Oil film thickness prediction method and device based on BP neural network, equipment and medium

Abstract

The invention provides an oil film thickness prediction method based on a BP neural network, and the method comprises the steps: firstly capturing a fluorescence signal corresponding to an overflow oil film through an ultraviolet fluorescence method in a non-contact measurement mode, reducing the influence of seawater reflection or ambient light on the fluorescence signal, and collecting synchronous environment data, preparation is made for further improving the prediction precision of the oil film thickness of the spilled oil; a BP neural network is further introduced, a relationship between an oil film thickness true value and oil film feature data is learned in advance, then oil film feature information is extracted based on an oil spill signal in an actual application scene through a learned BP neural network model, and the oil film thickness of an oil spill oil film is predicted. The oil spill oil film monitoring process is simplified, the prediction efficiency of the oil spill oil film thickness is improved, and the accuracy and robustness of oil film thickness prediction are improved.

Description

Technical Field

[0001] The present invention relates to the field of environmental monitoring, and in particular to a method, device, computer equipment and computer-readable storage medium for predicting oil film thickness based on a BP neural network. Background Art

[0002] The damage caused by marine oil spills to the marine environment and ecosystem is becoming increasingly serious. Therefore, it is necessary to assess the amount of oil spilled in the ocean where the marine oil spill occurs, formulate emergency measures and carry out environmental restoration based on the assessed amount of oil spilled, and the accurate measurement of oil film thickness is crucial in assessing the amount of oil spilled. The traditional method of monitoring the thickness of oil spilled oil film mainly relies on the change of reflected light and manual observation. This method is not only inefficient, but also has low measurement accuracy for the thickness of oil film. With the rise of environmental awareness and the development of remote sensing technology, infrared spectroscopy and synthetic aperture radar (SAR) are becoming the mainstream methods. Among them, infrared spectroscopy is to identify oil film by monitoring the absorption and reflection characteristics of infrared light by oil spills. It is easily affected by environmental interference, such as infrared absorption of water bodies and weather conditions, and is not suitable for monitoring thin oil films; synthetic aperture radar (SAR) monitors oil spills through the interaction between radar waves and oil films, which is not suitable for monitoring thin oil films and small areas of oil spills.

[0003] Therefore, how to solve the low accuracy of predicting the thickness of oil spill film has become a technical problem that needs to be solved urgently. Summary of the invention

[0004] The main purpose of the present invention is to provide a method, device, equipment and computer-readable storage medium for predicting oil film thickness based on BP neural network, aiming to solve the current technical problems of low efficiency and accuracy in monitoring the thickness of oil spilled oil films.

[0005] To achieve the above object, the present invention provides an oil film thickness prediction method based on BP neural network, and the oil film thickness prediction method comprises the following steps:

[0006] When an oil spill signal is detected, a target fluorescence signal corresponding to the oil spill film is obtained;

[0007] Performing signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal as the target periodic signal;

[0008] Based on the target periodic signal, extracting oil film characteristic information corresponding to the oil spill film;

[0009] Based on the BP neural network model and the oil film characteristic information, the oil film thickness corresponding to the oil spill film is predicted.

[0010] In addition, to achieve the above-mentioned purpose, the present invention also provides an oil film thickness prediction device based on a BP neural network, and the oil film thickness prediction device based on a BP neural network comprises:

[0011] A fluorescence signal acquisition module is used to acquire a target fluorescence signal corresponding to the oil spill film when an oil spill signal is detected;

[0012] A periodic signal generating module, used for performing signal analysis on the target fluorescent signal to obtain a periodic signal corresponding to the target fluorescent signal as the target periodic signal;

[0013] An oil film feature extraction module, used for extracting oil film feature information corresponding to the oil spill film based on the target periodic signal;

[0014] The oil film thickness prediction module is used to predict the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information.

[0015] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer device, which includes a processor, a memory, and an oil film thickness prediction program based on a BP neural network stored in the memory and executable by the processor, wherein when the oil film thickness prediction program based on a BP neural network is executed by the processor, the steps of the oil film thickness prediction method based on a BP neural network as described above are implemented.

[0016] In addition, to achieve the above-mentioned purpose, the present invention also provides a computer-readable storage medium, on which is stored an oil film thickness prediction program based on a BP neural network, wherein when the oil film thickness prediction program based on a BP neural network is executed by a processor, the steps of the oil film thickness prediction method based on a BP neural network as described above are implemented.

[0017] The present application provides a method for predicting the thickness of an oil film based on a BP neural network. The method obtains a target fluorescence signal corresponding to the oil spill film when an oil spill signal is detected; performs signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal as the target periodic signal; extracts the oil film characteristic information corresponding to the oil spill film based on the target periodic signal; and predicts the thickness of the oil film corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information. In the above manner, the present application first captures the fluorescence signal corresponding to the oil spill film in a non-contact measurement mode by an ultraviolet fluorescence method, reduces the influence of seawater reflection or ambient light on the fluorescence signal, improves the success rate of capturing the fluorescence signal, and improves the prediction efficiency of the thickness of the oil spill film; and further introduces a BP neural network to learn the relationship between the oil film thickness and the oil film characteristic data in advance, and then extracts the oil film characteristic information based on the target fluorescence signal through the learned BP neural network model to predict the thickness of the oil spill film. Not only does it simplify the monitoring process of the oil spill film, improve the prediction efficiency of the oil spill film thickness, but it also improves the accuracy and robustness of the prediction of the oil film thickness. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 It is a flow chart of the first embodiment of the oil film thickness prediction method based on BP neural network of the present invention;

[0019] Figure 2 It is a schematic diagram of the block diagram of each module in the marine oil spill monitoring system of the present invention;

[0020] Figure 3 This is a schematic diagram of the model training process based on the BP neural network model of the present invention;

[0021] Figure 4 This is a schematic diagram of the data preprocessing process of the present invention;

[0022] Figure 5 It is a schematic diagram of model inversion based on BP neural network model of the present invention;

[0023] Figure 6 It is a schematic diagram of the flow chart of the method for predicting the thickness of oil spill film of the present invention;

[0024] Figure 7 This is an oil film thickness prediction diagram based on BP neural network according to an embodiment of the present invention;

[0025] Figure 8 This is a percentage error diagram of the actual value and predicted value of the oil film thickness in an embodiment of the present invention;

[0026] Fig. 9 It is a schematic diagram of an oil film thickness prediction device based on BP neural network of the present invention;

[0027] Fig.10 The figure is a schematic diagram of the hardware structure of a computer device involved in an embodiment of the present invention.

[0028] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0030] The flowcharts shown in the accompanying drawings are only examples and do not necessarily include all the contents and operations / steps, nor must they be executed in the order described. For example, some operations / steps may also be decomposed, combined or partially merged, so the actual execution order may change according to actual conditions.

[0031] It should be understood that the terms used in this application specification are only for the purpose of describing specific embodiments and are not intended to limit the application. As used in this application specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0032] It should be further understood that the term “and / or” used in the specification and appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0033] The oil film thickness prediction method based on BP neural network involved in the embodiment of the present invention is mainly applied to computer equipment, which can be a PC, a portable computer, a mobile terminal or other equipment with display and processing functions.

[0034] In conjunction with the accompanying drawings, some embodiments of the present application are described in detail below. In the absence of conflict, the following embodiments and features in the embodiments can be combined with each other.

[0035] Reference Figure 1 , Figure 1 It is a flow chart of the first embodiment of the oil film thickness prediction method based on BP neural network of the present invention.

[0036] like Figure 1As shown, an embodiment of the present invention provides an oil film thickness prediction method based on a BP neural network. The oil film thickness prediction method based on a BP neural network includes steps S10 to S30.

[0037] The oil film thickness prediction method based on BP neural network is applied to offshore oil spill monitoring system, such as Figure 2 As shown, the offshore oil spill monitoring system includes a microcontroller, an ultraviolet light source excitation module, a fluorescence signal receiving module, an environmental signal receiving module, a signal processing module, etc.

[0038] The microcontroller is used to execute the program of the oil film thickness prediction method and is responsible for coordinating and managing the workflow of the entire system. The microcontroller specifically receives the transmission data from various sensors and modules and processes and analyzes the received data in real time. Specifically, it can be used to determine whether an oil spill has occurred according to the preset algorithm and logic, and predict the oil film thickness information when an oil spill occurs.

[0039] In a specific embodiment, an STM32 single-chip microcomputer can be used as a microcontroller, and a random access memory is used to store data and results during operation.

[0040] The ultraviolet light source excitation module is used to irradiate a preset sea surface area through an ultraviolet light source to excite the oil film that may exist on the sea surface to generate a fluorescent signal.

[0041] Specifically, an LED light panel can be used as the light source, and the wavelength of the ultraviolet light source can be 365nm with a band center of 5Hz, so as to effectively improve the sensitivity and accuracy of oil spill monitoring.

[0042] The fluorescence signal receiving module is used to collect and amplify the fluorescence signal generated by the oil spill film, to accurately capture the fluorescence signal in a complex marine environment, and to amplify the fluorescence signal through a signal amplifier.

[0043] Specifically, a SiPM photoelectric converter may be used to convert the fluorescence signal, and an incident filter with a central wavelength of 410 nm may be used to filter the fluorescence signal to obtain an effective fluorescence signal.

[0044] Furthermore, the offshore oil spill monitoring system also includes:

[0045] The communication module is used to realize remote monitoring of oil spill monitoring equipment and data transmission of sensors. It can communicate via wireless communication, satellite communication or wired communication, and can transmit monitoring data in various marine environments.

[0046] In a specific embodiment, data transmission and communication with a computer can be achieved through RS232, and various data can be sent to a cloud server by interacting with GPS.

[0047] The environmental signal receiving module is used to collect environmental parameters including background light intensity, temperature, salinity, wind speed, etc., which are used to correct and optimize the prediction results of oil spill film thickness. In a specific embodiment, the environmental signal receiving module is used in conjunction with the fluorescent signal receiving module.

[0048] The signal processing module is used to pre-process the fluorescence signals and environmental parameters transmitted by different sensors, and extract the oil film characteristics of the oil film based on the pre-processed fluorescence signals and environmental parameters, and predict the oil film thickness based on the oil film characteristics;

[0049] Specifically, the thickness of the oil spill film can be predicted by the BP neural network model trained in the signal processing module.

[0050] The power module can use an isolated DC regulated power supply as the power module. By adopting an isolation design, external electromagnetic interference can be effectively isolated to ensure the working stability and monitoring accuracy of the monitoring system.

[0051] In a specific embodiment, the input voltage and output voltage of the power module can be flexibly adjusted within a preset range according to different application scenarios.

[0052] Step S10: when an oil spill signal is detected, obtaining a target fluorescence signal corresponding to the oil spill film;

[0053] In this embodiment, after the offshore oil spill monitoring system is started, it enters the trigger measurement mode, that is, it determines whether the oil spill signal is received according to the preset time interval. If the offshore oil spill monitoring system detects the oil spill signal, that is, when the oil spill is discovered, it starts to execute the oil film thickness prediction program. If the oil spill signal is not received, it continues to wait until the preset time interval to determine whether the oil spill signal is received. After the oil spill film thickness prediction is performed, the oil film thickness prediction program is further repeatedly executed according to the preset time interval until the offshore oil spill ends.

[0054] Specifically, the offshore oil spill monitoring system creates a sampling queue data structure for storing the sampling data of each sensor. When an oil spill signal is detected, a target fluorescence signal corresponding to the oil spill film is obtained.

[0055] The target fluorescence signal is obtained by preprocessing the initial fluorescence data generated by the offshore oil spill monitoring system on the oil spill film.

[0056] Furthermore, in the case where the oil spill signal is monitored, before obtaining the target fluorescent signal corresponding to the oil spill film, the method further includes:

[0057] Based on the ultraviolet light source, a preset flashing frequency and a preset wavelength are used to illuminate a preset sea surface area, so that the oil spill film generates a fluorescent signal;

[0058] The fluorescent signal corresponding to the oil spill film is collected based on the fluorescent signal receiving module, and the fluorescent signal corresponding to the oil spill film is amplified by a signal amplifier;

[0059] The amplified fluorescence signal is preprocessed to obtain the target fluorescence signal, and the target fluorescence signal is stored in a sampling queue data structure.

[0060] In this embodiment, based on the light source flashing frequency and sampling frequency, an ultraviolet light source is set to illuminate a fixed area of ​​the sea surface at a fixed frequency. If an oil spill occurs, the leaked oil film will emit fluorescence under ultraviolet light excitation. The fluorescence signal generated by the oil spill film is collected through the fluorescence signal receiving module SiPM and a specific filter. A sampling queue data structure is created to store the sampling data of each sensor, and the stored sampling data is integrated and preprocessed to obtain the target fluorescence signal.

[0061] Specifically, in order to effectively improve the sensitivity and accuracy of oil spill monitoring, the wavelength and intensity of the ultraviolet light source are optimized, that is, an LED light board is used as the light source, the band center is set to 365nm, and the flashing frequency is set to 5Hz.

[0062] Furthermore, in the case where the oil spill signal is monitored, obtaining the target fluorescent signal corresponding to the oil spill film specifically includes:

[0063] When the oil spill signal is detected, an initial fluorescence signal corresponding to the oil spill film is obtained;

[0064] Based on a silicon photoelectric converter and an incident filter, converting the initial fluorescence signal into a digital signal;

[0065] The converted initial fluorescence signal is subjected to signal separation to obtain the background light signal and the oil film fluorescence signal, and the oil film fluorescence signal is used as the target fluorescence signal.

[0066] In this embodiment, when the oil spill signal is detected, the fluorescence signal generated by the oil spill film is collected through the fluorescence signal receiving module, and then the effective fluorescence signal is filtered out through the incident filter with a central wavelength of 410nm, which is recorded as the initial fluorescence signal. The initial fluorescence signal is amplified by the signal amplifier, and the amplified initial fluorescence signal is converted into a digital signal through the silicon photoelectric converter SiPM. In this way, the weak fluorescence signal can be accurately captured in the complex marine environment through the high-sensitivity sensor and signal amplifier.

[0067] The converted fluorescence signal is separated by a data analysis method to obtain a background light signal and an oil film fluorescence signal, and the oil film fluorescence signal is used as the target fluorescence signal.

[0068] Specifically, by using wavelet function to discretely transform the digital signal time series into a coefficient matrix, denoising the coefficient matrix to obtain a denoised matrix, and restoring the denoised matrix to a smooth, clear and low-noise denoised signal time series, it is easier to accurately separate the background light signal and the oil film fluorescence signal, thereby improving the accuracy of the oil spill film thickness prediction results.

[0069] It can be understood that the denoised signal time series is, for example: d[n] = {d1, d2, d3, ..., dn}, wherein d1, d2, d3, ..., dn represent the fluorescence signal intensity of a preset sea surface area at different time nodes, and a signal intensity curve can be constructed using the fluorescence signal intensity at different time nodes in the denoised signal time series (i.e., d1, d2, d3, ..., dn).

[0070] Based on the fact that the ultraviolet light source is a periodically flashing light signal, the constructed signal intensity curve is a periodically fluctuating curve with clear curve peaks and curve troughs. When the ultraviolet light source is irradiated to the preset sea surface area, the emitted fluorescence signal is mainly composed of fluorescence stimulated by the spilled oil floating on the water surface. Therefore, the curve trough represents the background signal corresponding to the preset sea surface area, and the curve peak represents the target signal corresponding to the oil film on the surface of the preset sea surface area.

[0071] After identifying the peaks and troughs in the signal intensity curve, the background light signal is extracted according to the troughs of the curve, and the oil spill film signal is extracted according to the peaks of the curve.

[0072] Step S20, performing signal analysis on the target fluorescent signal to obtain a periodic signal corresponding to the target fluorescent signal as the target periodic signal;

[0073] In this embodiment, the time series of the target fluorescence signal is analyzed using any algorithm of cluster analysis, regression analysis or wavelet transform to obtain the signal periodic variation frequency, and based on the signal periodic variation frequency, the periodic signal of the target fluorescence signal is identified and recorded as the target periodic signal.

[0074] Specifically, a digital signal time series matching the spectral information is generated according to the spectral information of the target fluorescent signal, and the digital signal time series is used as the target periodic signal.

[0075] Step S30, extracting oil film characteristic information corresponding to the oil spill film based on the target periodic signal;

[0076] In this embodiment, a preset wavelet function is used to perform discrete transformation on the target periodic signal (i.e., digital signal time series) to obtain a coefficient matrix, and the elements in the coefficient matrix correspond to the signal characteristics of the target fluorescence signal at multiple scales and / or frequencies. The coefficient matrix is ​​recorded as the oil film characteristic information corresponding to the oil spill film.

[0077] Specifically, using a preset wavelet function to perform discrete transformation on the target periodic signal specifically includes:

[0078] A convolution operation is performed on the digital signal time series using a preset wavelet function to obtain a plurality of wavelet coefficients, and different wavelet coefficients are used to characterize the signal characteristics of the target fluorescence signal at different scales and / or different frequencies;

[0079] A coefficient matrix is ​​constructed according to the wavelet coefficients, wherein the elements in the coefficient matrix are the wavelet coefficients.

[0080] It can be understood that the discrete transformation (wavelet transformation) of the digital signal time series using the preset wavelet function is to decompose the digital signal time series into a series of wavelet functions, so as to analyze the signal or function at different scales. The discrete transformation of the digital signal time series using the preset wavelet function includes determining the preset wavelet function, and then performing inner product of the digital signal time series with various translations and expansions of the preset wavelet function to obtain a plurality of different decimal coefficients. The decimal coefficients are used to describe the signal characteristics of the target fluorescence signal at different scales and / or different frequencies.

[0081] Step S40: predicting the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information.

[0082] In this embodiment, the oil film characteristic information is input into the BP neural network model, and the oil film thickness corresponding to the oil spill oil film is output through the correlation relationship between the oil film characteristic information and the oil film thickness learned by the BP neural network model.

[0083] Furthermore, before predicting the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information, the method further includes:

[0084] The oil film characteristic data set is divided into a training set and a test set according to a preset ratio;

[0085] Initializing the model parameters of the BP neural network model, optimizing the model weights of the BP neural network model based on a gradient descent algorithm, and calculating the difference between the predicted value and the standard value of the training set output by the BP neural network model based on a preset loss function;

[0086] Based on a preset learning rate, a preset training round and the training set, the BP neural network model is trained until the model weight or the difference value reaches a preset convergence condition.

[0087] In this embodiment, the model parameters of the BP neural network model are first initialized and assigned values, and the model parameters include the number of layers, batch size, number of iterations, activation function, learning rate, weight, bias, etc.

[0088] Specifically, Figure 3 As shown, this embodiment adopts random initialization when training the BP neural network model for the first time, assigns initial values ​​to the unit weights and bias items in the model, that is, randomly sets the weight matrix and bias vector corresponding to the model parameters to values ​​within a preset range. Momentum initialization is adopted in subsequent training of the BP neural network model, that is, the training results of the previous model parameters are set as the weight matrix and bias vector of the next model parameters.

[0089] The training set data corresponding to the oil film characteristic data set is imported into the input layer of the BP neural network model to train the model parameters. k , n=0.

[0090] Specifically, the training data (i.e., data value) corresponding to the first input layer is multiplied by the corresponding initial parameter weight, and then added to the corresponding bias vector to obtain the output result corresponding to the first input layer. The input signal u and output signal v of each layer of neurons are calculated. Then, based on the forward propagation algorithm, the output result corresponding to the first input layer is input into the activation function for calculation, and the calculation result obtained by the activation function is passed to the next node of the hidden layer, and so on, to complete the parameter training of all input layers in the BP neural network model until the BP neural network model reaches the corresponding convergence condition.

[0091] In a specific embodiment, when the mean square error function value is less than a preset value (ie, E(n) < ζ), or the preset training round k is greater than the maximum training round (ie, k > Maxloop), it is determined that the BP neural network model has reached the corresponding convergence condition. When the above conditions are not met, the local gradient δ of each layer of neurons is reversely calculated, and the weight correction amount (ie, n = n + 1) is calculated to perform the next round of model training, that is, re-input the next training sample in the training set.

[0092] Furthermore, in the forward propagation algorithm, the output result of the previous input layer is added to the activation function and used as the input of the next input layer, where the sigmoid activation function f(x) is:

[0093]

[0094] The derivative of the sigmoid activation function f(x) is:

[0095] f′(x)=f(x)·[1-f(x)]

[0096] Furthermore, the forward propagation algorithm is used to input the output data of the last node into the multilayer perceptron to obtain the final output (i.e., the oil film thickness corresponding to each oil spill film feature), where the calculation formula of the output signal v is:

[0097]

[0098] k is the maximum training round, i is the training round number, u i is the input signal of each layer of neurons in round i, W i 、b i is the weight correction of round i.

[0099] in, η is the preset learning rate.

[0100] Furthermore, the difference between the output signal v and the expected result is calculated based on the error function E(n), where the calculation formula of E(n) is:

[0101]

[0102] Furthermore, in order to improve the training efficiency of the BP neural network model, the gradient descent method is used, that is, the local gradient of the hidden layer is calculated backward layer by layer (i.e., reverse calculation) through the chain rule, where the calculation formula of the local gradient δ is:

[0103]

[0104] This embodiment provides a method for predicting the thickness of an oil film based on a BP neural network. The method obtains a target fluorescence signal corresponding to the oil spill film when an oil spill signal is detected; performs signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal as the target periodic signal; extracts the oil film characteristic information corresponding to the oil spill film based on the target periodic signal; and predicts the thickness of the oil film corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information. In the above manner, the present application first captures the fluorescence signal corresponding to the oil spill film in a non-contact measurement mode by using an ultraviolet fluorescence method to reduce the influence of seawater reflection or ambient light on the fluorescence signal, and collects synchronous environmental data at the same time, so as to prepare for further improving the prediction accuracy of the thickness of the oil spill film; and further introduces a BP neural network to learn the relationship between the true value of the oil film thickness and the oil film characteristic data in advance, and then extracts the oil film characteristic information based on the target fluorescence signal through the learned BP neural network model to predict the thickness of the oil spill film. It not only simplifies the monitoring process of the oil spill film, improves the prediction efficiency of the oil spill film thickness, but also improves the accuracy and robustness of the prediction of the oil film thickness.

[0105] Furthermore, based on the above Figure 1 In the embodiment shown, in this embodiment, the step S20 specifically includes:

[0106] Acquire the oil film type of the oil spill film, and determine the target analysis algorithm corresponding to the oil spill film based on the oil film type;

[0107] Based on the target analysis algorithm, the time series corresponding to the target fluorescence signal is analyzed to obtain the signal periodic change frequency corresponding to the target fluorescence signal;

[0108] The target periodic signal is generated based on the signal periodic change frequency corresponding to the target fluorescent signal.

[0109] In this embodiment, the oil film types include crude oil, diesel, gasoline, toluene, xylene, and lubricating oil. The analysis algorithms include cluster analysis, regression analysis, and wavelet transform. Signal data of different oil film types are suitable for different analysis algorithms. Because, in order to further improve the accuracy of the prediction result of the oil film thickness, the target analysis algorithm suitable for the oil film is further determined according to the oil film type of the oil spill. Take the target analysis algorithm as the wavelet transform algorithm as an example for explanation:

[0110] A preset wavelet function is used to perform a discrete transformation on the digital signal time series corresponding to the target fluorescence signal. That is, a convolution operation is performed on the digital signal time series using the preset wavelet function to obtain multiple wavelet coefficients, and different wavelet coefficients are used to characterize the signal characteristics of the target fluorescence signal at different scales and / or different frequencies; and a coefficient matrix is ​​constructed based on the wavelet coefficients, wherein the elements in the coefficient matrix are wavelet coefficients.

[0111] Among them, the preset wavelet function is Ricker wavelet (Mexican hat wavelet), and its function expression is as follows:

[0112]

[0113] Wherein, t represents the sampling interval, and f represents the center frequency of the preset wavelet function; the offshore oil spill monitoring system, microcontroller or signal processing module that executes the present method generates a Ricker wavelet with the preset center frequency and sampling frequency, and then performs a convolution operation on the digital signal time series D[n] to obtain the wavelet coefficients, and finally constructs a coefficient matrix as the output sequence φ(t, f) using the wavelet coefficients as matrix elements, so that the elements in the coefficient matrix correspond to the signal characteristics of the target fluorescence signal at multiple scales and / or frequencies.

[0114] Wherein, generating a digital signal time series matching the spectral information according to the spectral information of the target fluorescence signal comprises:

[0115] Performing filtering and amplification processing on the target fluorescence signal to obtain an optimized fluorescence signal;

[0116] The optimized fluorescence signal is sampled according to a preset sampling period to obtain multiple groups of time-discrete voltage data;

[0117] Multiple sets of voltage data are combined into a digital signal time series based on the time order of time nodes.

[0118] It can be understood that each set of voltage data represents the light intensity of the target fluorescence signal at different time nodes, and the different time nodes are discrete from each other, wherein the higher the voltage data, the higher the light intensity of the target fluorescence signal.

[0119] Furthermore, the performing signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal specifically includes:

[0120] Performing data cleaning on the target fluorescence signal, and performing wavelet transformation and signal decomposition on the cleaned target fluorescence signal to obtain a periodic variation frequency corresponding to the target fluorescence signal;

[0121] Based on the periodic variation frequency corresponding to the target fluorescent signal, performing noise processing on the target fluorescent signal to remove high-frequency noise data in the sampled data;

[0122] Based on the periodic transformation frequency corresponding to the processed target fluorescent signal, a periodic signal corresponding to the target fluorescent signal is obtained.

[0123] In this embodiment, Figure 4 As shown, the specific process of data preprocessing in this embodiment includes:

[0124] 1. First, perform data cleaning on the target fluorescence signal, such as filtering out qualified signals through incident light filters.

[0125] 2. Further noise processing is performed on the filtered effective fluorescence signal, that is, outliers or noise data are identified and processed in the effective fluorescence signal. For example, wavelet transform is used for noise processing, that is, wavelet transform is performed on the effective fluorescence data through the wavelet transform algorithm, the effective fluorescence signal is decomposed, and the high-frequency noise components in the decomposed fluorescence signal are removed to retain the main fluorescence signal.

[0126] It is understandable that the flashing frequency of the ultraviolet light source in the offshore oil spill monitoring system is known to be 5 Hz, and the sampling frequency is 100 Hz, that is, the time interval between two adjacent sampling points is If the sampling time is set to 1 minute, then monitoring the oil film thickness requires at least points.

[0127] 3. The fluorescence signal after noise processing is further processed by default value, that is, the missing values ​​in the fluorescence signal data set are processed. For example, the default value processing can be performed by deleting missing data and interpolating missing data.

[0128] It can be understood that when the number of missing collected data points (i.e., fluorescence signals) is small, individual cases (i.e., a small number of signals that have little impact on the prediction results) are eliminated; when the number of missing collected data points is large, default value processing is performed through the mean replacement interpolation method; when the number of missing collected data points is too large, resampling is required.

[0129] 4. Further outlier processing is performed on the sampling points after default value processing, that is, outliers in the data are identified and processed to ensure the consistency of the sampling data. For example, the 3σ criterion can be used to delete outliers, that is, according to the statistical distribution law, data exceeding the range of ±3 standard deviations of the mean value is deleted to prevent it from affecting the data analysis results.

[0130] 5. Perform data normalization on the processed fluorescence signal, and scale the fluorescence signal to a uniform scale according to its corresponding voltage value. For example, the Min-Max normalization method is used to scale the voltage value corresponding to the fluorescence signal to a specified range (such as 0 to 1) to make the fluorescence signal comparable. The normalization formula for each fluorescence signal x is as follows:

[0131]

[0132] Furthermore, based on the above Figure 1 In the embodiment shown, in this embodiment, the step S30 specifically includes:

[0133] Collecting environmental signals of the environment to which the oil spill film belongs through an environmental signal receiving module, wherein the environmental signals include at least one of temperature, salinity and wind speed;

[0134] The oil film characteristic information is extracted based on the target periodic signal, the background light signal and the environmental signal.

[0135] In this embodiment, different oil film types, oil film fluorescence signals, background light signals, temperature, salinity, wind speed and other characteristics are used to reflect the changes in the thickness of the oil spill film in the actual marine environment.

[0136] It can be understood that, similar to the above-mentioned model training process, the BP neural network model is pre-trained based on the target periodic signal, the background light signal and the environmental signal.

[0137] Specifically, Figure 5 As shown, an ultraviolet light source is set to illuminate a fixed area of ​​the sea surface at a fixed frequency. If an oil spill is detected, the leaked oil film will emit fluorescence under the excitation of ultraviolet light. The fluorescence signal excited by the oil spill film is collected through the offshore oil spill monitoring system and the incident filter, and the fluorescence signal is converted into a digital signal through the SiPM and the AD converter. The digital signal time series corresponding to the fluorescence signal is analyzed by the data analysis method, and the fluorescence signal is separated according to the analysis results to obtain the background light signal and the oil film fluorescence signal. And synchronize the environmental signals collected by other sensors, including temperature, salinity, etc. Then integrate and preprocess all the read signals, and correspond the characteristic information of the oil film to the thickness one by one. Finally, based on the oil film thickness inversion, laboratory data, and theoretical modeling data, an oil film characteristic inversion model based on the BP neural network is constructed to periodically update the BP neural network model.

[0138] Furthermore, if Figure 6As shown, the offshore oil spill information data set is divided into a training data set and a test data set according to a preset ratio (such as 8:2). Then, the fluorescence signal time series is analyzed using wavelet transform to obtain the signal periodic change frequency, thereby identifying the periodic signal. The extracted periodic signal is input into the hidden layer to extract the oil film characteristic information of the oil spill oil film.

[0139] Furthermore, wavelet transform can be used to judge the stability of mutation signals and remove mutation signals to ensure the quality of the data set.

[0140] Finally, the training set data is imported into the BP neural network model for training, and the BP neural network model is verified with the test set data. If the evaluation index of the test set data verification result meets the expected requirements, it means that the model training is successful, otherwise the model training is repeated.

[0141] Furthermore, the types of oil films are divided into 6 categories according to different oily hazardous chemicals, namely crude oil, diesel, gasoline, toluene, xylene, and lubricating oil. 6,000 data points are used as a sample, and 100 samples are selected for each category, for a total of 600 experimental samples. As shown in the following table:

[0142]

[0143] As shown in the table above, the ranges corresponding to the parameters shown in the table above are the ranges of various oil film characteristics defined in this embodiment. The above characteristic data can affect the optical properties of the oil film, thereby affecting the measurement results of the oil film thickness. The above characteristic data include different oil film types, oil film thickness, background light relative intensity, temperature, salinity, wind speed and other diversified characteristics, which are used to truly reflect the changes in oil film thickness in the actual offshore environment.

[0144] Furthermore, combined with Figure 7 as well as Figure 8 , the oil film thickness prediction method is further explained:

[0145] First, the collected oil film feature data set is divided into a training set and a test set in a ratio of 8:2, of which 480 data are used for model training and 120 data are used to verify the accuracy and stability of the model.

[0146] Design the structure of BP neural network, where:

[0147] The input layer of the neural network contains 5 nodes, and the input features include oil film fluorescence intensity, background light relative intensity, temperature, salinity, and wind speed;

[0148] The hidden layer adopts a single-layer design, containing 16 nodes, and the output layer contains 1 node, which is the predicted oil film thickness;

[0149] The learning rate η is set to 0.01, the number of training rounds is set to 20,000, and the loss function is used to measure the difference between the predicted value and the actual value. The network weights are optimized by using the gradient descent algorithm until convergence is reached on the training set;

[0150] The trained model is then applied to the 120 data in the test set, and the error between the oil film thickness value predicted by the model and the actual oil film thickness value is calculated, and the following evaluation indicators are obtained:

[0151] The MSE on the test set is 0.656, which indicates that the model has a small prediction error in different thickness ranges.

[0152] The MAE on the test set is 0.750, which proves the prediction accuracy of the model.

[0153] According to the above test data analysis results, the BP neural network model shows high accuracy in predicting the thickness of all categories of oil products by classifying and analyzing the test data of different categories, especially in the thicker oil film thickness range (20 microns to 100 microns), the prediction error of the model is less than 10%.

[0154] Comparative analysis with traditional optical monitoring methods shows that:

[0155] Traditional methods mainly predict the oil film thickness by the change of reflected light intensity. Although it has a certain degree of accuracy in simple environments, it has a large error in complex sea conditions and different oil types. For example:

[0156] The MSE of the traditional method is 2.37 and the MAE is 3.43.

[0157] In comparison, the oil film thickness prediction method based on the BP neural network model provided in this embodiment shows lower errors in both MSE and MAE indicators, and the error rate of oil film thickness above 20% is reduced by 63%. This shows that the BP neural network method performs better in dealing with complex nonlinear relationships and has advantages in the accuracy and robustness of oil film thickness prediction.

[0158] Furthermore, the diversity and quality of the dataset have a significant impact on the model performance, and the prediction ability of the model can be further improved by increasing the amount of training data and enriching the diversity of the data.

[0159] See also Fig. 9 , Fig. 9 The present invention provides a schematic diagram of the functional modules of an oil film thickness prediction device based on a BP neural network.

[0160] like Fig. 9 As shown, the oil film thickness prediction device 900 based on BP neural network includes:

[0161] The fluorescence signal acquisition module 10 is used to acquire the target fluorescence signal corresponding to the oil spill film when the oil spill signal is detected;

[0162] A periodic signal generating module 20, configured to perform signal analysis on the target fluorescent signal to obtain a periodic signal corresponding to the target fluorescent signal as the target periodic signal;

[0163] An oil film feature extraction module 30, for extracting oil film feature information corresponding to the oil spill film based on the target periodic signal;

[0164] The oil film thickness prediction module 40 is used to predict the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information.

[0165] It should be noted that those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working process of the above-described device and each module can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0166] The above-mentioned device can be implemented in the form of a computer program. Fig.10 Runs on the computer device shown.

[0167] See also Fig.10 , Fig.10 1 is a schematic block diagram of a computer device provided in an embodiment of the present application. The computer device may be a server.

[0168] See also Fig.10 The computer device includes a processor, a memory and a network interface connected through a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0169] The non-volatile storage medium can store an operating system and a computer program. The computer program includes program instructions, and when the program instructions are executed, the processor can execute any oil film thickness prediction method based on the BP neural network.

[0170] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0171] The internal memory provides an environment for the operation of the computer program in the non-volatile storage medium. When the computer program is executed by the processor, the processor can execute any oil film thickness prediction method based on the BP neural network.

[0172] The network interface is used for network communication, such as sending assigned tasks, etc. Those skilled in the art will understand that Fig.10 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0173] It should be understood that the processor may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Among them, the general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0174] The processor is used to run a computer program stored in a memory to implement any one of the oil film thickness prediction methods based on a BP neural network provided in the embodiments of the present application.

[0175] A computer-readable storage medium is also provided in an embodiment of the present application, wherein the computer-readable storage medium stores a computer program, wherein the computer program includes program instructions, and the processor executes the program instructions to implement any one of the oil film thickness prediction methods based on the BP neural network provided in the embodiments of the present application.

[0176] The computer-readable storage medium may be an internal storage unit of the computer device described in the foregoing embodiment, such as a hard disk or memory of the computer device. The computer-readable storage medium may also be an external storage device of the computer device, such as a plug-in hard disk, a smart memory card (SmartMedia Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), etc., equipped on the computer device.

[0177] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the technical field can easily think of various equivalent modifications or replacements within the technical scope disclosed in the present application, and these modifications or replacements should be included in the protection scope of the present application. Therefore, the protection scope of the present application shall be based on the protection scope of the claims.

Claims

1. A method for predicting oil film thickness based on BP neural network, characterized in that: The oil film thickness prediction method comprises the following steps: When an oil spill signal is detected, a target fluorescence signal corresponding to the oil spill film is obtained; Performing signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal as the target periodic signal; Based on the target periodic signal, extracting oil film characteristic information corresponding to the oil spill film; Based on the BP neural network model and the oil film characteristic information, the oil film thickness corresponding to the oil spill film is predicted.

2. The oil film thickness prediction method according to claim 1, characterized in that: The performing of signal analysis on the target fluorescent signal to obtain a periodic signal corresponding to the target fluorescent signal as the target periodic signal specifically includes: Acquire the oil film type of the oil spill film, and determine the target analysis algorithm corresponding to the oil spill film based on the oil film type; Based on the target analysis algorithm, the time series corresponding to the target fluorescence signal is analyzed to obtain the signal periodic change frequency corresponding to the target fluorescence signal; The target periodic signal is generated based on the signal periodic change frequency corresponding to the target fluorescent signal.

3. The oil film thickness prediction method according to claim 1, characterized in that: In the case where the oil spill signal is detected, obtaining the target fluorescent signal corresponding to the oil spill film specifically includes: When the oil spill signal is detected, an initial fluorescence signal corresponding to the oil spill film is obtained; Based on a silicon photoelectric converter and an incident filter, converting the initial fluorescence signal into a digital signal; The converted initial fluorescence signal is subjected to signal separation to obtain the background light signal and the oil film fluorescence signal, and the oil film fluorescence signal is used as the target fluorescence signal.

4. The oil film thickness prediction method according to claim 3, characterized in that: The step of extracting the oil film characteristic information corresponding to the oil spill film based on the target periodic signal specifically includes: Collecting environmental signals of the environment to which the oil spill film belongs through an environmental signal receiving module, wherein the environmental signals include at least one of temperature, salinity and wind speed; The oil film characteristic information is extracted based on the target periodic signal, the background light signal and the environmental signal.

5. The oil film thickness prediction method according to claim 1, characterized in that: The performing signal analysis on the target fluorescence signal to obtain a periodic signal corresponding to the target fluorescence signal specifically includes: Performing data cleaning on the target fluorescence signal, and performing wavelet transformation and signal decomposition on the cleaned target fluorescence signal to obtain a periodic variation frequency corresponding to the target fluorescence signal; Based on the periodic variation frequency corresponding to the target fluorescent signal, performing noise processing on the target fluorescent signal to remove high-frequency noise data in the sampled data; Based on the periodic transformation frequency corresponding to the processed target fluorescent signal, a periodic signal corresponding to the target fluorescent signal is obtained.

6. The oil film thickness prediction method according to claim 1, characterized in that: Before predicting the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information, the method further includes: The oil film characteristic data set is divided into a training set and a test set according to a preset ratio; Initializing the model parameters of the BP neural network model, optimizing the model weights of the BP neural network model based on a gradient descent algorithm, and calculating the difference between the predicted value and the standard value of the training set output by the BP neural network model based on a preset loss function; Based on a preset learning rate, a preset training round and the training set, the BP neural network model is trained until the model weight or the difference value reaches a preset convergence condition.

7. The oil film thickness prediction method according to any one of claims 1 to 6, characterized in that: In the case where the oil spill signal is detected, before obtaining the target fluorescent signal corresponding to the oil spill film, the method further includes: Based on the ultraviolet light source, a preset flashing frequency and a preset wavelength are used to illuminate a preset sea surface area, so that the oil spill film generates a fluorescent signal; The fluorescent signal corresponding to the oil spill film is collected based on the fluorescent signal receiving module, and the fluorescent signal corresponding to the oil spill film is amplified by a signal amplifier; The amplified fluorescence signal is preprocessed to obtain the target fluorescence signal, and the target fluorescence signal is stored in a sampling queue data structure.

8. An oil film thickness prediction device based on BP neural network, characterized in that: The oil film thickness prediction device based on BP neural network comprises: A fluorescence signal acquisition module is used to acquire a target fluorescence signal corresponding to the oil spill film when an oil spill signal is detected; A periodic signal generating module, used for performing signal analysis on the target fluorescent signal to obtain a periodic signal corresponding to the target fluorescent signal as the target periodic signal; An oil film feature extraction module, used for extracting oil film feature information corresponding to the oil spill film based on the target periodic signal; The oil film thickness prediction module is used to predict the oil film thickness corresponding to the oil spill film based on the BP neural network model and the oil film characteristic information.

9. A computer device, characterized in that: The computer device includes a processor, a memory, and an oil film thickness prediction program based on a BP neural network stored in the memory and executable by the processor, wherein when the oil film thickness prediction program based on a BP neural network is executed by the processor, the steps of the oil film thickness prediction method based on a BP neural network as described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores an oil film thickness prediction program based on a BP neural network, wherein when the oil film thickness prediction program based on a BP neural network is executed by a processor, the steps of the oil film thickness prediction method based on a BP neural network as described in any one of claims 1 to 7 are implemented.

Images