Sea wave height prediction method and device, storage medium and electronic equipment

Abstract

The embodiment of the invention provides a sea wave height prediction method and device, a storage medium and electronic equipment, and the method comprises the steps: fitting an autoregressive integrated moving average (ARIMA) model according to first height observation values of sea waves of a target region at a plurality of first time points of a first time period, predicting first height prediction values of the sea waves at a plurality of second time points in a second time period according to the ARIMA model; under the condition of determining that a nonlinear relationship exists in residual distribution corresponding to the ARIMA model, training a random forest model according to a residual corresponding to the ARIMA model and an environment variable of the target area in the first time period, and predicting second height prediction values of sea waves at a plurality of second time points in the second time period according to the trained random forest model; and according to the first height predicted value and the second height predicted value, determining a target height predicted value of the sea wave at a plurality of second time points of the second time period.

Description

Technical Field

[0001] This application relates to the field of ocean monitoring technologies, and in particular, to a method and device for predicting wave height, a storage medium, and an electronic device. Background Art

[0002] In fields such as offshore wind power, ocean engineering, and navigation, accurately predicting wave height is crucial for ensuring the safe operation of offshore facilities and improving the efficiency of offshore activities. Traditionally, the monitoring and prediction of wave height mainly rely on a single physical model and single-sensor data. For example, sensors such as buoys, radars, or laser rangefinders measure wave height using hydrodynamic models, the principle of electromagnetic wave reflection, or the principle of rectilinear propagation of light, respectively. However, these methods have exposed a series of technical problems in practical applications, especially the challenge of low calculation accuracy.

[0003] First of all, a single physical model often has difficulty accurately describing the wave characteristics under complex sea conditions. The formation of waves is affected by multiple factors such as wind speed, wind direction, ocean current, tide, and seabed topography, and it is difficult for a single physical model to comprehensively consider these variables. Especially when the sea conditions change violently or there are multiple wave systems interacting, the prediction accuracy of the model will decrease significantly.

[0004] Secondly, the limitations of a single sensor also affect the accuracy of the data. For example, although a buoy can provide direct wave height measurements, it may be damaged or displaced under extreme weather conditions, affecting the continuity and reliability of the data. Although radars and laser rangefinders have the advantages of long-distance monitoring and high-precision measurement, they are respectively restricted by the characteristics of electromagnetic wave propagation or the optical propagation environment. For example, in rain, fog, strong wind, or wave turbulence, the signal quality will decline, resulting in an increase in measurement error.

[0005] Regarding the problem of low calculation accuracy in the method of predicting wave height relying on a single physical model in the related art, no effective solution has been proposed yet.

[0006] Therefore, it is necessary to improve the related technologies to overcome the defects in the related technologies. Summary of the Invention

[0007] Embodiments of this application provide a method and device for predicting wave height, a storage medium, and an electronic device, so as to at least solve the problem of low calculation accuracy in the method of predicting wave height relying on a single physical model in the related art.

[0008] According to an embodiment of the present application, a method for predicting the sea wave height is provided, including: fitting an autoregressive integrated moving average (ARIMA) model according to the first height observation values of the sea waves in a target area at a plurality of first time points in a first time period, and predicting the first height prediction values of the sea waves at a plurality of second time points in a second time period according to the ARIMA model; in the case where a non-linear relationship exists in the residual distribution corresponding to the ARIMA model, training a random forest model according to the residuals corresponding to the ARIMA model and the environmental variables of the target area in the first time period, and predicting the second height prediction values of the sea waves at a plurality of second time points in the second time period according to the trained random forest model; determining the target height prediction values of the sea waves at a plurality of second time points in the second time period according to the first height prediction values and the second height prediction values.

[0009] In an exemplary embodiment, fitting an autoregressive integrated moving average (ARIMA) model according to the first height observation values of the sea waves in a target area at a plurality of first time points in a first time period includes: converting a plurality of first height observation values into a time series according to each first time point corresponding to each first height observation value; decomposing the time series to extract a trend component in the time series; removing the trend component in the time series through a differencing operation to obtain a differenced stationary series, and determining the differencing order corresponding to the ARIMA model; determining an autocorrelation graph and a partial autocorrelation graph corresponding to the differenced stationary series; determining the moving average order corresponding to the ARIMA model according to the autocorrelation graph, and determining the autoregressive order corresponding to the ARIMA model according to the partial autocorrelation graph; fitting the ARIMA model according to the differencing order, the moving average order, and the autoregressive order.

[0010] In an exemplary embodiment, before training the random forest model according to the residuals and the environmental variables of the target area in the first time period, the method further includes: obtaining second height observation values of the sea waves at a plurality of third time points in a third time period before the first time period, and inputting the second height observation values into the ARIMA model to enable the ARIMA model to output third height prediction values of the sea waves at a plurality of first time points in the first time period; determining the residuals corresponding to the ARIMA model according to the first height observation values and the third height prediction values; determining whether a non-linear relationship exists in the residual distribution corresponding to the ARIMA model according to the residuals.

[0011] In an exemplary embodiment, training the random forest model according to the residual and the environmental variables in the target area during the first time period includes: performing feature learning on the environmental variables to determine the feature data corresponding to the environmental variables, and determining the initial hyperparameters of the random forest model, where the initial hyperparameters include at least one of the following: the number of decision trees included in the random forest model, the minimum number of samples for splitting nodes corresponding to the random forest model, and the minimum number of samples for leaf nodes; constructing decision trees in the random forest model based on the feature data, the environmental variables, and the residual; inputting the residual and the environmental variables into the random forest model to enable the random forest model to further predict the residual and obtain a target residual prediction value; and training the random forest model according to the residual and the target residual prediction value.

[0012] In an exemplary embodiment, training the random forest model according to the residual and the target residual prediction value includes: determining a first error metric between the residual and the target residual prediction value; in the case where it is determined that the first error metric is greater than or equal to a preset threshold, repeatedly performing an update operation until it is determined that the first error metric is less than the preset threshold, where the update operation is: updating the initial hyperparameters through grid search, and determining the residual prediction value corresponding to the random forest with the updated initial hyperparameters as the target residual prediction value; in the case where it is determined that the first error metric is less than the preset threshold, stopping training the random forest model.

[0013] In an exemplary embodiment, predicting the second height prediction values of the sea wave at multiple second time periods during the second time period according to the trained random forest model includes: inputting the first height observation value and the environmental variables into the trained random forest model to determine each prediction value of each decision tree in the random forest model; determining the second height prediction value according to a first formula, where the first formula is: is the second height prediction value, N is the number of decision trees included in the random forest model, and T i(x) is each prediction value.

[0014] In an exemplary embodiment, determining the target height prediction values of the sea wave at multiple second time points during the second time period according to the first height prediction value and the second height prediction value includes: determining a first weight value corresponding to the ARIMA model and a second weight value of the trained random forest model; determining the target height prediction value according to a second formula, where the second formula is: Y is the target height prediction value, is the first height prediction value, and α1 is the first weight value. is the second height prediction value, and α2 is the second weight value.

[0015] According to another embodiment of the present application, there is provided a device for predicting the sea wave height, including: a fitting module, configured to fit an autoregressive integrated moving average (ARIMA) model according to the first height observation values at multiple first time points of the sea waves in a target area within a first time period, and predict the first height prediction values at multiple second time points of the sea waves in a second time period according to the ARIMA model; a training module, configured to train a random forest model according to the residuals corresponding to the ARIMA model and the environmental variables of the target area within the first time period when it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, and predict the second height prediction values at multiple second time points of the sea waves in the second time period according to the trained random forest model; a determination module, configured to determine the target height prediction values at multiple second time points of the sea waves in the second time period according to the first height prediction values and the second height prediction values.

[0016] According to still another embodiment of the present application, there is also provided a computer-readable storage medium, in which a computer program is stored, and the computer program is configured to execute the steps in any one of the above method embodiments when running.

[0017] According to still another embodiment of the present application, there is also provided an electronic device, including a memory and a processor, a computer program is stored in the memory, and the processor is configured to run the computer program to execute the steps in any one of the above method embodiments.

[0018] According to still another embodiment of the present application, there is also provided a computer program product, including a computer program, and the computer program implements the steps in any one of the above method embodiments when executed by a processor.

[0019] According to the embodiments of the present application, an autoregressive integrated moving average model (ARIMA model) is fitted based on the first height observation values of the ocean waves in the target area at multiple first time points in the first time period, and the first height prediction values of the ocean waves at multiple second time points in the second time period are predicted according to the ARIMA model; when it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, a random forest model is trained according to the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and the second height prediction values of the ocean waves at multiple second time points in the second time period are predicted according to the trained random forest model; the target height prediction values of the ocean waves at multiple second time points in the second time period are determined according to the first height prediction values and the second height prediction values. That is to say, the embodiments of the present application predict the height of the ocean waves through multiple models (i.e., the ARIMA model and the random forest model), and then determine the final prediction result of the ocean wave height according to the prediction results of the multiple models. By combining the short-term dynamic prediction ability of the ARIMA model with the non-linear fitting ability of the random forest model in the embodiments of the present application, the problem of low calculation accuracy in the method of predicting the ocean wave height by relying on a single physical model in the related art is solved, and the prediction accuracy is improved by predicting the ocean wave height through the ARIMA model and the random forest model. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] The accompanying drawings herein are incorporated into the specification and form a part of the specification, showing embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application.

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the accompanying drawings required for use in the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, other drawings can also be obtained based on these drawings without creative efforts.

[0022] Figure 1 It is a hardware structure block diagram of a computer terminal device for a method of predicting the height of ocean waves according to an embodiment of the present application;

[0023] Figure 2 It is a flowchart of a method of predicting the height of ocean waves according to an embodiment of the present application;

[0024] Figure 3 It is a structure block diagram of a device for predicting the height of ocean waves according to an embodiment of the present application. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0025] The embodiments of the present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0026] It should be noted that the terms "first", "second", etc. in the description, claims and the above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily describe a specific order or sequence.

[0027] The method embodiments provided in the embodiments of this application can be executed in a computer terminal device or a similar computing device. Taking the operation on a computer terminal device as an example, Figure 1 is a hardware structure block diagram of a computer terminal device for a method of predicting sea wave height according to an embodiment of this application. As Figure 1 shown, the computer terminal device may include one or more ( Figure 1 only one is shown in Figure 1 figures) processors 102 (the processors 102 may include, but are not limited to, processing devices such as microprocessors or programmable logic devices FPGA, etc.) and a memory 104 for storing data. Among them, the above-mentioned computer terminal device may further include a transmission device 106 for communication functions and an input / output device 108. Those of ordinary skill in the art can understand that Figure 1 the structure shown in Figure 1 figures is only schematic and does not limit the structure of the above-mentioned computer terminal device. For example, the computer terminal device may further include more or fewer components than

[0028] shown in

[0029] The memory 104 can be used to store computer programs, for example, software programs and modules of application software, such as the computer program corresponding to the method of predicting sea wave height in the embodiments of this application. The processor 102 executes various functional applications and data processing by running the computer programs stored in the memory 104, that is, the above-mentioned method is realized. The memory 104 may include a high-speed random access memory, and may further include a non-volatile memory, such as one or more magnetic storage devices, flash memories, or other non-volatile solid-state memories. In some instances, the memory 104 may further include a memory remotely disposed relative to the processor 102, and these remote memories may be connected to the computer terminal device through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an enterprise intranet, a local area network, a mobile communication network, and combinations thereof.The transmission device 106 is used to receive or send data via a network. Specific examples of the above-mentioned network may include a wireless network provided by a communication provider of a computer terminal device. In one example, the transmission device 106 includes a network adapter (Network Interface Controller, abbreviated as NIC), which can be connected to other network devices through a base station so as to communicate with the Internet. In one example, the transmission device 106 may be a Radio Frequency (RF) module, which is used to communicate with the Internet wirelessly.

[0030] This embodiment provides a method for predicting the wave height. Figure 2 It is a flowchart of the method for predicting the wave height according to the embodiment of the present application, as Figure 2 shown. The process includes the following steps:

[0031] Step S202: Fit an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the waves in the target area at multiple first time points in the first time period, and predict the first height prediction values of the waves at multiple second time points in the second time period according to the ARIMA model;

[0032] Step S204: In the case of determining that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, train a random forest model based on the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and predict the second height prediction values of the waves at multiple second time points in the second time period according to the trained random forest model;

[0033] Step S206: Determine the target height prediction values of the waves at multiple second time points in the second time period according to the first height prediction values and the second height prediction values.

[0034] Through the above steps, an autoregressive integrated moving average model (ARIMA model) is fitted based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and the first height prediction values of the sea waves at multiple second time points in the second time period are predicted according to the ARIMA model; in the case where a non-linear relationship exists in the residual distribution corresponding to the ARIMA model, the random forest model is trained according to the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and the second height prediction values of the sea waves at multiple second time points in the second time period are predicted according to the trained random forest model; the target height prediction values of the sea waves at multiple second time points in the second time period are determined according to the first height prediction values and the second height prediction values. That is to say, in the embodiment of the present application, the height of the sea waves is predicted by multiple models (i.e., the ARIMA model and the random forest model), and then the final prediction result of the sea wave height is determined according to the prediction results of the multiple models. By combining the short-term dynamic prediction ability of the ARIMA model with the non-linear fitting ability of the random forest model in the embodiment of the present application, the problem of low calculation accuracy existing in the method of predicting the sea wave height relying on a single physical model in the related art is solved, and then the height of the sea waves is predicted by the ARIMA model and the random forest model, thereby improving the prediction accuracy.

[0035] Optionally, step S202 of fitting an autoregressive integrated moving average ARIMA model according to the first height observation values of the sea waves in the target area at multiple first time points in the first time period includes: converting the multiple first height observation values into a time series according to each first time point corresponding to each first height observation value; decomposing the time series to extract the trend component in the time series; removing the trend component in the time series through a differencing operation to obtain a differenced stationary series, and determining the differencing order corresponding to the ARIMA model; determining the autocorrelation plot and partial autocorrelation plot corresponding to the differenced stationary series; determining the moving average order corresponding to the ARIMA model according to the autocorrelation plot, and determining the autoregressive order corresponding to the ARIMA model according to the partial autocorrelation plot; fitting the ARIMA model according to the differencing order, the moving average order, and the autoregressive order.

[0036] It can be understood that the ARIMA model is used to process data with trends and seasonality. The following is the method for fitting the ARIMA model:

[0037] Time series conversion: For the first height observation values of the sea waves in the target area at multiple first time points in the first time period, arrange them in chronological order to construct a time series data.

[0038] Trend decomposition: Decompose the constructed time series of wave height, aiming to isolate the trend component within it. The trend component reflects the long-term variation trend of wave height over time and is usually extracted by fitting with linear, polynomial, or exponential functions.

[0039] Differencing operation and stationarity test: To fit an ARIMA model, it is necessary to check whether the time series is stationary. A stationary series is one whose statistical properties (such as mean and variance) do not change over time. If the time series has an obvious trend, a differencing operation is usually required to remove the trend component to obtain a stationary series. The choice of the differencing order needs to be based on the results of the stationarity test of the data (such as the ADF test) to determine whether the data has reached a stationary state or how many times of differencing are required to achieve stationarity.

[0040] Determine the autocorrelation function (ACF) and partial autocorrelation function (PACF): Calculate the ACF and PACF of the differenced stationary series. The ACF shows the correlation between data at different time points in the series, while the PACF shows the independent correlation between the data at a certain time point and the data at other time points, that is, the correlation after removing the influence of intermediate time points. By observing the ACF and PACF, the autoregressive and moving average characteristics of the series can be extracted to determine the parameters of the model.

[0041] Determine the parameters of the ARIMA model: Determine the order of the moving average (MA) part in the ARIMA model, that is, the q value of MA(q), according to the ACF; determine the order of the autoregressive (AR) part, that is, the p value of AR(p), according to the PACF. At the same time, combined with the previously determined differencing order (d), the parameter configuration of the ARIMA(p, d, q) model is finally obtained.

[0042] Fit the ARIMA model: Use the determined differencing order d, moving average order q, and autoregressive order p to fit the ARIMA model to the differenced stationary series. The fitting process estimates the model parameters by minimizing the prediction error of the model (such as the sum of squared residuals).

[0043] Optionally, before training the random forest model according to the residuals and the environmental variables of the target area in the first time period in step S204 above, the method further includes: obtaining the second height observation values at multiple third time points in the third time period before the first time period of the wave, and inputting the second height observation values into the ARIMA model so that the ARIMA model outputs the third height prediction values for multiple first time points of the wave in the first time period; determining the residuals corresponding to the ARIMA model according to the first height observation values and the third height prediction values; and determining whether there is a non-linear relationship in the residual distribution corresponding to the ARIMA model according to the residuals.

[0044] It is understandable that the ARIMA model can be used to predict the wave height, and the linear prediction ability of the ARIMA model can be evaluated through the prediction residuals, so as to judge whether it is necessary to introduce a non-linear model (such as a random forest) for further optimization. Specifically:

[0045] Obtaining the predicted values of the ARIMA model: Before training the random forest model, the ARIMA model is first used to predict the wave height. Specifically, multiple second height observations at multiple third time points of the waves in the third time period (i.e., a time window before the first time period), that is, historical wave height data, are collected. These historical data are input into the previously trained ARIMA model, so that the model outputs the third height predicted values for multiple first time points within the first time period (i.e., the current target prediction time period).

[0046] Determining the residuals of the ARIMA model: Next, the third height predicted values of the ARIMA model are compared with the actual first height observations within the first time period, and the difference between the two, that is, the residuals, is calculated. The residuals reflect the accuracy of the ARIMA model prediction. If the residuals are small, it indicates that the model fitting effect is good; if the residuals are large, it indicates that the model prediction ability is limited, and there may be uncaught non-linear relationships or complex patterns.

[0047] Evaluating non-linear relationships in the residual distribution: Further analyze the residual distribution of the ARIMA model to judge whether there are non-linear relationships. This usually uses statistical methods, such as scatter plot analysis of residuals, autocorrelation plot and partial autocorrelation plot analysis of residuals, and non-linear tests (such as Kolmogorov-Smirnov test, Anderson-Darling test, etc.). If the residual distribution shows obvious non-linear laws or there are complex patterns, then only using the ARIMA model may not be able to fully capture these characteristics, and at this time, non-linear models such as random forests need to be introduced for supplementary training and optimization.

[0048] Introducing a random forest for training: If the evaluation results show that there are non-linear relationships in the residual distribution of the ARIMA model, then the residuals and the environmental variables (such as wind speed, wind direction, tide, etc.) in the target area within the first time period are used as inputs to train the random forest model. The random forest is a machine learning method based on decision trees, which is particularly suitable for dealing with non-linear relationships and high-dimensional feature data. By training the random forest model, its non-linear prediction ability can be used to make up for the linear prediction limitations of the ARIMA model, and improve the overall prediction accuracy and the robustness of the model.

[0049] Optionally, training the random forest model according to the residual and the environmental variables of the target area in the first time period in step S202 includes: performing feature learning on the environmental variables to determine the feature data corresponding to the environmental variables, and determining the initial hyperparameters of the random forest model, where the initial hyperparameters include at least one of the following: the number of decision trees included in the random forest model, the minimum number of samples for splitting nodes, and the minimum number of samples for leaf nodes corresponding to the random forest model; constructing decision trees in the random forest model based on the feature data, the environmental variables, and the residual; inputting the residual and the environmental variables into the random forest model so that the random forest model further predicts the residual to obtain a target residual prediction value; and training the random forest model according to the residual and the target residual prediction value.

[0050] Among them, training the random forest model according to the residual and the target residual prediction value includes: determining a first error metric between the residual and the target residual prediction value; in the case where it is determined that the first error metric is greater than or equal to a preset threshold, repeatedly performing an update operation until it is determined that the first error metric is less than the preset threshold, where the update operation is: updating the initial hyperparameters through grid search and determining the residual prediction value corresponding to the random forest after updating the initial hyperparameters as the target residual prediction value; and in the case where it is determined that the first error metric is less than the preset threshold, stopping training the random forest model.

[0051] It can be understood that the residual of the ARIMA model and the environmental variables of the target area can be combined, and the random forest model can be used for training to improve the prediction accuracy of the sea wave height. Specifically:

[0052] Feature learning and hyperparameter initialization: Perform feature learning on the environmental variables of the target area in the first time period, aiming to identify and extract the feature data most relevant to the sea wave height change in these variables, such as wind speed, wind direction, ocean current, water temperature, etc. At the same time, determine the initial hyperparameters of the random forest model, including the number of decision trees included in the model, the minimum number of samples for splitting nodes, and the minimum number of samples for leaf nodes.

[0053] Decision tree construction: Based on the feature data, environmental variables, and the residual of the ARIMA model, construct decision trees in the random forest model. Each tree will be trained using randomly selected features and data samples to learn the non-linear relationship of the sea wave height change. The construction process of the decision tree follows recursive splitting, and each node makes a decision based on different environmental variable features until the minimum number of samples for leaf nodes is reached or further splitting is not possible.

[0054] Model Prediction and Evaluation: The residuals of the ARIMA model are used as the target variable and input into the random forest model together with the environmental variables. The random forest model further predicts the residuals to obtain the predicted values of the target residuals. The prediction ability of the random forest model is evaluated by the error between the predicted residual values and the actual residuals.

[0055] Model Training and Hyperparameter Optimization: The random forest model is trained based on the first error metric (such as mean squared error or mean absolute error) between the residuals and the predicted target residuals. If the first error metric is greater than or equal to a preset threshold (i.e., the gap between the residuals predicted by the model and the actual residuals exceeds the acceptable range), then the hyperparameter update operation is performed. This is done through grid search (GS), that is, within a predefined hyperparameter range, all possible combinations are exhausted, the performance of the model under each combination is evaluated, and the best combination is selected. The specific operation is to update the number of decision trees, the minimum number of samples for splitting nodes, and the minimum number of samples for leaf nodes, and use the predicted residual values corresponding to the updated random forest model as the new predicted target residual values. The evaluation and update are performed iteratively until the first error metric is less than the preset threshold, indicating that the prediction ability of the model is good enough and the training can be stopped.

[0056] Training Termination: Once the prediction error (the first error metric) of the model is below the preset threshold, it means that the random forest model has been able to effectively capture and explain the non - linear relationships and complex patterns that the ARIMA model cannot handle. At this time, the training process of the model is stopped.

[0057] Optionally, predicting the second height prediction values of the sea wave in the second time period according to the trained random forest model in step S204 includes: inputting the first height observation values and the environmental variables into the trained random forest model to determine each prediction value of each decision tree in the random forest model; determining the second height prediction value according to the first formula, where the first formula is: is the second height prediction value, N is the number of decision trees included in the random forest model, and T i(x) is each prediction value.

[0058] It can be understood that the trained random forest model can be used to predict the sea wave height to obtain the second height prediction value. Specifically:

[0059] Generation of predicted values: The first height observation values of the target area at multiple first time points in the first time period, as well as the environmental variables (such as wind speed, wind direction, tide, etc.) at these time points, are used as input data and input into the trained random forest model. The random forest model consists of multiple decision trees. Each decision tree will generate a predicted value for the input data according to its own training characteristics and splitting rules. This means that for each set of input data, each decision tree in the random forest model will give a prediction result of the wave height.

[0060] Integration of predicted values: The random forest model generates the final predicted value by integrating the prediction results of these decision trees. The predicted values of each decision tree will be collected, and then the second height predicted values of the waves at multiple second time points in the second time period are determined according to the first formula. The first formula is: Where, represents the second height predicted value at the second time point t in the second time period; T is the number of decision trees in the random forest model; T i(x) represents the predicted value of the i-th decision tree at the time point t.

[0061] Optionally, determining the target height predicted values of the waves at multiple second time points in the second time period according to the first height predicted value and the second height predicted value in step S206 includes: determining the first weight value corresponding to the ARIMA model and the second weight value of the trained random forest model; determining the target height predicted value according to the second formula, where the second formula is: Y is the target height predicted value, is the first height predicted value, α1 is the first weight value, is the second height predicted value, α2 is the second weight value.

[0062] It can be understood that the prediction results of the ARIMA model and the random forest model can be combined, and the final target height predicted value can be determined through weight allocation to improve the accuracy and reliability of the prediction. Specifically:

[0063] Determination of model weights: In order to integrate the prediction results of the ARIMA model (linear prediction model) and the random forest model (non-linear prediction model), the weights of the two need to be determined. The first weight value (α1) and the second weight value (α2) respectively correspond to the relative importance of the ARIMA model and the random forest model. The determination of the weights can be based on the prediction performance of the models on the validation set. For example, by comparing the mean squared error, mean absolute error, etc. of the two, the better model is selected and given a higher weight. More complex methods, such as grid search, Bayesian optimization, etc., can also be used to find the optimal weight combination.

[0064] Fusion of predicted values: Using the determined weight values, fuse the first height predicted value of the ARIMA model and the second height predicted value of the random forest model, and calculate the final target height predicted value through the following second formula:

[0065] To better understand the process of the above-mentioned method for predicting wave height, the following further describes the implementation method flow of the above-mentioned wave height prediction in combination with optional embodiments, but it is not used to limit the technical solutions of the embodiments of the present application.

[0066] Wave height is an important indicator characterizing the marine dynamic environment. Real-time monitoring and prediction play a crucial role in ensuring the operation of offshore wind farms, shipping safety, and extreme weather warnings. Traditional methods rely on physical models and single sensors, suffering from problems such as complex calculations, poor real-time performance, and limited accuracy.

[0067] To address the above technical problems, the optional embodiments of the present application combine statistical time series modeling and machine learning regression analysis, providing a wave height monitoring method based on time series modeling and random forest regression. It mainly combines time series modeling and machine learning, combines the short-term dynamic prediction ability of the time series model with the non-linear fitting ability of the random forest regression, breaking through the limitations of traditional single methods. Moreover, various statistical features (such as trends, periods, frequencies) are extracted from the time series of wave height, combined with environmental factors (such as wind speed, air pressure, etc.) to comprehensively characterize the wave change pattern; in addition, by dynamically updating the random forest model, the adaptability of the system under different seasons, climates, and regions is ensured, improving the monitoring accuracy and reliability. That is to say, the optional embodiments of the present application are superior to traditional methods in terms of accuracy, robustness, and prediction ability, and can be widely applied to scenarios such as offshore wind farms, port construction, and marine resource development. Specifically:

[0068] Step 1: Data collection and transmission.

[0069] Multi-source data collection:

[0070] Wave height sensor: Use the ultrasonic or laser ranging equipment of the buoy system to collect real-time wave height data (i.e., the first height observation values at multiple first time points in the first time period), which serve as the core input variables of the time series.

[0071] Environmental variable sensor: Obtain relevant environmental variables such as wind speed, wind direction, air pressure, and sea temperature through a meteorological station to assist in describing the external factors affecting wave changes.

[0072] Data transmission and storage: Transmit the data to the central server using wireless transmission or satellite communication. Build a distributed database on the server side to store historical time series data for subsequent modeling and analysis.

[0073] Step 2: Data preprocessing and feature extraction.

[0074] Data cleaning: Check and remove missing values and outliers, and fill in the missing data using linear interpolation or moving average method.

[0075] For mutation anomalies (such as instrument false alarms), use an outlier detection algorithm based on the interquartile range (IQR) to remove the influencing factors.

[0076] Time series decomposition: Decompose the time series of wave height to extract the trend component, periodic component, and random perturbation component;

[0077] Trend component: Capture the overall change trend through the moving average method.

[0078] The trend component represents the long-term trend of the time series, such as linear growth, decline, or more complex trends. In the case of wave height, the trend component reflects the long-term change of wave height over time and may contain information such as seasonal changes and long-term climate patterns.

[0079] Periodic component: Use the fast Fourier transform to extract the main fluctuation frequency and amplitude information.

[0080] The periodic component focuses on the recurring periodic patterns in the time series, such as daily changes, annual changes, or other periodic fluctuations. The periodic component may involve regular changes such as tidal effects and seasonal wave patterns in wave monitoring.

[0081] Random component: Analyze the distribution and characteristics of the noise.

[0082] The random component covers the random fluctuations and noise in the time series that cannot be explained by trends or periodicity. The random component is usually considered white noise, but in some cases, it may exhibit non-linear characteristics or have certain specific distribution patterns.

[0083] Statistical feature extraction:

[0084] Extract the following features from the original time series: mean, variance, kurtosis, skewness; wave period, significant wave height, wave crest and wave trough distribution; change rate of local maximum and minimum values.

[0085] Step 3: Time series modeling.

[0086] Model selection and fitting: Autoregressive Integrated Moving Average model (ARIMA model): Model the trend and stationary series separately to predict the change trend of short-term wave height.

[0087] The model parameters are selected by the Akaike Information Criterion (AIC) and the Bayesian Information Criterion (BIC) to ensure the generalization ability and stability of the model.

[0088] Residual analysis: Check the residual distribution of the ARIMA model to ensure that the random component follows the white noise property. If there is a non-linear pattern in the residuals, enter the stage of random forest regression (i.e., random forest model) to further optimize the prediction results.

[0089] Step 4, Random forest regression modeling.

[0090] Model training: Use the residuals of the time series and the environmental variables as the input of the random forest model, and the actual value of the wave height as the output. The random forest uses the integrated structure of decision trees to capture the non-linear relationship between the input features and the wave height.

[0091] Formula description: Where, N: The number of decision trees in the random forest; T i(x) : The predicted output of the i-th decision tree; The final prediction result of the random forest.

[0092] The model is optimized through the following steps:

[0093] Feature importance analysis: Calculate the contribution degree of the input variables according to the Gini Coefficient (Gini) impurity or information gain.

[0094] Hyperparameter tuning: Optimize parameters such as the number of decision trees, maximum depth, and minimum splitting nodes through grid search.

[0095] Online update: Use the latest data to adjust the random forest model in real time to ensure adaptability to dynamic sea conditions.

[0096] Step 5, Multi-step prediction and dynamic adjustment.

[0097] Multi-step prediction: Gradually input the predicted values of the time series into the model to achieve the prediction of wave heights at multiple future time points.

[0098] Dynamically adjust the weights to fuse the results of short-term prediction (e.g., ARIMA model) and long-term prediction (e.g., random forest).

[0099] Real-time update mechanism: As new data is added, the system retrains the random forest model every fixed time.

[0100] Optimize the parameters of the ARIMA model using the Bayesian update method to ensure the accuracy of the prediction results.

[0101] Step 6, Extreme wave event detection.

[0102] Threshold monitoring: Set a dynamic threshold for the significant wave height to identify possible extreme wave events in real time.

[0103] Use the classification ability of the random forest to classify and alarm the detected abnormal wave events.

[0104] Short-term warning: Combine time series prediction and environmental factor analysis to issue an extreme wave warning in advance and provide decision-making support for ocean operations.

[0105] Step 7, System integration and remote monitoring.

[0106] Cloud platform support: Integrate data and models into the cloud platform to achieve remote access and real-time visualization.

[0107] The platform supports historical data playback, statistical analysis, and dynamic trend display.

[0108] Cross-departmental collaboration: The monitoring results can be directly used in multiple scenarios such as ship scheduling, wind power equipment control, and port management.

[0109] Provide an API interface to link with the weather forecast and disaster emergency system to form an integrated solution.

[0110] In summary, the optional embodiments of the present application achieve real-time monitoring and multi-step prediction of wave height through the combination of time series modeling and random forest regression. Through the extraction and fusion of multi-dimensional features, the combination of short-term dynamic prediction and long-term non-linear fitting, and the introduction of a dynamic update mechanism. Compared with traditional methods, the optional embodiments of the present application have significantly improved monitoring accuracy, application scope, and extreme event response capabilities, and can be widely used in the fields of marine economy and safety guarantee.

[0111] Through the description of the above embodiments, those skilled in the art can clearly understand that the method according to the above embodiments can be implemented by means of software plus a necessary general hardware platform. Of course, it can also be implemented by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. The computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), including several instructions for causing a terminal device (which can be a mobile phone, computer, server, or network device, etc.) to execute the methods described in the various embodiments of the present application.

[0112] In this embodiment, a device for predicting the sea wave height is further provided. This device is used to implement the above-mentioned embodiments and preferred implementation manners, and those that have been described will not be repeated. As used hereinafter, the term "module" can be a combination of software and / or hardware that can achieve a predetermined function. Although the devices described in the following embodiments are preferably implemented in software, implementation in hardware, or a combination of software and hardware is also possible and contemplated.

[0113] Figure 3 is a structural block diagram of a device for predicting the sea wave height according to an embodiment of the present application. As Figure 3 shown, the device includes:

[0114] A fitting module 32, configured to fit an autoregressive integrated moving average (ARIMA) model according to the first height observation values at multiple first time points of the sea waves in the target area during a first time period, and predict the first height prediction values at multiple second time points of the sea waves during a second time period according to the ARIMA model;

[0115] A training module 34, configured to train a random forest model according to the residuals corresponding to the ARIMA model and the environmental variables of the target area during the first time period when it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, and predict the second height prediction values at multiple second time points of the sea waves during the second time period according to the trained random forest model;

[0116] A determination module 36, configured to determine the target height prediction values at multiple second time points of the sea waves during the second time period according to the first height prediction values and the second height prediction values.

[0117] Through the above device, an autoregressive integrated moving average model (ARIMA model) is fitted based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and the first height prediction values of the sea waves at multiple second time points in the second time period are predicted according to the ARIMA model; in the case where a non-linear relationship exists in the residual distribution corresponding to the ARIMA model, the random forest model is trained according to the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and the second height prediction values of the sea waves at multiple second time points in the second time period are predicted according to the trained random forest model; the target height prediction values of the sea waves at multiple second time points in the second time period are determined according to the first height prediction values and the second height prediction values. That is to say, in the embodiments of the present application, the height of the sea waves is predicted through multiple models (i.e., the ARIMA model and the random forest model), and then the final prediction result of the sea wave height is determined according to the prediction results of the multiple models. By combining the short-term dynamic prediction ability of the ARIMA model with the non-linear fitting ability of the random forest model in the embodiments of the present application, the problem of low calculation accuracy in the method of predicting the sea wave height relying on a single physical model in the related art is solved, and thus the height of the sea waves is predicted through the ARIMA model and the random forest model, improving the prediction accuracy.

[0118] In an exemplary embodiment, the fitting module 32 is further configured to convert the multiple first height observation values into a time series according to each first time point corresponding to each first height observation value; decompose the time series to extract a trend component in the time series; remove the trend component in the time series through a difference operation to obtain a stationary series after difference, and determine the difference order corresponding to the ARIMA model; determine an autocorrelation graph and a partial autocorrelation graph corresponding to the stationary series after difference; determine the moving average order corresponding to the ARIMA model according to the autocorrelation graph, and determine the autoregressive order corresponding to the ARIMA model according to the partial autocorrelation graph; fit the ARIMA model according to the difference order, the moving average order, and the autoregressive order.

[0119] In an exemplary embodiment, the fitting module 32 is further configured to obtain the second height observation values of the sea waves at multiple third time points in a third time period before the first time period, and input the second height observation values into the ARIMA model, so that the ARIMA model outputs the third height prediction values of the sea waves at multiple first time points in the first time period; determine the residuals corresponding to the ARIMA model according to the first height observation values and the third height prediction values; determine whether there is a non-linear relationship in the residual distribution corresponding to the ARIMA model according to the residuals.

[0120] In an exemplary embodiment, the training module 34 is further configured to perform feature learning on the environmental variables to determine feature data corresponding to the environmental variables, and determine initial hyperparameters of the random forest model, where the initial hyperparameters at least include one of the following: the number of decision trees included in the random forest model, the minimum number of samples for splitting nodes corresponding to the random forest model, and the minimum number of samples for leaf nodes; construct decision trees in the random forest model based on the feature data, the environmental variables, and the residuals; input the residuals and the environmental variables into the random forest model so that the random forest model further predicts the residuals to obtain a target residual prediction value; and train the random forest model according to the residuals and the target residual prediction value.

[0121] In an exemplary embodiment, the training module 34 is further configured to determine a first error metric between the residuals and the target residual prediction value; in a case where it is determined that the first error metric is greater than or equal to a preset threshold, repeatedly perform an update operation until it is determined that the first error metric is less than the preset threshold, where the update operation is: update the initial hyperparameters through grid search, and determine the residual prediction value corresponding to the random forest after updating the initial hyperparameters as the target residual prediction value; and in a case where it is determined that the first error metric is less than the preset threshold, stop training the random forest model.

[0122] In an exemplary embodiment, the training module 34 is further configured to input the first height observation value and the environmental variables into the trained random forest model to determine each prediction value of each decision tree in the random forest model; determine the second height prediction value according to a first formula, where the first formula is: is the second height prediction value, N is the number of decision trees included in the random forest model, and T i(x) is each prediction value.

[0123] In an exemplary embodiment, the determination module 36 is further configured to determine a first weight value corresponding to the ARIMA model and a second weight value of the trained random forest model; determine the target height prediction value according to a second formula, where the second formula is: Y is the target height prediction value, is the first height prediction value, α1 is the first weight value, is the second height prediction value, and α2 is the second weight value.

[0124] It should be noted that the above-mentioned modules can be implemented by software or hardware. For the latter, it can be implemented in the following ways, but not limited to this: the above-mentioned modules are all located in the same processor; or, the above-mentioned modules are respectively located in different processors in any combination form.

[0125] An embodiment of the present application also provides a computer-readable storage medium, in which a computer program is stored. Among them, the computer program is set to execute the steps in any one of the above method embodiments when running.

[0126] Optionally, in this embodiment, the above storage medium can be set to store program code for executing the following steps:

[0127] S1. Fit an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and predict the first height prediction values of the sea waves at multiple second time points in the second time period according to the ARIMA model;

[0128] S2. When it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, train a random forest model according to the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and predict the second height prediction values of the sea waves at multiple second time points in the second time period according to the trained random forest model;

[0129] S3. Determine the target height prediction values of the sea waves at multiple second time points in the second time period according to the first height prediction values and the second height prediction values.

[0130] In an exemplary embodiment, the above computer-readable storage medium may include, but is not limited to: various media such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs that can store computer programs.

[0131] An embodiment of the present application also provides an electronic device, including a memory and a processor. A computer program is stored in the memory, and the processor is set to run the computer program to execute the steps in any one of the above method embodiments.

[0132] In an exemplary embodiment, the above electronic device may further include a transmission device and an input / output device, where the transmission device is connected to the above processor, and the input / output device is connected to the above processor.

[0133] Optionally, in this embodiment, the above-mentioned processor may be set to execute the following steps through a computer program:

[0134] S1. Fit an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and predict the first height prediction values of the sea waves at multiple second time points in the second time period according to the ARIMA model;

[0135] S2. In the case where a non-linear relationship exists in the residual distribution corresponding to the ARIMA model, train a random forest model based on the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and predict the second height prediction values of the sea waves at multiple second time points in the second time period according to the trained random forest model;

[0136] S3. Determine the target height prediction values of the sea waves at multiple second time points in the second time period according to the first height prediction values and the second height prediction values.

[0137] An embodiment of the present application also provides a computer program product. The above computer program product includes a computer program, and when the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0138] An embodiment of the present application also provides another computer program product, including a non-volatile computer-readable storage medium. The non-volatile computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the steps in any one of the above method embodiments are implemented.

[0139] An embodiment of the present application also provides a computer program. The computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium; a processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the steps in any one of the above method embodiments.

[0140] Optionally, in this embodiment, the above-mentioned processor may be set to execute the following steps through a computer program:

[0141] S1. Fit an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and predict the first height prediction values of the sea waves at multiple second time points in the second time period according to the ARIMA model;

[0142] S2. When it is determined that there is a non - linear relationship in the residual distribution corresponding to the ARIMA model, train a random forest model based on the residuals corresponding to the ARIMA model and the environmental variables of the target area in the first time period, and predict the second - height prediction values of the sea wave at multiple second - time points in the second time period according to the trained random forest model;

[0143] S3. Determine the target - height prediction values of the sea wave at multiple second - time points in the second time period according to the first - height prediction values and the second - height prediction values.

[0144] Specific examples in this embodiment can refer to the examples described in the above - mentioned embodiments and exemplary embodiments, and will not be repeated here.

[0145] Obviously, those skilled in the art should understand that the above - mentioned modules or steps of the present application can be implemented by a general - purpose computing device. They can be concentrated on a single computing device or distributed on a network composed of multiple computing devices. They can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device. And in some cases, the steps shown or described can be executed in a different order than here, or they can be separately fabricated into individual integrated - circuit modules, or multiple modules or steps of them can be fabricated into a single integrated - circuit module to be implemented. In this way, the present application is not limited to any specific combination of hardware and software.

[0146] The above - mentioned are only the preferred embodiments of the present application and are not used to limit the present application. For those skilled in the art, the present application can have various changes and modifications. Any modification, equivalent replacement, improvement, etc. made within the principle of the present application shall be included in the protection scope of the present application.

Claims

1. A method for predicting the wave height, characterized in that, Including: Fitting an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, and predicting the first height prediction values of the sea waves at multiple second time points in the second time period according to the ARIMA model; In the case where it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, training a random forest model according to the residuals corresponding to the ARIMA model and the environmental variables in the target area in the first time period, and predicting the second height prediction values of the sea waves at multiple second time points in the second time period according to the trained random forest model; Determining the target height prediction values of the sea waves at multiple second time points in the second time period according to the first height prediction values and the second height prediction values.

2. The method according to claim 1, wherein Fitting an autoregressive integrated moving average (ARIMA) model based on the first height observation values of the sea waves in the target area at multiple first time points in the first time period, including: Converting multiple first height observation values into a time series according to each first time point corresponding to each first height observation value; Decomposing the time series to extract the trend component in the time series; Removing the trend component in the time series through a differencing operation to obtain a differenced stationary series, and determining the differencing order corresponding to the ARIMA model; Determining the autocorrelation plot and partial autocorrelation plot corresponding to the differenced stationary series; Determining the moving average order corresponding to the ARIMA model according to the autocorrelation plot, and determining the autoregressive order corresponding to the ARIMA model according to the partial autocorrelation plot; Fitting the ARIMA model according to the differencing order, the moving average order, and the autoregressive order.

3. The method according to claim 1, wherein Before training the random forest model according to the residuals and the environmental variables in the target area in the first time period, the method further includes: Obtaining the second height observation values of the sea waves at multiple third time points in the third time period before the first time period, and inputting the second height observation values into the ARIMA model so that the ARIMA model outputs the third height prediction values of the sea waves at multiple first time points in the first time period; Determining the residuals corresponding to the ARIMA model according to the first height observation values and the third height prediction values; Determining whether there is a non-linear relationship in the residual distribution corresponding to the ARIMA model according to the residuals.

4. The method according to claim 1, wherein Training a random forest model according to the residuals and the environmental variables in the target area in the first time period, including: Performing feature learning on the environmental variables to determine the feature data corresponding to the environmental variables, and determining the initial hyperparameters of the random forest model, where the initial hyperparameters include at least one of the following: the number of decision trees included in the random forest model, the minimum number of samples for splitting nodes corresponding to the random forest model, and the minimum number of samples for leaf nodes; Constructing decision trees in the random forest model based on the feature data, the environmental variables, and the residuals. Input the residual and the environmental variable into the random forest model so that the random forest model further predicts the residual to obtain a target residual prediction value; Train the random forest model according to the residual and the target residual prediction value.

5. The method according to claim 4, wherein Training the random forest model according to the residual and the target residual prediction value includes: Determine a first error metric between the residual and the target residual prediction value; When it is determined that the first error metric is greater than or equal to a preset threshold, loop and execute an update operation until it is determined that the first error metric is less than the preset threshold, where the update operation is: update the initial hyperparameters through grid search, and determine the residual prediction value corresponding to the random forest after updating the initial hyperparameters as the target residual prediction value; When it is determined that the first error metric is less than the preset threshold, stop training the random forest model.

6. The method according to claim 1, characterized in that, Predicting the second height prediction values of the sea wave at multiple second time periods in the second time period according to the trained random forest model includes: Input the first height observation value and the environmental variable into the trained random forest model to determine each prediction value of each decision tree in the random forest model; determine the second height prediction value according to the first formula, where the first formula is: is the second height prediction value, N is the number of decision trees included in the random forest model, and T i(x) is each prediction value.

7. The method according to claim 1, wherein Determining the target height prediction values of the sea wave at multiple second time points in the second time period according to the first height prediction value and the second height prediction value includes: Determine a first weight value corresponding to the ARIMA model and a second weight value of the trained random forest model; Determine the target height prediction value according to the second formula, where the second formula is: Y is the predicted value of the target height, is the first predicted height value, and α1 is the first weight value, is the second predicted height value, and α2 is the second weight value.

8. A device for predicting the height of ocean waves, characterized in that, Includes: A fitting module, configured to fit an autoregressive integrated moving average (ARIMA) model according to the first height observation values of the sea wave at multiple first time points in the first time period of the target area, and predict the first height prediction values of the sea wave at multiple second time points in the second time period according to the ARIMA model; A training module, configured to train a random forest model according to the residual corresponding to the ARIMA model and the environmental variable of the target area in the first time period when it is determined that there is a non-linear relationship in the residual distribution corresponding to the ARIMA model, and predict the second height prediction values of the sea wave at multiple second time points in the second time period according to the trained random forest model; A determination module, configured to determine the target height prediction values of the sea wave at multiple second time points in the second time period according to the first height prediction value and the second height prediction value.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, where the program, when running, executes the method according to any one of claims 1 to 7.

10. An electronic device, comprising a memory and a processor, characterized in that, A computer program is stored in the memory, and the processor is configured to execute the method according to any one of claims 1 to 7 through the computer program.

Images