4endmembers) based on the actual number of endmembers. Considering the actual needs of ocean mixing processes, this solution supports calculation scenarios with 3 to 6 endmembers, basically covering the vast majority of ocean research needs. Its modular design allows users to easily extend higher-dimensional calculation functions based on the existing code architecture, demonstrating good adaptability and scalability.

[0100] S4. Results Analysis

[0101] 4.1) Evaluate the relative error of each parameter;

[0102] 4.1.1) Obtain the measured data of tracer parameters from step 2.1), and obtain the contribution ratio opt_f of each endmember from step S3;

[0103] 4.1.2) The theoretical calculated values ​​of the sample tracer parameters are obtained by multiplying the contribution ratio of each endmember by the measured data of the endmember tracer.

[0104] 4.1.3) Extract the actual measured values ​​of the sample tracer parameters from the sample_data array;

[0105] 4.1.4) Compare and analyze the theoretically calculated values ​​of the sample tracer parameters with the actual measured values ​​of the sample tracer parameters to quantitatively assess the degree of influence of non-conservative processes on each tracer parameter during the multidimensional mixing process.

[0106] 4.2) Assess the extent of the impact of non-conservative processes;

[0107] 4.2.1) Obtain the measured data of the non-conservative parameters in step 2.1), and obtain the contribution ratio opt_f of each endmember in step S3;

[0108] 4.2.2) Obtain the theoretically calculated value of the sample's non-conservative parameters by multiplying the contribution ratio of each endmember by the measured data of the endmember non-conservative parameters;

[0109] 4.2.3) Extract the actual measured values ​​of non-conservative parameters of the sample from the sample_data array;

[0110] 4.2.4) Compare and analyze the theoretically calculated values ​​of the non-conservative parameters of the sample with the actual measured values ​​of the non-conservative parameters of the sample to quantitatively evaluate the degree of influence of the non-conservative process on each non-conservative parameter in the multidimensional mixing process.

[0111] In this embodiment, specifically, this step inherits the endmember measured data array `endmember_data` and the sample measured data array `sample_data` from step 2.1, and calculates the relative errors for tracer parameters and non-conservative parameters respectively. The theoretical calculated values ​​for the samples are obtained by multiplying the contribution ratio array `opt_f` (obtained in step S3) of each endmember by the measured data of each endmember tracer (columns 1-15 of `endmember_data`) and the measured data of each endmember non-conservative parameter (columns 16-18 of `endmember_data`), respectively. The actual measured values ​​of the samples are directly extracted from the `sample_data` array. Through comparative analysis of theoretical and measured values, the influence of non-conservative processes (such as biological activity, particulate matter dissolution, etc.) on each tracer parameter and each non-conservative parameter during multidimensional mixing can be quantitatively assessed, and the relative error arrays `alpha_P` and `alpha_UC` for each tracer parameter and non-conservative parameter are output.

[0112] This step serves a dual purpose: first, to verify the conservation of tracer parameters and screen suitable tracers for computation; second, to analyze the non-conservative process effects of parameters such as nutrients and dissolved oxygen, expanding the application scope of the method. Its error analysis mechanism is similar to the mass participation accuracy assessment method in kinetic analysis, ensuring the reliability of the calculation results by quantifying deviations. While the current embodiment uses a specific mixing process as an example, its modular design supports parameter expansion to adapt to different research scenarios.

[0113] S5. Data Output: Generate an analysis report containing the mixing ratio of each endmember and parameter errors.

[0114] In this embodiment, specifically, this step is responsible for integrating the optimal solution calculation results generated in step S3 (including the endmember contribution ratio array opt_f and the minimum residual array min_R), and the relative error array generated in step S4 (including the relative error array alpha_P for each tracer parameter and the relative error array alpha_UC for each non-conservative parameter). This step systematically integrates core indicators such as endmember mixing ratio (Fraction), model residuals, tracer parameter relative errors, and non-conservative parameter relative errors through a result function, and automatically converts them into a standardized report in XLSX format. This structured output significantly improves data visualization, enabling researchers to intuitively obtain the analysis results of the mixture model, while also supporting secondary data processing and visualization operations on the Excel platform. The module employs dynamic data binding technology to ensure that the output report remains synchronized with the original calculation results in real time, providing a convenient data interface for subsequent mixture process verification and non-conservative effect evaluation.

[0115] An optimization device for global analysis of ocean multidimensional mixing processes, adapted to the method described above, includes:

[0116] An input device, the input device being used to acquire multidimensional parameter data of endogenous water masses and seawater samples;

[0117] A data input unit, electrically connected to an input device, is used to input multidimensional parameter data of endmember water masses and seawater samples (to the data storage unit); (corresponding to S1)

[0118] A data storage unit, electrically connected to the data input unit, is used to read and store multidimensional parameter data of endogenous water masses and seawater samples, and to store analysis result report data; (corresponding to S2 and S5)

[0119] A data processing unit, electrically connected to the data storage unit, is used to perform data standardization and multi-level data refinement calculations, accelerate multi-dimensional hybrid process calculations, and generate analysis reports; (corresponding to S2, S3, and S4)

[0120] The result output unit is electrically connected to both the data storage unit and the display, and is used to output an analysis report. (Corresponding to S5)

[0121] The device runs on MATLAB R2017b.

[0122] A computer-readable storage medium having a computer program stored thereon that, when executed by a processor, implements the steps of the method described above.

[0123] An electronic device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the method described above.

[0124] The overall implementation method in this embodiment is as follows:

[0125] like Figure 2 As shown, according to an embodiment of the present invention, the code operation logic from initialization to result output, and the related parameter iteration will be explained.

[0126] clc; clear; close all; clear the workspace; format short g; set the output format.

[0127] Read the data from the Data.xlsx file and assign it to the arrays endmember_data and sample_data according to the end metadata and sample data.

[0128] Check if both arrays are empty. If they are not empty, proceed to the next step. If both arrays are empty, report "Data reading failed. Please check if the data.xlsx file exists and is in the correct format".

[0129] Read the data entered by the user in sequence and assign it to variables e, n, m, and array W respectively.

[0130] Read the first 15 columns of data from the array `endmember_data`, calculate the mean `endmember_mean` and standard deviation `endmember_std` for each column, and normalize the parameters using the `zscore` function. Assign the results to the array `endmember_norm`. Then, using the same algorithm, normalize the first 15 columns of the array `sample_data`, and assign the results to the array `sample_norm`.

[0131] Based on variable e, the corresponding function name is selected; in this example, it is `Function calculate_4endmembers`. Substituting `endmember_norm`, `sample_norm`, and the `W` array into the function, the output result arrays `opt_f` and `min_R` are calculated. The core calculation function uses an exhaustive search combined with multi-level refinement loops. This ensures finding the global optimum while improving calculation accuracy by refining the search step size. A specific implementation of the calculation function will be discussed in conjunction with... Figure 3 A detailed introduction will follow.

[0132] Combining the arrays opt_f, endmember_data, and sample_data, the relative errors of each parameter of the sample are calculated, i.e., alpha_P = (opt_f(:,:)*endmember_data(1:e,:)-sample_data(1:m,:)). / sample_data(1:m,:). The first 15 columns of the output array alpha_P represent the relative errors of the 15 tracer parameters, and the columns after the 16th column represent the relative errors of the non-conservative parameters.

[0133] Read the current date and time, and name the output file MOMPA_Results_yyyymmdd_HHMM.xlsx. Write the arrays opt_f, min_R, and alpha_P to this output file in sequence.

[0134] like Figure 3 As shown, according to an embodiment of the present invention, the multi-level refined OMPA calculation process is described.

[0135] In this embodiment, the number of endmembers is 4, i.e., e = 4. Therefore, the core calculation function is selected as function[opt_f,min_R] = calculate_4endmembers(endmember_data,sample_data,W).

[0136] Define the step size array `stepSizes` for the three-level refinement loop as [0.05, 0.01, 0.001], and define the initial uniform scaling `opt_f` as [0.5, 0.5, 0.5, 0.5]. Define the initial residual `min_R` as infinity `inf`.

[0137] Execute the first loop, stepSizes[1] = 0.05;

[0138] The second loop is executed, traversing the variable space of the first terminator contributing the Friction variable f1 with a step size of 0.05. The space range is max(0,0.5-10×0.05):min(1,0.5+10×0.05), which is 0-1.

[0139] The third loop is executed, traversing the variable space of the second terminator contributing the Friction variable f2 with a step size of 0.05. The space range is max(0,0.5-10×0.05):min(1-f1,0.5+10×0.05), which is 0-1.

[0140] Execute the fourth - layer loop, traversing the variable space of the third - endmember contribution Fraction variable f3 with a step size of 0.05. The space range is max(0,0.5 - 10×0.05):min(1,0.5 + 10×0.05), that is, 0 - 1.

[0141] The fourth - endmember contribution Fraction variable f4 can be calculated from the variables f1, f2, and f3, that is, f4 = 1 - f1 - f2 - f3.

[0142] Judge whether the fourth endmember is positive. If it is positive, calculate the calculated value array of each parameter predicted = Σfi*endmember_data(i,:), and the residual R = sum(W.*(sample_data - predicted)^2). If the residual R < min_R, then update min_R with the R value and update the opt_f array to [f1, f2, f3, f4].

[0143] Return to the previous - layer loop, traverse the third - endmember contribution Fraction variable f3. After completion, return to the loop of the layer above again, and traverse the parameter spaces of all endmembers in turn to find the opt_f array when min_R is minimized (here assumed to be [0.6, 0.3, 0.05, 0.1]).

[0144] Return to the first - layer loop and refine the step size stepSizes[2]=0.01.

[0145] Execute the second - layer loop. The range of f1 is max(0,0.6 - 10×0.01):min(1,0.6 + 10×0.01), that is, 0.5 - 0.7.

[0146] Execute the third - layer loop. The range of f2 is max(0,0.3 - 10×0.01):min(1 - 0.6,0.3 + 10×0.01), that is, 0.2 - 0.4.

[0147] Execute the fourth - layer loop. The range of f3 is max(0,0.05 - 10×0.01):min(1 - f1 - f2,0.05 + 10×0.01), that is, 0 - 0.15.

[0148] Judge whether the fourth endmember is positive. If it is positive, calculate the calculated value array of each parameter predicted = Σfi*endmember_data(i,:), and the residual R = sum(W.*(sample_data - predicted)^2). If the residual R < min_R, then update min_R with the R value and update the opt_f array to [f1, f2, f3, f4].

[0149] Return to the first loop, refine the step size stepSizes[3] = 0.001, and repeat the above loop.

[0150] Output opt_f and min_R, return to step S3, compute the least squares solution for all samples in parallel, and assign the solution to the arrays opt_f and min_R.

[0151] like Figure 4 As shown, this embodiment of the invention provides a device for global analysis of multidimensional mixing processes in the ocean. The device runs on a computer, requiring an operating system and a Matlab program (Matlab R2017b or later is recommended). The storage unit stores the data file Data.xlsx and the result file MOMPA_Results_yyyymmdd_HHMM.xlsx. The computing core employs a CPU parallel processing architecture. To improve computational performance, it is recommended to further optimize the CPU clock speed and memory access rate.

[0152] Example 2:

[0153] Example 2 compares the computational efficiency of the optimization method of this invention with that of the traditional global exhaustive search method in the case of more complex endmember mixing in the East China Sea shelf region. The system was configured to process 6 endmembers, 13 parameters, and 8 samples. The test was conducted in the Matlab R2017b environment, and the hardware platform was a Lenovo Y7000P laptop (equipped with an 8-core CPU). The performance comparison shows that when the accuracy requirement is 0.1%, the optimization algorithm takes about 20 seconds to process 8 samples, with an average calculation time of less than 3 seconds per sample. In contrast, the traditional exhaustive search method takes 6 minutes and 37 seconds to complete the same calculation at 1% accuracy, which is 20 times longer than the method of this invention. If the accuracy of the exhaustive search method is increased to 0.1%, the calculation time is expected to increase exponentially (at least 10^6 times), completely losing its practical value. The method of this invention can still achieve a reduction of about 95% in calculation time even with a tenfold increase in accuracy.

[0154] Actual verification in typical sea areas such as the East China Sea shelf shows that when processing 8 samples under the condition of 6 endmembers and 13 parameters, this method can improve the calculation accuracy by ten times (to 0.1%) while still saving 95% of the calculation time compared with the traditional exhaustive method (20 seconds vs. 397 seconds). It provides a technical solution that combines theoretical rigor and engineering practicality for the precise quantitative study of marine mixing processes.

[0155] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Those skilled in the art should understand that appropriate modifications can be made to the embodiments or equivalent substitutions can be made to some technical features without departing from the spirit and essence of the technical solutions of the present invention, and such adjustments all fall within the protection scope of the claims of the present invention.

[0156] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered illustrative and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the scope of the invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

[0157] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. An optimization method for global analysis of oceanic multi-dimensional mixing processes, characterized in that, The method comprises the following steps: S1, data collection: obtaining multi-dimensional parameter data of end-member water mass and seawater sample; S2, data initialization processing 2.1), reading the collected multi-dimensional parameter data; 2.2), standardizing all the collected multi-dimensional parameter data; S3, data calculation: using a multi-level refinement architecture and a parallel computing framework to calculate the global optimal solution of the standardized data; S4, result analysis 4.1), evaluating the relative error of each parameter; 4.2), evaluating the influence degree of non-conservative process; S5, data output: generating an analysis report containing the mixing ratio of each end-member, residual error and parameter error.

2. The method of claim 1, wherein, The multi-dimensional parameter data in step S1 includes tracer parameter data and other non-conservative seawater parameter data, wherein the tracer data includes the temperature and salinity of the seawater sample, the multi-parameters of eleven rare earth elements and two hydrogen and oxygen isotopes in the seawater sample, and the multi-parameters of nutrients, dissolved oxygen and the like.

3. The method of claim 2, wherein, In step 2.1), the specific content of reading the collected multi-dimensional parameter data is: 2.1.1), reading the multi-dimensional parameter data of the end-member water mass and the measured multi-dimensional parameter data of the seawater sample, and assigning the end-member data and the sample data to the arrays endmember_data and sample_data; 2.1.2), verifying the validity of the read data; 2.1.3), reading the number of end-members, the number of parameters, the number of samples and the weight of each tracer parameter input by the user.

4. The method of claim 3, wherein, In step 2.2), the data is standardized by using the zscore method, and the specific content is: 2.2.1), assigning the data input by the user to variables e, n, m and W, wherein e is the number of end-members, n is the number of parameters, m is the number of samples, and W is the weight array of the parameters; 2.2.2), bringing the multi-dimensional parameter data into the following formula: endmember_norm=(endmember_data-endmember_mean) / endmember_std wherein endmember_mean represents the mean value of each parameter, endmember_std represents the standard deviation of each parameter, and endmember_norm represents the standardized end-member parameter, and endmember_data represents the original end-member parameter; 2.2.3), bringing the multi-dimensional parameter data into the following formula: sample_norm=(sample_data-endmember_mean) / endmember_std wherein sample_norm represents the standardized sample parameter, and sample_data represents the original sample parameter.

5. The method of claim 1, wherein, In step S3, the multi-level refinement architecture adopts a three-level step optimization mechanism, and the steps are 0.05, 0.01 and 0.001 in turn.

6. The method of claim 5, wherein, The parallel computing framework uses the parfor parallel computing algorithm of MATLAB to fully utilize the multi-core computing capability of CPU.

7. A method of optimization of global analysis of a multi-dimensional process in the ocean according to claim 1, 2, 3, 4, 5 or 6, characterized in that, In step S3, the specific content of global optimal solution calculation is: 3.1) Dynamically select the calculation function according to the number of endmembers input by the user, and calculate the different sample data according to the selected function; 3.2) Substitute endmember_norm, sample_norm, and W array into the selected function; 3.3) Calculate the global optimal solution by using the optimized exhaustive method and multi-level refinement search; 3.4) Output the global optimal solution set.

8. The method of claim 7, wherein, In the step 4.1), the specific content of evaluating the relative error of each tracer parameter is: 4.1.1) Obtain the tracer parameter measured data content in step 2.1), and obtain the contribution ratio of each endmember in step 3.4); 4.1.2) Obtain the theoretical calculation value by the product of the contribution ratio of each endmember and the endmember tracer measured data; 4.1.3) Extract the actual measured value from the sample_data array; 4.1.4) Compare and analyze the theoretical calculation value and the actual measured value, and quantitatively evaluate the influence degree of each tracer parameter in the multi-dimensional mixing process by the non-conservative process.

9. The method of claim 8, wherein, In the step 4.2), the specific content of evaluating the influence degree of the non-conservative process includes: 4.2.1) Obtain the non-conservative parameter measured data content in step 2.1), and obtain the contribution ratio of each endmember in step 3.4); 4.2.2) Obtain the theoretical calculation value by the product of the contribution ratio of each endmember and the endmember non-conservative parameter measured data; 4.2.3) Extract the actual measured value from the sample_data array; 4.2.4) Compare and analyze the theoretical calculation value and the actual measured value, and quantitatively evaluate the influence degree of each non-conservative parameter in the multi-dimensional mixing process by the non-conservative process.

Images