Method, program, device, and storage medium for correcting a local marine environment field information prediction model based on online observations

Abstract

The present application belongs to the technical field of marine environment modeling and prediction, and particularly relates to a method, program, device and storage medium for correcting a local marine environment field information prediction model based on online observation. The present application continuously obtains marine environment field information within an observation grid over a period of time through an environment perception module to obtain a time series of true values of marine environment field information of the observation grid, and then constructs an adaptive ARIMA model, and obtains a time series of observation estimated values of marine environment field information of the observation grid in a future period of time through the adaptive ARIMA model. According to the true values and the observation estimated values of the marine environment field information of the observation grid, a prediction error is calculated and its spatio-temporal influence range is determined, and finally the prediction values of the marine environment field information of the related grid are corrected in combination with a prediction bias time series and a neighborhood bias model.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of marine environment modeling and prediction, and specifically relates to a method, program, device and storage medium for correcting a local marine environment field information prediction model based on online observation. BACKGROUND

[0002] With the increasingly severe global climate change and marine environmental problems, the prediction and monitoring of marine environments have gradually become the focus of scientific research and engineering fields. In the prediction of marine environments, especially in terms of marine environmental fields and ocean dynamics, numerical simulation and remote sensing technologies have made certain progress. However, the dynamic changes of the ocean have strong nonlinearity and spatiotemporal complexity, which leads to many challenges in the accuracy and real-time performance of existing prediction methods.

[0003] Currently, most marine environment prediction systems rely on numerical simulation methods based on mathematical models. These methods usually predict future marine conditions based on observation data and preliminary marine states. For example, common numerical models include ocean general circulation models (OGCMs), regional ocean models (ROMS), and high-resolution coupled models. These models perform well in simulating large-scale ocean processes, but they often struggle to provide high enough accuracy when facing complex marine conditions in local areas. Specifically, the specific information of marine environments such as sea wind, sea wave, sea current, and solar radiation intensity varies greatly between different regions and time points, which significantly increases the difficulty of predicting local marine environments. In addition, traditional models usually have systematic errors and parameter uncertainties, which lead to certain biases and uncertainties in the prediction results.

[0004] Accurate prediction of marine environments has important implications for many engineering applications. For example, marine robots rely on accurate marine environment information to optimize their motion trajectories, avoid obstacles, and save energy consumption when performing path planning and track following tasks. If there are large errors in the environment prediction, it will directly affect the efficiency and safety of the robot task. In addition, the construction and maintenance of offshore wind farms, the optimization of shipping routes, the development and utilization of marine resources, and other engineering activities also rely on high-precision marine environment prediction to ensure their smooth progress and reduce operational risks.

[0005] In order to improve the accuracy of marine prediction, in recent years, researchers try to correct the prediction by fusing real-time observation data and simulation model. The introduction of observation data can effectively correct the deviation in model prediction, especially in the case of local area where the marine condition changes dramatically. By combining real-time observation information with the results of numerical model, the accuracy and adaptability of prediction can be greatly improved, especially in short-term prediction and local scale accurate prediction. However, the existing observation-based correction method still has some shortcomings in practical application. First of all, the traditional correction method mostly depends on statistical model or simple linear regression model, which lacks enough flexibility and adaptability for complex marine dynamic changes. Secondly, how to effectively integrate observation data from different sources and optimize in different spatial and temporal scales is still a technical problem to be solved. SUMMARY

[0006] The purpose of the present application is to overcome the problem that the existing marine environment prediction method cannot provide high-precision prediction when the local marine area faces complex nonlinear dynamic changes, especially the technical problem of lack of effective correction in short-term prediction and real-time updating process, and to provide a correction method, program, device and storage medium for local marine environment field information prediction model based on online observation.

[0007] The correction method for local marine environment field information prediction model based on online observation comprises the following steps:

[0008] Step 1: Obtain the grid map of the target sea area, and obtain the marine environment field information prediction value of each grid in a future period of time by using the real-time global environment field prediction model;

[0009] Step 2: Determine the grid where the environmental perception module is located in the grid map of the target sea area, denoted as the observation grid; obtain the real value time series of marine environment field information in the observation grid by continuously obtaining the marine environment field information in the observation grid through the environmental perception module in a period of time;

[0010] Step 3: Construct an adaptive ARIMA model according to the real value time series of marine environment field information in the observation grid, and obtain the observation estimation value time series of marine environment field information in the observation grid in a future period of time according to the adaptive ARIMA model;

[0011] Step 4: Calculate the prediction deviation of the real-time global environment field prediction model at each time point of the observation grid according to the real value and observation estimation value of the marine environment field information in the observation grid, and obtain the prediction deviation time series;

[0012] Step 5: Determine the spatio-temporal influence range of the prediction deviation, and establish a neighborhood deviation model by using the prediction deviation time series;

[0013] Step 6: Correct the predicted value of the marine environment field information of the observation grid according to the predicted bias time series;

[0014] According to the neighborhood bias model, the predicted value of the marine environment field information of the other grids in the spatio-temporal influence range of the predicted bias, except for the observation grid, is corrected.

[0015] Further, in step 3, the adaptive ARIMA model is constructed according to the true value time series of the marine environment field information of the observation grid, specifically:

[0016] Step 3.1: Determine the time window, preprocess the true value time series of the marine environment field information of the observation grid, and initialize the difference order d = 0;

[0017] Step 3.2: Test the stationarity of the true value time series of the marine environment field information of the observation grid. If it is not stationary, perform step 3.3; if it is stationary, perform step 3.4;

[0018] Step 3.3: After difference processing the true value time series of the marine environment field information of the observation grid according to the difference order d, set d = d + 1 and return to step 3.2;

[0019] Step 3.4: According to the parameter candidate set of the autoregressive order p and the moving average order q, use the BIC criterion to obtain the combination of the autoregressive order p and the moving average order q with the minimum BIC criterion value;

[0020] Step 3.5: Construct an adaptive ARIMA model according to the autoregressive order p, the moving average order q, and the difference order d.

[0021] Further, in step 3.2, the ADF test method is used to test the stationarity of the true value time series of the marine environment field information of the observation grid.

[0022] Further, the step 4 is specifically:

[0023]

[0024] Where (x1, y1) represents the position information of the observation grid; z represents the time, z1 represents the starting time when the environmental perception module starts observation, z2 represents the termination time when the environmental perception module ends observation, and z3 represents the maximum time for observation estimation by the preset adaptive ARIMA model; Δv wd (x1, y1, z) represents the predicted bias at time z; Lv wd (x1, y1, z) represents the marine environment field information obtained by the environmental perception module in the observation grid at time z; Yv represents the observation estimation value of the marine environment field information in the observation grid at time z obtained by the adaptive ARIMA model. wd (x1, y1, z) represents the prediction value of the marine environment field information in the observation grid at time z obtained by the real-time global environment field prediction model.

[0025] Further, the spatio-temporal influence range of the prediction bias in step 5 is a hemispherical region with (x1, y1, z1) as the center and a preset spatio-temporal influence distance r as the radius, that is, z≥z1 in the region.

[0026] Further, the neighborhood bias model in step 5 is:

[0027]

[0028] wherein, represents the neighborhood bias of the grid with position information (x i ,y i ) at time z; a is a real number with a value range of 0 to 1; l kt (x i ,y i ,z) represents the spatio-temporal distance of the grid with position information (x i ,y i ) at time z from the observation grid.

[0029]

[0030] wherein, l k (x i ,y i ) represents the straight-line distance of the grid with position information (x i ,y i ) from the observation grid; J z represents the set of the remaining grids except the observation grid within the spatio-temporal influence range of the prediction bias at time z.

[0031] Further, the step 6 is specifically:

[0032] correcting the prediction value of the marine environment field information of the observation grid according to the prediction bias time series;

[0033]

[0034] correcting the prediction value of the marine environment field information of the grids other than the observation grid within the spatio-temporal influence range of the prediction bias according to the neighborhood bias model;

[0035]

[0036] A computer device / equipment / system includes a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above-mentioned method for correcting the local ocean environment field information prediction model based on online observation.

[0037] A computer-readable storage medium stores a computer program / instruction, which, when executed by a processor, implements the steps of the method for correcting the local ocean environment field information prediction model based on online observation.

[0038] A computer program product includes a computer program / instruction, which, when executed by a processor, implements the steps of the method for correcting the local ocean environment field information prediction model based on online observation.

[0039] The beneficial effects of the present invention are:

[0040] The present invention can effectively improve the accuracy of short-term predictions of the marine environment, especially by performing real-time correction of prediction errors in complex local sea environments. This is of great significance for improving the path planning and trajectory tracking capabilities of autonomous systems such as marine robots, and at the same time provides more reliable technical support for marine resource development and environmental protection. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is an overall flow chart of the correction method of the local ocean environment field information prediction model based on online observation.

[0042] Figure 2 It is a flowchart for building an adaptive ARIMA model.

[0043] Figure 3 This is a conceptual diagram of the neighborhood deviation sphere model.

[0044] Figure 4 Yes k (x i ,y i ) is a schematic diagram of the calculation principle. DETAILED DESCRIPTION

[0045] The present invention will be further described below with reference to the accompanying drawings.

[0046] The traditional prediction method and observation method have advantages and disadvantages, and cannot meet the high requirements of data accuracy and collection range, resulting in certain deviation and uncertainty of the prediction result. Therefore, in order to improve the accuracy of local marine environment prediction, the application proposes a correction method of local marine environment field information prediction model based on online observation, which firstly initializes the grid map of fixed sea area, and obtains the prediction value of each grid by using the real-time global environment field prediction model; the real environment field information is obtained by observing a specific grid by a sensor, and an adaptive ARIMA model is constructed to generate future observation estimate value; the prediction error is obtained by calculating the difference between the observation value and the prediction value, the spatio-temporal influence range of the error is determined, the prediction value in the grid is corrected by using the neighborhood deviation model, and high-precision prediction correction of local marine environment is realized.

[0047] The application can effectively improve the accuracy of short-term prediction of marine environment field, especially real-time correction of prediction error in complex local sea area. This has important significance for improving the path planning and trajectory tracking ability of autonomous systems such as marine robots. At the same time, it provides more reliable technical support for marine resource development and environmental protection.

[0048] The correction method of local marine environment field information prediction model based on online observation comprises the following steps:

[0049] Step 1: Obtain the grid map of the target sea area, and obtain the marine environment field information prediction value of each grid in a future period of time by using the real-time global environment field prediction model;

[0050] Step 2: Determine the grid where the environment perception module is located in the grid map of the target sea area, and mark it as the observation grid; obtain the real value time series of the marine environment field information in the observation grid by continuously obtaining the marine environment field information in the observation grid in a period of time through the environment perception module;

[0051] Step 3: Construct an adaptive ARIMA model according to the real value time series of the marine environment field information of the observation grid, and obtain the observation estimate value time series of the marine environment field information of the observation grid in a future period of time according to the adaptive ARIMA model;

[0052] The adaptive ARIMA model is based on the traditional difference autoregressive moving average model (Autoregressive Integrated Moving Average Model, ARIMA), which analyzes the time series data in real time, and adaptively selects the parameters of the ARIMA model, including the difference order d, the autoregressive order p and the moving average order q, so as to avoid the tedious process of manual parameter adjustment and improve the accuracy of the model.

[0053] In the traditional ARIMA model construction, the order d, p and q of the model need to be determined by artificial observation, which is not conducive to automatic processing and is greatly affected by the observer's subjective judgment. In order to solve this problem, the present application introduces Bayesian Information Criterion (BIC), based on the goodness of fit of the model and the number of model parameters, to evaluate the fitting ability and prediction ability of the model. By comparing the BIC values of different models, the optimal model parameters are more accurately selected, and the adaptive ARIMA model is obtained, the specific steps are as follows:

[0054] Step 3.1: Determine the time window, preprocess the time series of the real value of the observed grid ocean environment field information, and initialize the difference order d=0;

[0055] Step 3.2: The ADF test method is used to test the stationarity of the time series of the real value of the observed grid ocean environment field information, if it is not stationary, step 3.3 is executed, if it is stationary, step 3.4 is executed;

[0056] Step 3.3: After the difference order d is determined, the difference order d is set to d+1, and the step 3.2 is returned;

[0057] Step 3.4: According to the parameter candidate set of autoregressive order p and moving average order q, the BIC criterion is used to obtain the combination of autoregressive order p and moving average order q with the minimum BIC criterion value;

[0058] Step 3.5: According to the autoregressive order p, the moving average order q and the difference order d, the adaptive ARIMA model is constructed;

[0059] Step 4: According to the real value and the observation estimate value of the observed grid ocean environment field information, the prediction deviation of the real-time global environment field prediction model of the observed grid at each time point is calculated, and the prediction deviation time series is obtained;

[0060]

[0061] Wherein, (x1, y1) represents the position information of the observed grid; z represents the time, z1 represents the starting time when the environmental perception module starts to observe, z2 represents the termination time when the environmental perception module ends to observe, and z3 represents the maximum time of observation estimation of the preset adaptive ARIMA model; Δv wd (x1, y1, z) represents the prediction deviation at z time; Lv wd (x1, y1, z) represents the ocean environment field information in the observed grid obtained by the environmental perception module at z time; Yv represents the observation estimation value of the marine environment field information in the observation grid at z time obtained by the adaptive ARIMA model; wd (x1,y1,z) represents the prediction value of the marine environment field information in the observation grid at z time obtained by the real-time global environment field prediction model;

[0062] In one embodiment provided by the application, z1=0, z2=0.5h and z3=2.5h can be taken.

[0063] Step 5: determining the spatio-temporal influence range of the prediction bias: taking (x1,y1,z1) as the center and the upper hemispherical region with a preset spatio-temporal influence distance r as the radius, that is, the region in which z≥z1;

[0064] establishing the neighborhood bias model by using the prediction bias time series:

[0065]

[0066] wherein, represents the neighborhood bias of the grid with the position information (x i ,y i ) at z time; a is a real number with the value range of 0 to 1; l kt (x i ,y i ,z) represents the spatio-temporal distance of the grid with the position information (x i ,y i ) at z time from the observation grid;

[0067]

[0068] wherein, l k (x i ,y i ) represents the straight-line distance of the grid with the position information (x i ,y i ) from the observation grid; J z represents the set of the grids other than the observation grid in the spatio-temporal influence range of the prediction bias at z time.

[0069] Step 6: correcting the prediction value of the marine environment field information of the observation grid according to the prediction bias time series;

[0070]

[0071] correcting the prediction value of the marine environment field information of the grids other than the observation grid in the spatio-temporal influence range of the prediction bias according to the neighborhood bias model;

[0072]

[0073] The application uses a prediction model to obtain the prediction value of the environmental field (meteorological / hydrological) information in each grid in the selected sea area grid map, and through the environmental perception module, the actual environmental field data of the fixed grid is observed and recorded in real time. Based on the historical real data, an adaptive ARIMA model is constructed to predict the future observation value, and then the prediction error is calculated and its spatio-temporal influence range is determined. Finally, combined with the prediction error and the neighborhood bias model, the environmental field prediction value of the related grid is dynamically corrected, thereby significantly improving the accuracy and reliability of the marine environment prediction, thereby effectively supporting the marine environment monitoring, disaster warning, marine assembly operation and other engineering applications.

[0074] The above only describes the preferred embodiments of the present application and is not intended to limit the present application. For those skilled in the art, the present application can have various modifications and changes. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A correction method for a local ocean environment field information prediction model based on online observation, characterized in that: The following steps are involved: Step 1: Obtain a grid map of the target sea area and use a real-time global environmental field prediction model to obtain the predicted value of the marine environmental field information for each grid in the future; Step 2: Determine the grid where the environmental perception module is located in the grid map of the target sea area, which is recorded as the observation grid; use the environmental perception module to continuously obtain the marine environmental field information in the observation grid over a period of time to obtain the time series of the true value of the marine environmental field information in the observation grid; Step 3: Build an adaptive ARIMA model based on the true value time series of the ocean environment field information of the observation grid, and obtain the time series of the estimated value of the ocean environment field information of the observation grid in the future period based on the adaptive ARIMA model; Step 4: Based on the true value and the estimated value of the ocean environment field information of the observation grid, calculate the prediction deviation of the real-time global environment field forecast model of the observation grid at each time point to obtain the prediction deviation time series; Where (x1, y1) represents the location information of the observation grid; z represents the time, z1 represents the starting time when the environment perception module starts observation, z2 represents the ending time when the environment perception module ends observation, and z3 represents the maximum time for the preset adaptive ARIMA model to perform observation estimation; Δv wd (x1, y1, z) represents the prediction deviation at time z; Lv wd (x1, y1, z) represents the ocean environment field information within the observation grid obtained by the environment perception module at time z; Yv represents the estimated value of the ocean environment field information in the observation grid at time z obtained through the adaptive ARIMA model; wd (x1, y1, z) represents the predicted value of the ocean environment field information in the observation grid at time z obtained through the real-time global environment field prediction model; Step 5: Determine the spatiotemporal impact range of the forecast deviation and establish a neighborhood deviation model using the forecast deviation time series; The spatiotemporal impact range of the prediction deviation is the upper hemispherical area with (x1, y1, z1) as the center and the preset spatiotemporal impact distance r as the radius, that is, z ≥ z1 in the area; The neighborhood bias model is: in, Indicates the position information as (x i ,y i ) grid at time z; a is a real number ranging from 0 to 1; l kt (x i ,y i ,z) indicates that the position information is (x i ,y i )’s grid at time z and the observation grid’s temporal and spatial distance; Among them, l k (x i ,y i ) indicates that the position information is (x i ,y i ) grid and the straight-line distance between the observation grid; J z It represents the set of grids other than the observation grid at time z within the spatiotemporal influence range of the prediction deviation; Step 6: According to the prediction deviation time series, the predicted value of the ocean environment field information of the observation grid is corrected; According to the neighborhood deviation model, the predicted values ​​of the ocean environment field information of grids other than the observation grid within the spatiotemporal influence range of the prediction deviation are corrected; 2. The method for correcting the local ocean environment field information prediction model based on online observation according to claim 1 is characterized in that: In step 3, an adaptive ARIMA model is constructed based on the real value time series of the ocean environment field information of the observation grid, specifically: Step 3.1: Determine the time window, preprocess the time series of the true value of the ocean environment field information of the observation grid, and initialize the difference order d = 0; Step 3.2: Check the stability of the time series of the true value of the ocean environment field information of the observation grid. If it is not stable, proceed to step 3.3; if it is stable, proceed to step 3.4; Step 3.3: After performing difference processing on the true value time series of the ocean environment field information of the observation grid according to the difference order d, set d = d + 1 and return to step 3.2; Step 3.4: Based on the parameter candidate set of the autoregressive order p and the moving average order q, use the BIC criterion to obtain the combination of the autoregressive order p and the moving average order q that has the smallest BIC criterion value; Step 3.5: Construct an adaptive ARIMA model based on the autoregressive order p, the moving average order q, and the differencing order d.

3. The method for correcting the local ocean environment field information prediction model based on online observation according to claim 2, characterized in that: In step 3.2, the ADF test method is used to test the stationarity of the time series of the true value of the ocean environment field information of the observation grid.

4. A computer device comprising a memory, a processor, and a computer program stored in the memory, wherein: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 3.

5. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

6. A computer program product comprising computer instructions, characterized in that: When the computer instructions are executed by a processor, the steps of the method according to any one of claims 1 to 3 are implemented.

Images