Marine environment treatment early warning method and system based on deep neural network

Abstract

The invention relates to the technical field of treatment and early warning, and particularly discloses a treatment and early warning method and system for a marine environment based on a deep neural network, and the method comprises the steps: building a marine environment monitoring relation database, and dividing the database into a training set and a test set; performing data preprocessing on historical data of the monitoring parameters and the prediction parameters; determining the order of the marine environment parameter dynamic prediction model, and obtaining the marine environment parameter dynamic prediction model; inputting the monitoring parameters into the marine environment parameter dynamic prediction model, and adopting a multi-step prediction method to obtain prediction parameters; performing early warning on the marine environment based on the prediction parameter, and generating a marine environment early warning signal based on an early warning result; according to the method, the change rule of the prediction error under different environmental conditions can be reflected more accurately, so that the early warning system can judge the abnormal condition of the marine environmental parameters more accurately, misjudgment and missed judgment caused by environmental factors are reduced, and the reliability and practicability of marine environmental governance early warning are improved.

Description

Technical Field

[0001] The present invention relates to the field of governance and early warning technology, and in particular to a governance and early warning method and system for a marine environment based on a deep neural network. Background Art

[0002] In today's era, the ocean plays a vital role in the global ecological balance and the sustainable development of human society. The ocean is not only a treasure trove of rich resources, providing humans with important resources such as fisheries, oil and gas, but also a key factor in climate regulation, profoundly affecting the global climate and ecological environment. However, with the acceleration of industrialization, the increasing frequency of maritime trade and the overexploitation of marine resources, the marine environment is facing unprecedented challenges. Problems such as marine pollution, ecosystem destruction, and frequent marine disasters continue to emerge, seriously threatening the health of the marine ecology and human well-being.

[0003] In this context, marine environmental monitoring and early warning technology is particularly important. Accurate early warning can provide timely and effective information support for marine environmental protection and governance, help to take countermeasures in advance, and reduce the negative impact of marine environmental problems. However, the current marine environmental early warning methods have many shortcomings. On the one hand, the existing early warning models are difficult to accurately capture the dynamic change trends of complex and changeable marine environmental parameters. The marine environment is affected by a variety of factors, including ocean currents, tides, climate change, etc. These factors are intertwined, making the marine environmental parameters show highly complex dynamic change characteristics. However, due to the lack of comprehensive consideration and effective modeling of these complex factors, traditional methods have resulted in inaccurate predictions of marine environmental parameters and are unable to timely and accurately reflect the actual changes in the marine environment.

[0004] On the other hand, traditional early warning methods often ignore the impact of marine environmental factors on prediction errors when determining prediction error thresholds and making warning judgments. The diversity and complexity of the marine environment make the characteristics of prediction errors significantly different under different environmental conditions. For example, in different seasons and different sea areas, the changing patterns of marine environmental parameters and the distribution of prediction errors are different. However, the existing early warning methods fail to fully consider the impact of these environmental factors, and only determine the prediction error threshold and make warning judgments based on a single error data. This leads to the early warning system being prone to misjudgment and missed judgments under different marine environmental conditions, reducing the reliability and practicality of the early warning.

[0005] In summary, the existing marine environmental early warning technology is difficult to meet the actual needs of current marine environmental protection and governance. Developing a new marine environmental governance early warning method that can fully consider the dynamic change characteristics of marine environmental parameters and the impact of environmental factors on prediction errors has become a key issue that needs to be solved in this field. Summary of the Invention

[0006] The object of the present invention is to provide a governance early warning method and system for the marine environment based on a deep neural network to solve at least one of the above-mentioned prior art problems.

[0007] In a first aspect, the present invention provides a governance early warning method for the marine environment based on a deep neural network, including the following steps:

[0008] Establish a marine environment monitoring relational database, and divide the data in the marine environment monitoring relational database into a training set and a test set;

[0009] Among them, the marine environment monitoring relational database includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, prediction parameters;

[0010] Perform data preprocessing on the historical data of the monitoring parameters and prediction parameters;

[0011] Model establishment module: Determine the order of the dynamic prediction model of the marine environment parameters by using the criterion function order determination method, and obtain the dynamic prediction model of the marine environment parameters by using the least square estimation method;

[0012] Among them, when the prediction error of the dynamic prediction model of the marine environment parameters exceeds the prediction error threshold, adjust the order so that its prediction error does not exceed the prediction error threshold;

[0013] Input the monitoring parameters into the dynamic prediction model of the marine environment parameters, and use the multi-step prediction method to obtain the prediction parameters;

[0014] Based on the prediction parameters, give an early warning to the marine environment, and generate a marine environment early warning signal based on the early warning result.

[0015] In a second aspect, the present invention provides a governance early warning system for the marine environment based on a deep neural network, and the system includes:

[0016] Database establishment module: Establish a marine environment monitoring relational database, and divide the data in the marine environment monitoring relational database into a training set and a test set;

[0017] Among them, the marine environment monitoring relational database includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, prediction parameters;

[0018] Data processing module: Perform data preprocessing on the historical data of the monitoring parameters and prediction parameters;

[0019] Model establishment module: Determine the order of the dynamic prediction model of the marine environment parameters by using the criterion function order determination method, and obtain the dynamic prediction model of the marine environment parameters by using the least square estimation method;

[0020] Among them, when the prediction error of the marine environmental parameter dynamic prediction model exceeds the prediction error threshold, the order is adjusted so that its prediction error does not exceed the prediction error threshold;

[0021] Data prediction module: Input the monitoring parameters into the marine environmental parameter dynamic prediction model, and use a multi-step prediction method to obtain prediction parameters;

[0022] Environmental warning module: Based on the prediction parameters, give a warning about the marine environment, and generate a marine environment warning signal based on the warning result.

[0023] Advantages of the present invention:

[0024] By extracting the monitored environments of each error window and performing normalization processing, analyzing the same environmental characteristics therein, and identifying the regulated environment, it means that the determination of the prediction error threshold not only depends on the error data itself, but also combines the actual characteristics of the marine environment. The present invention can more accurately reflect the variation law of the prediction error under different environmental conditions, so that the warning system can more accurately judge the abnormal conditions of marine environmental parameters, reduce misjudgments and missed judgments caused by environmental factors, and improve the reliability and practicality of marine environment governance warnings. Description of the drawings

[0025] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0026] Figure 1 It is a flowchart of a governance warning method for a marine environment based on a deep neural network provided in Embodiment 1 of the present invention;

[0027] Figure 2 It is a schematic structural diagram of a governance warning system for a marine environment based on a deep neural network provided in Embodiment 2 of the present invention;

[0028] Figure 3 It is a schematic structural diagram of a governance warning device for a marine environment based on a deep neural network provided in Embodiment 3 of the present invention. Detailed implementation manners

[0029] To enable those skilled in the art to better understand the solution of the present invention, the following will clearly and completely describe the technical solution in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0030] Embodiment 1

[0031] Figure 1 It is a flowchart of a method for governing and warning of the marine environment based on a deep neural network provided in Embodiment 1 of the present invention. The embodiments of the present invention are applicable to a situation of governing and warning of the marine environment based on a deep neural network. The method for governing and warning of the marine environment based on a deep neural network can be executed by a system for governing and warning of the marine environment based on a deep neural network. The system for governing and warning of the marine environment based on a deep neural network can be implemented by software and / or hardware, and the system for governing and warning of the marine environment based on a deep neural network can be configured in a device for governing and warning of the marine environment based on a deep neural network. Optionally, a device for governing and warning of the marine environment based on a deep neural network can be an electronic device, and the electronic device can be a notebook, a desktop computer, a smart tablet, etc. The embodiments of the present invention do not limit this.

[0032] A method for governing and warning of the marine environment based on a deep neural network provided by the embodiments of the present invention specifically includes the following steps:

[0033] Step 1: Establish a relational database for marine environment monitoring, and randomly extract a certain proportion (for example, 70%) of the data from the relational database for marine environment monitoring as a training set for model training, and the remaining data as a test set for evaluating the prediction error of the model;

[0034] Among them, the relational database for marine environment monitoring includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, prediction parameters;

[0035] It should be explained that the monitoring parameters include but are not limited to: water temperature, pH value, turbidity, nitrate nitrogen, and the monitoring environment includes but is not limited to: environmental temperature, environmental season;

[0036] Step 2: Perform data preprocessing on the historical data of the monitoring parameters and prediction parameters;

[0037] Among them, data preprocessing includes: performing linear normalization on the monitoring parameters to map the values of all monitoring parameters into the interval [0, 1], which is convenient for subsequent model training, and performing logarithmic normalization on the prediction parameters. Logarithmic normalization can effectively handle the large numerical range that may exist in the prediction parameters and make the data distribution more uniform;

[0038] Step 3: Use the criterion function order determination method (the criterion function order determination method can be: Akaike information criterion, Bayesian information criterion) to determine the order of the dynamic prediction model of ocean environmental parameters, and use the least squares estimation method to obtain the dynamic prediction model of ocean environmental parameters;

[0039] Among them, when the prediction error of the dynamic prediction model of ocean environmental parameters exceeds the prediction error threshold, adjust the order so that its prediction error does not exceed the prediction error threshold;

[0040] Exemplarily, the preset monitoring parameter sequence is X1, X2, … X m and the prediction parameter sequence y1, y2, … y m , and a set of coefficients β0, β1, … β p are obtained through the least squares method, so that is minimized, where m is the total number of monitoring parameters in the training set, j is the jth monitoring parameter in the training set, and the value of j is 1, 2, …… m, and p is the total number of coefficients;

[0041] In some implementation schemes, the process of obtaining the prediction error threshold is as follows:

[0042] Extract the monitoring parameters and their corresponding prediction parameters at different monitoring times for each monitoring location in the test set, and perform difference calculation to obtain the prediction error value;

[0043] Based on any one monitoring location, sort the prediction error values corresponding to all its monitoring times in ascending order according to the numerical size to obtain the prediction error sequence, and extract the 95% quantile in the prediction error sequence as the initial prediction error threshold;

[0044] Preset multiple sliding window periods so that the prediction error fluctuation values of all prediction error values within the sliding window period do not exceed the prediction error fluctuation threshold;

[0045] It should be explained that, based on any one monitoring location, arrange the prediction error values corresponding to all monitoring times at this location in the order of time sequence, and set a sliding window period in the arrangement so that the sliding window period is as long as possible and ensure that the prediction error fluctuation values do not exceed the prediction error fluctuation threshold; that is, the lengths of each sliding window period are not the same;

[0046] Among them, the process of obtaining the prediction error fluctuation value is:

[0047] Calculate the variance of all prediction error values within the sliding window period to obtain the prediction error fluctuation value;

[0048] Based on the error distribution characteristics within each sliding window period, divide the sliding window period into a high-error window, a low-error window, and a fluctuation window. The specific process is as follows:

[0049] Preset the window period duration threshold, and compare and analyze the duration of the sliding window period with the window period duration threshold respectively;

[0050] If the duration of the sliding window period ≤ the window period duration threshold, mark the corresponding sliding window period as an invalid window period;

[0051] If the duration of the sliding window period > the window period duration threshold, mark the corresponding sliding window period as a valid window period;

[0052] Merge the consecutive invalid window periods to obtain the fluctuation window;

[0053] At the same time, based on any valid window period, extract all prediction error values of the valid window period and calculate the mean value to obtain the prediction error window mean value;

[0054] Preset the prediction error window threshold, and compare and analyze the prediction error window mean value with the prediction error window threshold;

[0055] If the prediction error window mean value ≤ the prediction error window threshold, it indicates that the error degree within this valid window period is relatively low, and mark this valid window period as a low-error window;

[0056] If the prediction error window mean value > the prediction error window threshold, it indicates that the error degree within this valid window period is relatively high, and mark this valid window period as a high-error window;

[0057] At the same time, extract the monitored environment corresponding to each detection time and normalize the monitored environment;

[0058] Extract all types of monitored environments in each error window (where the error window includes the low-error window and the high-error window) respectively to obtain the error environment group;

[0059] Then, based on the error environment group of any one type of error window, analyze and extract the same environmental features;

[0060] Among them, the environmental features can be: the average value of the monitored environment at all detection times of the error window, and the change value of the monitored environment corresponding to all detection times of the error window;

[0061] Among all the error windows of the same type, extract the number of those containing the same environmental features, and calculate the ratio with the total number of all error windows of this type to obtain the same feature ratio;

[0062] Preset the same feature ratio threshold, and conduct a comparative analysis of the same feature ratio with the same feature ratio threshold;

[0063] If the same feature ratio ≤ the same feature ratio threshold, it indicates that under the corresponding environmental features, the errors of the prediction parameters are not similar;

[0064] If the same feature ratio > the same feature ratio threshold, it indicates that under the corresponding environmental features, the errors of the prediction parameters are similar, and the corresponding environmental features are marked as the regulated environment;

[0065] Based on the regulated environment, calculate the threshold optimization coefficient, so as to optimize the initial prediction error threshold and obtain the prediction error threshold;

[0066] Among them, the process of obtaining the threshold optimization coefficient is as follows:

[0067] Extract the corresponding prediction error values of the regulated environment in different error windows, calculate their Pearson correlation coefficients, and output the correlation coefficient values;

[0068] Then extract the correlation coefficient values of all regulated environments in any one error window and conduct normalization processing so that the sum of the correlation coefficient values of all regulated environments in the normalized error window is 1. Perform a weighted summation calculation of the normalized correlation coefficient values with their corresponding same feature ratios to obtain the threshold optimization coefficient;

[0069] The process of optimizing the initial prediction error threshold to obtain the prediction error threshold is as follows:

[0070] If the error window corresponding to the optimized value is a low-error window, in order to ensure the sensitivity of the early warning and avoid missed reports, the prediction error threshold should be appropriately reduced, that is, the prediction error threshold is the product between the threshold optimization coefficient and the initial prediction error threshold;

[0071] If the error window corresponding to the optimized value is a high-error window, at this time, in order to reduce false alarms, it is necessary to appropriately increase the prediction error threshold on the basis of the initial prediction error threshold, that is, the prediction error threshold is the product between the threshold optimization coefficient plus one and the initial prediction error threshold;

[0072] Step Four: Input the monitoring parameters into the dynamic prediction model of ocean environmental parameters, and adopt the multi-step prediction method to obtain the prediction parameters;

[0073] In some implementation schemes, repeat Step Three to train multiple different dynamic prediction models of ocean environmental parameters;

[0074] Input the monitoring parameters into each dynamic prediction model of marine environmental parameters respectively, and output the predicted parameters at different future monitoring times for each dynamic prediction model of marine environmental parameters to obtain the direct predicted parameters;

[0075] It should be noted that the direct predicted parameters are: input the monitoring parameters into the dynamic prediction model of marine environmental parameters, and directly output the predicted values at each future monitoring time based on the monitoring parameters;

[0076] Meanwhile, use the single-step prediction method to recursively predict the predicted parameters at different future monitoring times to obtain the recursive predicted parameters;

[0077] It should be noted that the recursive predicted parameters are: input the monitoring parameters into the dynamic prediction model of marine environmental parameters, and output the predicted value at the first future monitoring time based on the monitoring parameters, then input the predicted value at the first future monitoring time into the dynamic prediction model of marine environmental parameters, and based on the predicted value at the first future monitoring time, obtain the predicted value at the second future monitoring time, and so on;

[0078] Calculate the difference between the direct predicted parameters and the recursive predicted parameters at the same future detection time, and take the absolute value to obtain the prediction difference;

[0079] Then sum up the prediction differences at different future monitoring times within each dynamic prediction model of marine environmental parameters to obtain the total prediction difference;

[0080] Through the formula: Calculate to obtain the prediction weight WP of the i-th dynamic prediction model of marine environmental parameters i , where the value of i is 1, 2,..., n, and n is the total number of all dynamic prediction models of marine environmental parameters, and PE i is the total prediction difference of the i-th dynamic prediction model of marine environmental parameters;

[0081] Extract the direct predicted parameters of all different dynamic prediction models of marine environmental parameters, and obtain the direct weight predicted parameters at different future monitoring times through the method of weighted summation according to the prediction weights of each dynamic prediction model of marine environmental parameters;

[0082] Extract the recursive predicted parameters of all different dynamic prediction models of marine environmental parameters, and obtain the recursive weight predicted parameters at different future monitoring times through the method of weighted summation according to the prediction weights of each dynamic prediction model of marine environmental parameters;

[0083] Calculate the mean value of the direct weight predicted parameters and the corresponding recursive weight predicted parameters at the future monitoring time to obtain the predicted parameters at this future monitoring time;

[0084] Step Five: Based on the prediction parameters, give early warnings to the marine environment, and generate marine environment early warning signals based on the early warning results;

[0085] In some embodiments, preset the early warning parameter thresholds, and conduct a comparative analysis of the early warning parameters and the early warning parameter thresholds;

[0086] If the prediction parameter ≤ the early warning parameter threshold, generate a continuous prediction signal;

[0087] If the prediction parameter > the early warning parameter threshold, generate a marine environment early warning signal, and send the marine environment early warning signal to the manager's terminal;

[0088] The technical solution of the embodiment of the present invention is mainly: using a sliding window to dynamically analyze the prediction error, and dividing the window into a high-error window, a low-error window, and a fluctuation window according to the error fluctuation value and window duration within the window. Considering the dynamic change characteristics of the error in the time series, it can more accurately capture the fluctuation law of the error. On this basis, further identify the regulation environment, and calculate the Pearson correlation coefficient between the regulation environment and the prediction error value, and combine the same feature ratio for weighted summation to obtain the threshold optimization coefficient. According to different error window types, use the threshold optimization coefficient to targetedly adjust the initial prediction error threshold; in the process of determining the prediction error threshold, fully consider the monitoring environment factors, extract the monitoring environment of each error window and normalize it, analyze the same environmental features among them, and identify the regulation environment, which means that the determination of the prediction error threshold not only depends on the error data itself, but also combines the actual characteristics of the marine environment. The present invention can more accurately reflect the change law of the prediction error under different environmental conditions, so that the early warning system can more accurately judge the abnormal situation of marine environment parameters, reduce misjudgment and missed judgment caused by environmental factors, and improve the reliability and practicality of marine environment governance early warning.

[0089] Embodiment Two

[0090] On the basis of Embodiment 1, please refer to Figure 2 As shown, a marine environment governance early warning system based on a deep neural network described in the embodiment of the present invention includes:

[0091] Database establishment module: Establish a marine environment monitoring relational database, and divide the data in the marine environment monitoring relational database into a training set and a test set;

[0092] Among them, the marine environment monitoring relational database includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, prediction parameters;

[0093] Data processing module: Perform data preprocessing on the historical data of the monitoring parameters and prediction parameters;

[0094] Model establishment module: Determine the order of the dynamic prediction model of marine environmental parameters by using the criterion function order determination method, and obtain the dynamic prediction model of marine environmental parameters by using the least square estimation method;

[0095] Among them, when the prediction error of the dynamic prediction model of marine environmental parameters exceeds the prediction error threshold, adjust the order so that its prediction error does not exceed the prediction error threshold;

[0096] Data prediction module: Input the monitoring parameters into the dynamic prediction model of marine environmental parameters, and adopt the multi-step prediction method to obtain the prediction parameters;

[0097] Environmental warning module: Based on the prediction parameters, give a warning about the marine environment, and generate a marine environment warning signal based on the warning result.

[0098] Embodiment III

[0099] Refer to Figure 3 , the embodiment of the present invention also provides a computer device 3, including: a memory 302, a processor 301, and a computer program 303 stored on the memory 302. When the computer program 303 is executed on the processor 301, it implements a governance warning method for the marine environment based on a deep neural network as described in any one of the above methods.

[0100] The computer device 3 may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The computer device 3 may include, but is not limited to, a processor 301 and a memory 302. Those skilled in the art can understand that

[0101] Figure 3 This is only an example of the computer device 3 and does not constitute a limitation on the computer device 3. It may include more or fewer components than shown in the figure, or combine some components, or different components. For example, it may also include input / output devices, network access devices, etc.

[0102] The so-called processor 301 may be a central processing unit (CPU), and this processor 301 may also be other general-purpose processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or this processor may also be any conventional processor, etc.

[0103] In some embodiments, the memory 302 may be an internal storage unit of the computer device 3, such as the hard disk or memory of the computer device 3. In some other embodiments, the memory 302 may also be an external storage device of the computer device 3, such as a plug-in hard disk, a Smart Media Card (SMC), a Secure Digital (SD) card, a Flash Card, etc. equipped on the computer device 3. Further, the memory 302 may also include both the internal storage unit and the external storage device of the computer device 3. The memory 302 is used to store an operating system, application programs, a Boot Loader, data, and other programs, such as the program code of the computer program. The memory 302 may also be used to temporarily store data that has been output or is to be output.

[0104] Embodiment 4

[0105] An embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored. When the computer program is run by a processor, it implements a method for governance warning of a marine environment based on a deep neural network as described in any one of the above methods.

[0106] In this embodiment, if the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on such an understanding, to implement all or part of the processes in the above embodiment methods of the present application, a computer program can be used to instruct relevant hardware to complete. The computer program can be stored in a computer-readable storage medium. When the computer program is executed by a processor, it can implement the steps of the above various method embodiments. Among them, the computer program includes computer program code, and the computer program code can be in the form of source code, object code, an executable file, or some intermediate form, etc. The computer-readable medium may at least include: any entity or device that can carry the computer program code to the photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. For example, a USB flash drive, a mobile hard disk, a magnetic disk, or an optical disc, etc. In some jurisdictions, according to legislation and patent practice, the computer-readable medium may not be an electrical carrier signal and a telecommunication signal.

[0107] In the above embodiments, the descriptions of the respective embodiments have their own emphases. For parts not described or recorded in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0108] Those of ordinary skill in the art can realize that the units and algorithm steps of the examples described in combination with the embodiments disclosed herein can be implemented by electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of this application.

[0109] In the embodiments disclosed in this application, it should be understood that the disclosed apparatus / terminal device and method can be implemented in other ways. For example, the apparatus / terminal device embodiments described above are merely illustrative. For example, the division of the modules or units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical or other forms.

[0110] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place, or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0111] The above formulas are all dimensionless and take their numerical values for calculation. The formula is a formula obtained by collecting a large amount of data for software simulation to approximate the real situation as closely as possible. The preset parameters in the formula are set by technicians in this field according to the actual situation.

[0112] The above has described in detail an embodiment of the present invention, but the content described is only a preferred embodiment of the present invention and cannot be considered as used to limit the scope of implementation of the present invention. All equivalent changes and improvements made in accordance with the scope of the application of the present invention should still fall within the scope covered by the patent of the present invention.

Claims

1. A governance warning method for the marine environment based on a deep neural network, characterized in that, It includes the following steps: Establish a relational database for marine environmental monitoring, and divide the data in the relational database for marine environmental monitoring into a training set and a test set; Among them, the relational database for marine environmental monitoring includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, and prediction parameters; Perform data preprocessing on the historical data of monitoring parameters and prediction parameters; Model establishment module: Determine the order of the dynamic prediction model of marine environmental parameters using the criterion function order determination method, and obtain the dynamic prediction model of marine environmental parameters using the least squares estimation method; Among them, when the prediction error of the dynamic prediction model of marine environmental parameters exceeds the prediction error threshold, adjust the order so that its prediction error does not exceed the prediction error threshold; Input the monitoring parameters into the dynamic prediction model of marine environmental parameters, and use the multi-step prediction method to obtain the prediction parameters; Based on the prediction parameters, give an early warning for the marine environment, and generate a marine environment early warning signal based on the early warning result.

2. The governance warning method for a marine environment based on a deep neural network according to claim 1, wherein The process of obtaining the prediction error threshold is as follows: Extract the monitoring parameters and their corresponding prediction parameters at different monitoring times for each monitoring location in the test set, and perform difference calculation to obtain the prediction error value; Based on the prediction error values corresponding to all monitoring times at the same monitoring location, preset the initial prediction error threshold; Preset multiple sliding window periods so that the prediction error fluctuation values of all prediction error values within the sliding window period do not exceed the prediction error fluctuation threshold; Based on the error distribution characteristics within each sliding window period, divide the sliding window period into a high-error window, a low-error window, and a fluctuation window; At the same time, extract the monitoring environment corresponding to each detection time, and perform normalization processing on the monitoring environment; Extract all types of monitoring environments in each error window respectively to obtain an error environment group; Among them, the error window includes a low-error window and a high-error window; Then, based on the error environment group of any one type of error window, analyze and extract the same environmental characteristics; Based on all error windows of the same type, extract the number of error windows containing the same environmental characteristics, and calculate the ratio with the total number of all error windows of this type to obtain the same feature ratio; If the same feature ratio > the same feature ratio threshold, mark the corresponding environmental characteristics as regulated environments; Based on the regulated environment, calculate the threshold optimization coefficient, so as to optimize the initial prediction error threshold and obtain the prediction error threshold.

3. The governance and early warning method for marine environment based on deep neural network according to claim 2, wherein, The process of obtaining the initial prediction error threshold is as follows: Sort the prediction error values corresponding to all monitoring times at the same monitoring location in ascending order of numerical value to obtain a prediction error sequence, and extract the 95% quantile in the prediction error sequence as the initial prediction error threshold.

4. The governance warning method for a marine environment based on a deep neural network according to claim 2, characterized in that, The process of obtaining the prediction error fluctuation value is as follows: Calculate the variance of all prediction error values within the sliding window period to obtain the prediction error fluctuation value.

5. The governance and early warning method for a marine environment based on a deep neural network according to claim 2, characterized in that, Based on the error distribution characteristics within each sliding window period, divide the sliding window period into a high-error window, a low-error window, and a fluctuation window. The specific process is as follows: If the duration of the sliding window period ≤ the window period duration threshold, mark the corresponding sliding window period as an invalid window period; If the duration of the sliding window period > the window period duration threshold, mark the corresponding sliding window period as a valid window period; Merge consecutive invalid window periods to obtain a fluctuation window; Meanwhile, based on any valid window period, extract all prediction error values of the valid window period and calculate their mean to obtain the prediction error window mean; Preset a prediction error window threshold and compare and analyze the prediction error window mean with the prediction error window threshold; If the prediction error window mean ≤ the prediction error window threshold, mark this valid window period as a low error window; If the prediction error window mean > the prediction error window threshold, mark this valid window period as a high error window.

6. The governance warning method for a marine environment based on a deep neural network according to claim 2, wherein The process of obtaining the threshold optimization coefficient is as follows: Extract the prediction error values corresponding to different error windows in the regulation environment and calculate their Pearson correlation coefficients, and output the correlation coefficient values; Then extract the correlation coefficient values of all regulation environments in any one error window and perform normalization processing so that the sum of the correlation coefficient values of all regulation environments in the normalized error window is 1. Perform a weighted sum calculation on the normalized correlation coefficient values and their corresponding same feature ratios to obtain the threshold optimization coefficient.

7. A governance early warning method for the marine environment based on a deep neural network according to claim 6, characterized in that, The process of optimizing the initial prediction error threshold to obtain the prediction error threshold is as follows: If the error window corresponding to the optimized value is a low error window, the prediction error threshold is the product between the threshold optimization coefficient and the initial prediction error threshold; If the error window corresponding to the optimized value is a high error window, the prediction error threshold is the product between the threshold optimization coefficient plus one and the initial prediction error threshold.

8. A governance warning method for the marine environment based on a deep neural network according to claim 1, characterized in that The process of obtaining the prediction parameter is as follows: Train multiple different dynamic prediction models for ocean environment parameters; Input the monitoring parameters into each dynamic prediction model for ocean environment parameters respectively, and output the prediction parameters for different future monitoring times of each dynamic prediction model for ocean environment parameters to obtain direct prediction parameters; Meanwhile, use the single-step prediction method to recursively predict the prediction parameters for different future monitoring times to obtain recursive prediction parameters; Extract the direct prediction parameters of all different dynamic prediction models for ocean environment parameters, and obtain the direct weighted prediction parameters for different future monitoring times by means of weighted sum according to the prediction weights of each dynamic prediction model for ocean environment parameters; Extract the recursive prediction parameters of all different dynamic prediction models for ocean environment parameters, and obtain the recursive weighted prediction parameters for different future monitoring times by means of weighted sum according to the prediction weights of each dynamic prediction model for ocean environment parameters; Perform a mean calculation on the direct weighted prediction parameters and the corresponding recursive weighted prediction parameters for the future monitoring time to obtain the prediction parameter for this future monitoring time.

9. The governance warning method for a marine environment based on a deep neural network according to claim 8, wherein The process of obtaining the prediction weight is as follows: Perform a difference calculation on the direct prediction parameters and the recursive prediction parameters for the same future detection time and take the absolute value to obtain the prediction difference; Then perform a sum calculation on the prediction differences for different future monitoring times within each dynamic prediction model for ocean environment parameters to obtain the total prediction difference; Through the formula: The prediction weight WP of the i-th dynamic prediction model of ocean environmental parameters is obtained by calculation i , where the value of i is 1, 2,..., n, and n is the total number of all dynamic prediction models of ocean environmental parameters, and PE i is the total prediction difference of the i-th dynamic prediction model of ocean environmental parameters.

10. A governance early warning system for the marine environment based on a deep neural network, characterized in that, This system is used to execute the method described in any one of the above claims 1-9, and this system includes: Database establishment module: Establish a relational database for marine environmental monitoring, and divide the data in the relational database for marine environmental monitoring into a training set and a test set; Among them, the relational database for marine environmental monitoring includes: monitoring location, monitoring environment, monitoring parameters, monitoring time, prediction parameters; Data processing module: Perform data preprocessing on the historical data of monitoring parameters and prediction parameters; Model establishment module: Use the criterion function order determination method to determine the order of the dynamic prediction model of marine environmental parameters, and use the least squares estimation method to obtain the dynamic prediction model of marine environmental parameters; Among them, when the prediction error of the dynamic prediction model of marine environmental parameters exceeds the prediction error threshold, adjust the order so that its prediction error does not exceed the prediction error threshold; Data prediction module: Input the monitoring parameters into the dynamic prediction model of marine environmental parameters, and use the multi-step prediction method to obtain the prediction parameters; Environmental warning module: Based on the prediction parameters, give a warning for the marine environment, and generate a marine environmental warning signal based on the warning result.

Images