Industrial source VOCs distribution analysis system based on laser radar

Abstract

The invention relates to the field of environment monitoring, and discloses an industrial source VOCs distribution analysis system based on a laser radar, which comprises a data acquisition module, a spectrum extraction module, a preliminary modeling module, an inversion optimization module, an intelligent analysis module, a visualization module and a dynamic monitoring and early warning module. The data acquisition module acquires VOCs data of an industrial source area and processes the VOCs data; the spectrum extraction module is used for extracting spectrum characteristics of VOCs from the processed laser signals; the preliminary modeling module estimates the VOCs concentration; the inversion optimization module globally optimizes the preliminary model in combination with an atmospheric transmission model, a nonlinear inversion algorithm and a genetic algorithm; the intelligent analysis module further improves the concentration prediction precision; the visual module is used for displaying the spatial distribution of the VOCs; and the dynamic monitoring and early warning module gives an early warning for abnormal concentration change. According to the invention, high-sensitivity, real-time monitoring and accurate distribution analysis of industrial source VOCs are realized, and a scientific basis is provided for pollution control and risk management of an industrial environment.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and in particular to an industrial source VOCs distribution analysis system based on laser radar. Background Art

[0002] In industrial environments, VOCs (volatile organic compounds) emission monitoring and distribution analysis are of great significance for pollution control and health management. The release of VOCs not only has a negative impact on air quality, but may also cause harm to the ecological environment and human health. Therefore, how to accurately and efficiently monitor and analyze the distribution of VOCs from industrial sources has become a key issue in environmental protection and industrial management. Existing VOCs monitoring technologies mainly include portable gas analyzers, fixed station monitoring systems, and some spectral absorption methods. However, these methods have significant deficiencies in wide-area real-time monitoring and high-sensitivity spatial distribution analysis.

[0003] Although portable gas analyzers can achieve high-precision VOCs detection, they are limited to single-point measurement and cannot cover a large range of industrial areas. They also require manual operation and are difficult to meet the needs of continuous and real-time monitoring. Fixed-station monitoring systems rely on fixed sampling points, with limited spatial coverage, and it is difficult to accurately reflect the distribution and concentration changes of VOCs in the entire industrial area. Infrared or ultraviolet absorption spectroscopy technology based on spectral absorption can achieve remote detection of VOCs, but it is easily interfered with during transmission in the atmospheric environment, making it difficult to obtain high-precision spatial distribution data. At the same time, because traditional algorithms are sensitive to noise signals and complex backgrounds in the environment, the accuracy and reliability of monitoring data are low, especially in a changing industrial environment, it is difficult to maintain stability.

[0004] The defects of these existing technologies have shown limitations in industrial source VOCs monitoring, mainly reflected in the lack of real-time performance, low spatial resolution and insufficient adaptability to changing environments. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention provides an industrial source VOCs distribution analysis system based on lidar, which can achieve high-sensitivity, real-time monitoring and precise distribution analysis of industrial source VOCs, overcoming the shortcomings of the prior art in real-time, spatial resolution and environmental adaptability.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: an industrial source VOCs distribution analysis system based on laser radar, comprising:

[0007] The data acquisition module is used to collect VOCs distribution data in industrial source areas in a specific band through lidar and perform preprocessing;

[0008] Spectral extraction module, used to extract spectral information related to VOCs characteristics from the processed lidar signal;

[0009] A preliminary modeling module, used to establish a preliminary model of VOCs concentration based on spectral information;

[0010] Inversion optimization module, used to optimize the preliminary model of VOCs concentration through inversion algorithm to obtain accurate VOCs concentration distribution;

[0011] The intelligent analysis module is used to perform machine learning and deep learning analysis on VOCs concentration distribution to improve the accuracy of concentration prediction.

[0012] Preferably, the data acquisition module includes:

[0013] A laser emitting unit, used for emitting a laser signal of a specific wavelength band, wherein the wavelength band is an ultraviolet wavelength band of 266nm or 355nm or an infrared wavelength band of 1064nm;

[0014] A laser receiving unit is used to receive an echo signal of a laser signal emitted by a laser emitting unit after being scattered or absorbed in a VOCs area, and generate corresponding spectral data;

[0015] The preprocessing unit is used to perform normalization processing and noise filtering processing on the laser echo signal.

[0016] Preferably, the spectrum extraction module includes:

[0017] A Fourier transform unit, used to perform Fourier transform on the normalized lidar signal to extract the spectral characteristic signal of VOCs;

[0018] A characteristic peak identification unit, used to identify characteristic absorption peaks of VOCs from the frequency domain signal after Fourier transformation, and generate characteristic spectrum data related to VOCs;

[0019] The signal intensity calculation unit is used to perform integral calculation on the characteristic absorption peak to obtain the absorption intensity I in a specific band. i As an input parameter of the VOCs concentration model, the calculation formula of the absorption intensity is:

[0020]

[0021] Among them, S(f) is the frequency domain signal, which is used to represent the absorption characteristics at different frequencies, I i is the absorption intensity of the i-th band, and f1 and f2 are the start and end frequencies of the characteristic absorption peak.

[0022] Preferably, the preliminary modeling module includes:

[0023] The regression analysis unit is used to establish a multivariate linear regression model of VOCs concentration based on the characteristic spectrum data generated by the spectrum extraction module to achieve a preliminary prediction of VOCs concentration. The expression of the model is:

[0024]

[0025] Among them, C is the initial concentration prediction value of VOCs, I i is the absorption intensity of the i-th band, β0 and β i are the intercept and regression coefficient of the regression model respectively, ∈ is the error term;

[0026] Regression coefficient calculation unit, used to estimate the coefficient β of the regression model based on the least squares method i , so that the prediction error of the regression model for VOCs concentration is minimized;

[0027] The error evaluation unit is used to calculate the mean square error of the regression model prediction results to evaluate the accuracy of the preliminary modeling.

[0028] Preferably, the inversion optimization module comprises:

[0029] The atmospheric transmission model unit is used to establish an improved multiple scattering model to describe the propagation characteristics of the laser in the industrial source environment and perform a preliminary inversion calculation of the VOCs concentration. The atmospheric transmission model expression is:

[0030] I(x)=I0e -ax +∫0 x k(x′)I(x′)dx′

[0031] Where I(x) represents the intensity of the laser at a distance of x, I0 is the initial laser intensity, α is the atmospheric absorption coefficient, and k(x′) is the scattering coefficient. This model optimizes the light transmission path based on the consideration of the atmospheric multiple scattering effect.

[0032] The nonlinear inversion unit is used to optimize the preliminary inversion results of the atmospheric transport model based on the nonlinear inversion algorithm to improve the calculation accuracy of the VOCs concentration. The nonlinear inversion algorithm adopts the Levenberg-Marquardt optimization method, and its objective function is:

[0033]

[0034] Among them, C i is the actual observed VOCs concentration, f(I i ; θ) is the predicted concentration calculated according to the laser intensity, and θ is the model parameter to be optimized.

[0035] Preferably, the inversion optimization module further includes:

[0036] The parameter optimization unit is used to globally optimize the parameters in the inversion model based on the genetic algorithm to further improve the adaptability and accuracy of the model in various industrial environments. The specific implementation of the genetic algorithm includes the following steps:

[0037] Initialize the population and randomly generate a set of initial parameter sets θ;

[0038] Fitness evaluation, calculating the fitness value of each set of parameters based on the objective function;

[0039] Selection and crossover: select parameters according to fitness and perform crossover operations to generate new parameter groups;

[0040] Mutation, mutating the newly generated parameter group to increase the diversity of solutions;

[0041] Iterate and update, repeating the above steps until the optimization conditions are met.

[0042] Preferably, the intelligent analysis module includes:

[0043] A deep learning model unit is used to perform nonlinear inversion analysis based on the VOCs spectral data generated by the spectral extraction module. The deep learning model unit adopts a deep neural network structure, and the input is the spectral data feature X = {I1, I2, ..., I n}, the output is the predicted value C of VOCs concentration, and the mathematical expression of its neural network is:

[0044] C=f(W n ·σ(…σ(W2·σ(W1·X+b1)+b2)…)+b n )

[0045] Among them, W i and b i are the weight and bias of the i-th layer respectively, and σ is the activation function, which can learn and fit the nonlinear features in the spectral data, thereby improving the accuracy of concentration prediction;

[0046] Model training unit, used to train the deep neural network model based on the historical data of VOCs, and optimize the weights and biases of the network by minimizing the loss function between the predicted concentration and the actual concentration;

[0047] The optimizer unit is used to optimize the value of the loss function using the gradient descent method during model training;

[0048] The feature selection unit is used to screen out the characteristic bands that contribute more to the prediction of VOCs concentration from the spectral data, and select the optimal feature subset through the recursive feature elimination method.

[0049] Preferably, the system further comprises a visualization module for generating a three-dimensional thermal distribution map of VOCs.

[0050] Preferably, the system further comprises a dynamic monitoring and early warning module for real-time monitoring of changes in VOCs concentration and abnormal early warning.

[0051] Preferably, the dynamic monitoring and early warning module includes:

[0052] Time series analysis unit, used to establish a time series model based on historical VOCs concentration data to predict the changing trend of VOCs concentration in the future;

[0053] A concentration threshold judgment unit is used to compare the VOCs concentration value monitored in real time with a preset safety threshold, and trigger an early warning signal when it is detected that the concentration value exceeds the threshold;

[0054] The trend anomaly detection unit is used to identify the abnormal growth trend of VOCs concentration and determine whether there is an abnormality by calculating the concentration change rate;

[0055] The warning information output unit is used to output warning signals and abnormal monitoring results to the user interface, and provide the diffusion path and trend information of VOCs concentration, including the current concentration value, change rate and predicted concentration curve, so that users can respond and process in time.

[0056] The present invention provides an industrial source VOCs distribution analysis system based on laser radar. It has the following beneficial effects:

[0057] 1. The system of the present invention utilizes a spectral extraction module to perform Fourier transform and characteristic peak identification on the laser signal, automatically extracts spectral information related to VOCs concentration and efficiently calculates the absorption intensity, provides data support for subsequent modeling and analysis, and significantly improves the accuracy and work efficiency of VOCs concentration analysis.

[0058] 2. The present invention constructs a multivariate linear regression model through a preliminary modeling module. The system can provide a basic estimate of VOCs concentration, ensuring the initial accuracy of inversion optimization. At the same time, the regression coefficient is solved by the least squares method, which reduces the modeling error and enhances the stability of the system.

[0059] 3. The present invention uses an atmospheric transmission model combined with a nonlinear inversion algorithm and a genetic algorithm to globally optimize the preliminary model, thereby improving the adaptability of the system in complex industrial environments, ensuring the accuracy of the VOCs concentration inversion calculation, and helping to maintain consistent monitoring performance in a changing environment.

[0060] 4. The intelligent analysis module of the present invention analyzes spectral data through deep neural networks and uses feature selection algorithms to screen important bands, thereby improving the accuracy and computational efficiency of VOCs concentration prediction, while enhancing the system's ability to generalize data and being able to adapt to diverse concentration data in different industrial environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 Schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the specification of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Please refer to the attached Figure 1 The industrial source VOCs distribution analysis system based on laser radar in the embodiment of the present invention realizes high-precision real-time monitoring and analysis of VOCs in industrial source environments by integrating modules such as data acquisition, spectrum extraction, preliminary modeling, inversion optimization and intelligent analysis. The various modules of the industrial source VOCs distribution analysis system based on laser radar in the present invention are described in detail below.

[0064] Data acquisition module

[0065] The data acquisition module is used to collect VOCs distribution in industrial source environments with high sensitivity and provide accurate data input. The data acquisition module realizes remote monitoring of VOCs through lidar signals in specific bands, and performs necessary preprocessing after signal acquisition, including signal normalization and noise filtering, to ensure the data quality and stability of subsequent spectral analysis and concentration modeling.

[0066] In this embodiment, the specific components of the data acquisition module include a laser emitting unit, a laser receiving unit, a signal normalization processing unit and a noise filtering unit:

[0067] 1. Laser emission unit:

[0068] In this embodiment, the laser emitting unit is used to emit laser signals of a specific band to enhance the absorption sensitivity of VOCs. The selected bands include ultraviolet bands (e.g., 266nm or 355nm) or infrared bands (e.g., 1064nm), which have good absorption effects on specific chemical groups (such as CH bonds) in VOCs, and help improve the detection sensitivity of the system. When the laser signal is emitted in an industrial source environment and encounters VOCs molecules, characteristic absorption or scattering will occur. These characteristic signals are collected by the laser receiving unit as echo signals for subsequent processing.

[0069] 2. Laser receiving unit:

[0070] In this embodiment, the laser receiving unit is used to receive the echo signal after the laser emitting unit emits in the VOCs area. The unit converts the reflected laser signal into an electrical signal to represent the preliminary information of the VOCs distribution. The signal output of the laser receiving unit is directly transmitted to the signal normalization processing unit and the noise filtering unit to ensure the continuity of subsequent data processing.

[0071] 3.Signal normalization processing unit:

[0072] In this embodiment, the signal normalization processing unit is used to eliminate the influence of environmental conditions (such as temperature, humidity and background light intensity) on the laser echo signal, so as to keep the signal intensity of the data consistent. Specifically, the normalization processing is calculated using the following formula:

[0073]

[0074] Among them, s(x,t) is the original lidar signal, S norm (x, t) is the normalized signal, x is the spatial position, and t is the time. Through this normalization process, the interference caused by changes in the external environment is eliminated, so that subsequent spectral analysis and concentration modeling are based on stable input signals, ensuring the validity of the data.

[0075] 4. Noise filtering unit:

[0076] In this embodiment, the noise filtering unit is used to perform noise processing on the normalized laser signal to remove low-energy noise and retain the effective signal of VOCs. The noise filtering process adopts a combination of wavelet denoising and Kalman filtering, and the specific implementation steps are as follows:

[0077] Wavelet denoising: First, the normalized signal is decomposed into wavelet coefficients of different frequency scales through wavelet transform. After decomposition, the low energy coefficients are removed and the high energy coefficients are retained to reduce the interference of background noise.

[0078] Kalman filtering: Secondly, the signal quality is further optimized through Kalman filtering. The filtering process describes the signal characteristics of the system by setting the state equation and observation equation, calculates the Kalman gain in a recursive manner, and iteratively updates the signal estimate. The signal after Kalman filtering has a high signal-to-noise ratio, which meets the data requirements of the subsequent spectral extraction module.

[0079] Through the collaborative work of the above units, the data acquisition module can stably and accurately obtain VOCs distribution data in industrial source areas, providing high-quality initial input signals for subsequent modules of the system.

[0080] Spectral extraction module

[0081] The spectrum extraction module extracts characteristic spectrum information related to the VOCs concentration from the laser echo signal transmitted by the data acquisition module, providing basic data for subsequent concentration modeling and inversion optimization. The spectrum extraction module accurately identifies the characteristic absorption peaks of VOCs by performing frequency domain conversion and characteristic peak analysis on the laser signal, and calculates the absorption intensity of a specific band, providing reliable input for the concentration model. The following is a specific implementation of the spectrum extraction module in this embodiment.

[0082] In this embodiment, the spectrum extraction module includes a Fourier transform unit, a characteristic peak identification unit and a signal strength calculation unit:

[0083] 1. Fourier transform unit:

[0084] In this embodiment, the Fourier transform unit is used to perform frequency domain conversion on the normalized signal output by the data acquisition module to extract the characteristic spectral information of VOCs. After converting the time domain signal into the frequency domain signal, the characteristic absorption peaks of VOCs can be identified more intuitively. The calculation formula of Fourier transform is as follows:

[0085]

[0086] Among them, S(t) is the laser echo signal at time t, and S(f) is the frequency domain signal at frequency f. Through this frequency domain transformation, the interference noise in the time domain can be effectively filtered, and the characteristic frequency of VOCs can be clearly displayed, laying the foundation for subsequent characteristic peak identification.

[0087] 2. Characteristic peak recognition unit:

[0088] In this embodiment, the characteristic peak identification unit is used to identify the characteristic absorption peaks of VOCs from the frequency domain signal after Fourier transformation. The characteristic absorption peak is a significant absorption signal of VOCs in a specific band and is an important sign to distinguish the types and concentrations of VOCs. The characteristic peak identification unit sets an appropriate threshold to screen out the significant absorption peak position and confirms the characteristic peak based on the frequency and intensity of the peak to ensure data accuracy. By identifying these characteristic peaks, basic data of band absorption intensity is provided for subsequent concentration modeling.

[0089] 3. Signal strength calculation unit:

[0090] In this embodiment, the signal intensity calculation unit is used to integrate the intensity of the characteristic absorption peak to obtain the absorption intensity I of a specific band. i , providing the necessary input parameters for concentration modeling. The absorption intensity is calculated as follows:

[0091]

[0092] Among them, I i is the absorption intensity of the ith band, f1 and f2 are the start and end frequencies of the characteristic peak, and S(f) is the amplitude of the frequency domain signal. By integrating the characteristic peaks, the absorption intensity of each characteristic band can be obtained, providing quantitative spectral features and inputting stable and accurate data for the concentration modeling module.

[0093] Through the above units, the spectral extraction module realizes the specific function of extracting VOCs characteristic spectra from laser echo signals, providing high-quality data support for subsequent concentration modeling and inversion optimization.

[0094] Preliminary Modeling Module

[0095] The preliminary modeling module uses the characteristic spectral data provided by the spectrum extraction module to establish a basic model of VOCs concentration. This module uses multiple linear regression analysis to associate the absorption intensity of different bands with VOCs concentration, providing an initial concentration estimate for subsequent inversion optimization. The preliminary modeling module achieves a rough estimate of VOCs concentration, ensuring the data quality and model stability of subsequent optimization steps.

[0096] In this embodiment, the implementation of the preliminary modeling module includes a regression analysis unit, a regression coefficient calculation unit and an error evaluation unit:

[0097] 1. Regression analysis unit:

[0098] In this embodiment, the regression analysis unit establishes a preliminary estimation model of VOCs concentration through a multivariate linear regression model based on the characteristic spectrum data generated by the spectrum extraction module. This model is used to describe the relationship between VOCs concentration C and absorption intensity I in different bands.i The linear relationship between them can be expressed as:

[0099]

[0100] Among them, C represents the concentration of VOCs, β0 is the intercept term, β i is the regression coefficient of the absorption intensity of the i-th band, I i is the absorption intensity of the i-th band, and ∈ is the error term. This regression model estimates the preliminary concentration distribution of VOCs by capturing the relationship between spectral data and concentration, providing initial prediction values ​​for further optimization.

[0101] 2. Regression coefficient calculation unit:

[0102] In this embodiment, the regression coefficient calculation unit is used to calculate the coefficient β of the regression model using the least squares method. i The estimation is performed to minimize the error between the model prediction and the actual observation. The regression coefficient is calculated as follows:

[0103] β=(X T X) -1 X T Y

[0104] Among them, X is the absorption intensity matrix, each column of which corresponds to the absorption intensity data of a band, Y is the observed VOCs concentration vector, and β is the regression coefficient vector. The regression coefficient is calculated by this formula, which enables the model to optimally fit the relationship between spectral data and VOCs concentration, ensuring the accuracy of the preliminary concentration estimation.

[0105] 3. Error evaluation unit:

[0106] In this embodiment, the error evaluation unit is used to evaluate the prediction error of the regression model, and the accuracy of the preliminary modeling is determined by calculating the mean square error (MSE) of the model. The calculation formula of the mean square error is:

[0107]

[0108] Where N is the number of samples, C i is the actual observed VOCs concentration, is the VOCs concentration predicted by the model. The mean square error reflects the deviation between the model prediction value and the actual value. By evaluating the error, the reliability of the preliminary model can be effectively verified to ensure the stability of the parameter basis of the subsequent inversion optimization module.

[0109] Through the above steps, the preliminary modeling module achieved a preliminary prediction of VOCs concentration, provided effective initial parameters for the system's inversion optimization module, and ensured data consistency and model stability in the subsequent optimization process.

[0110] Inverse Optimization Module

[0111] The inverse optimization module further optimizes the VOCs concentration estimates provided by the preliminary modeling module to improve the accuracy and adaptability of the model. This module uses an improved atmospheric transport model and nonlinear optimization algorithm combined with a genetic algorithm to globally optimize the model parameters to obtain a more accurate estimate of the VOCs concentration distribution. The inverse optimization module is more adaptable in complex industrial environments and can effectively reduce the impact of noise interference on concentration estimates.

[0112] In this embodiment, the implementation of the inversion optimization module includes an atmospheric transmission model unit, a nonlinear inversion unit and a parameter optimization unit:

[0113] 1. Atmospheric transport model unit:

[0114] In this embodiment, the atmospheric transmission model unit constructs an atmospheric transmission model based on the multiple scattering theory to describe the absorption and scattering effects of laser propagation in an industrial environment. When the laser propagates in the atmosphere, its signal intensity will be attenuated by the VOCs molecules and particulate matter in the air. The improved atmospheric transmission model can simulate these effects more accurately, and its model expression is:

[0115] I(x)=I0e -αx +∫0 x k(x′)I(x′)dx′

[0116] Among them, I(x) represents the intensity of the laser at a distance x, I0 is the initial laser intensity, α is the VOCs absorption coefficient, and k(x′) is the scattering coefficient. Through this model, the signal attenuation in the laser propagation path can be calculated while considering atmospheric multiple scattering and non-uniform distribution, thereby improving the accuracy of VOCs concentration estimation.

[0117] 2. Nonlinear inversion unit:

[0118] In this embodiment, the nonlinear inversion unit uses the Levenberg-Marquardt optimization algorithm to optimize the preliminary concentration estimate of the atmospheric transport model. The Levenberg-Marquardt algorithm combines gradient descent and Newton's method and is suitable for nonlinear optimization problems. The unit optimizes the model parameters by minimizing the error between the model prediction value and the actual observation value. Its optimization objective function is:

[0119]

[0120] Among them, C i is the actual observed VOCs concentration, f(I i; θ) is the predicted concentration calculated based on the laser intensity, and θ is the model parameter to be optimized. The optimization process iteratively adjusts the parameters to make the model prediction value closer to the actual observed data, ensuring high-precision inversion of VOCs concentration.

[0121] 3. Parameter optimization unit:

[0122] In this embodiment, the parameter optimization unit uses a genetic algorithm to globally optimize the parameters of the inversion model to further improve the adaptability and accuracy of the model. The genetic algorithm can perform a global search in a multidimensional parameter space by simulating the selection, crossover and mutation operations in the biological evolution process. The optimization steps of the genetic algorithm include:

[0123] Initialize the population: randomly generate a set of parameter sets as the initial population;

[0124] Fitness evaluation: Evaluate the fitness of each set of parameters by calculating the objective function value;

[0125] Selection and crossover: Select appropriate parameters according to fitness for crossover operation to generate a new generation of parameters;

[0126] Mutation: mutate the generated parameter set to increase the diversity of the population;

[0127] Iterative update: Repeat the above steps until the optimization conditions are met.

[0128] Through the global search characteristics of genetic algorithms, we can effectively avoid falling into local optimality and ensure the global optimal solution of the model in complex environments.

[0129] The inversion optimization module achieves the optimization processing of VOCs concentration estimation through the synergistic effect of the above units, combined with the multiple scattering model, nonlinear inversion algorithm and genetic algorithm, making the system more adaptable and accurate in complex industrial environments.

[0130] Intelligent analysis module

[0131] The intelligent analysis module is used to conduct further in-depth analysis of VOCs concentration data to improve the accuracy of concentration prediction and the adaptability of the system. This module uses a deep learning model to perform nonlinear analysis on the spectral data of laser echoes, and combines feature selection and optimization methods to extract feature data closely related to VOCs concentration, thereby achieving accurate prediction of concentration distribution.

[0132] In this embodiment, the implementation of the intelligent analysis module includes a deep learning model unit, a model training unit, an optimizer unit, and a feature selection unit. The detailed description of each unit is as follows:

[0133] 1. Deep learning model unit:

[0134] In this embodiment, the deep learning model unit constructs a deep neural network (DNN) model based on the VOCs spectral data provided by the spectral extraction module to perform nonlinear inversion prediction on the VOCs concentration. The deep neural network adopts a multi-layer structure, and the input is the spectral data feature X = {I1, I2, ..., I n}, the output is the predicted VOCs concentration C. The mathematical expression of the model is:

[0135] C=f(W n ·σ(…σ(W2·σ(W1·X+b1)+b2)…)+b n )

[0136] Among them, W i and b i are the weight and bias of the i-th layer respectively, and σ is the activation function. The multi-layer nonlinear structure of the network can capture the nonlinear relationship between complex spectral features and VOCs concentration, thereby improving the accuracy of concentration prediction.

[0137] 2. Model training unit:

[0138] In this embodiment, the model training unit is used to train the deep neural network model and optimize the network parameters by minimizing the loss function between the predicted value and the actual value. The loss function adopted is the mean square error (MSE). By minimizing the loss function, the model training unit adjusts the weight and bias parameters of the deep neural network so that the concentration value output by the model is closer to the actual value, thereby improving the prediction accuracy.

[0139] 3. Optimizer unit:

[0140] In this embodiment, the optimizer unit uses the gradient descent method to optimize the parameters in the model training process. The gradient descent algorithm calculates the gradient of the loss function relative to each parameter to determine the adjustment direction and step size of the weights and biases. The parameter update formula of the gradient descent is:

[0141]

[0142] Among them, W i is the weight of the current iteration, and η is the learning rate. This method can quickly converge to the optimal solution, ensuring the stability of model parameters and training efficiency.

[0143] 4. Feature selection unit:

[0144] In this embodiment, the feature selection unit is used to select characteristic bands that contribute more to the prediction of VOCs concentration from the spectral data to optimize the input data of the model and improve the calculation efficiency. The feature selection adopts the recursive feature elimination (RFE) method to select the optimal feature subset by calculating the weight of each feature. The feature selection process of RFE can be expressed as:

[0145]

[0146] Among them, w j is the weight of the jth feature, m is the total number of features, and the feature with the highest score is selected for model training. The feature selection unit can effectively reduce the data dimension and improve the computational efficiency and generalization ability of the model.

[0147] The intelligent analysis module achieves accurate prediction of VOCs concentration and optimization of data processing through deep learning, model optimization and feature selection, enabling the system to more accurately identify and analyze the concentration distribution of VOCs from industrial sources and meet monitoring needs in complex environments.

[0148] Visualization Module

[0149] The visualization module is used to present the VOCs concentration distribution information obtained from the analysis in a graphical form, so as to intuitively display the distribution of pollutants in space. Through data interpolation, triangulation and three-dimensional rendering, the module converts VOCs concentration data into a three-dimensional thermal map that is easy to observe and analyze, providing users with a basis for real-time monitoring and analysis.

[0150] In this embodiment, the implementation of the visualization module includes a data interpolation unit, a triangulation unit, a three-dimensional rendering unit, and a dynamic update unit:

[0151] 1. Data interpolation unit:

[0152] In this embodiment, the data interpolation unit is used to perform spatial interpolation on the VOCs concentration data to generate a continuous concentration distribution. The interpolation calculation uses the Kriging interpolation method to convert discrete concentration data into estimated concentration values ​​for each point in space. The mathematical expression of Kriging interpolation is:

[0153]

[0154] in, is the estimated concentration at the spatial point (x, y, z), C i is the concentration value of the known data point, λ i is the weight coefficient, which is calculated by minimizing the estimation error. This interpolation method can effectively smooth the data distribution and provide continuous concentration data input for three-dimensional rendering.

[0155] 2. Triangulation unit:

[0156] In this embodiment, the triangulation unit is used to perform spatial grid division on the interpolated data, and generate a three-dimensional grid model through Delaunay triangulation to discretize the spatial data. The mathematical definition of Delaunay triangulation is: for any triangle in the point set P, its circumscribed circle does not contain other points in the point set. The triangulation unit divides the spatial point set according to this definition to ensure that the generated three-dimensional grid has good uniformity and data coverage.

[0157] 3. 3D rendering unit:

[0158] In this embodiment, the three-dimensional rendering unit is used to render the three-dimensional mesh model generated by the triangulation unit to generate a visualized three-dimensional heat map. The rendering process uses a color mapping method to correspond different VOCs concentration values ​​to different colors, thereby realizing the visualization of concentration. Through this mapping relationship, different concentration areas are displayed in different colors to form a heat map of concentration distribution, which is convenient for intuitively identifying high and low areas of pollutant concentration.

[0159] 4. Dynamic update unit:

[0160] In this embodiment, the dynamic update unit is used to dynamically and visually update the real-time monitored VOCs concentration data to ensure that the thermal map can reflect the changes in pollutant concentration in real time. After receiving the new concentration data, the dynamic update unit re-interpolates, triangulates and renders, and displays the updated data on the thermal map, so that the user can monitor the real-time diffusion path and concentration change trend of the VOCs concentration.

[0161] Through the above steps, the visualization module realizes the graphical display of VOCs concentration distribution, converts complex data information into intuitive three-dimensional thermal maps, and provides effective visualization support for real-time monitoring and analysis of the system.

[0162] Dynamic monitoring and early warning module

[0163] The dynamic monitoring and early warning module is used to monitor VOCs concentration in real time and issue early warning signals according to the set concentration threshold and concentration change trend. Through time series analysis, threshold judgment, trend detection and other functions, the module detects abnormal VOCs concentration and triggers early warnings to respond to possible pollution events in a timely manner.

[0164] In this embodiment, the implementation of the dynamic monitoring and early warning module includes a time series analysis unit, a concentration threshold judgment unit, a trend anomaly detection unit and an early warning information output unit:

[0165] 1. Time Series Analysis Unit:

[0166] In this embodiment, the time series analysis unit is used to perform time series modeling on the historical VOCs concentration data to predict the changing trend of VOCs concentration in the future. The unit uses the autoregressive moving average model (ARIMA) to generate predicted values ​​based on historical data to provide a reference for dynamic monitoring. The expression of the ARIMA model is:

[0167] C t =φ1C t-1 +φ2C t-2 +…+θ1∈ t-1 +θ2∈ t-2 +…+∈ t

[0168] Among them, C t is the VOCs concentration at time t, φ i and θ i are the autoregressive coefficient and the moving average coefficient, ∈ t is a white noise term. Through time series analysis, future concentration changes can be predicted, which is convenient for identifying potential concentration anomalies in advance.

[0169] 2. Concentration threshold judgment unit:

[0170] In this embodiment, the concentration threshold judgment unit is used to compare the VOCs monitoring concentration value with the set safety threshold C in real time. thredh When the concentration value exceeds the threshold, the system will trigger an early warning signal. The early warning condition expression is:

[0171] C t >C thresh

[0172] Among them, C t is the VOCs concentration monitored at the current moment, C thredh Safety thresholds set for the system. By monitoring the thresholds of real-time data, the unit can quickly issue an alarm when the concentration exceeds the safety limit, ensuring timely response.

[0173] 3. Trend anomaly detection unit:

[0174] In this embodiment, the trend anomaly detection unit is used to identify the abnormal growth trend of VOCs concentration. The concentration change rate ΔC is calculated to determine whether the concentration has an abnormal increase trend. The change rate calculation formula is:

[0175]

[0176] Among them, C t and C t-1are the VOCs concentrations at the current moment and the previous moment, respectively, and Δt is the time interval. When ΔC exceeds the preset growth rate threshold, the system determines that the concentration growth is abnormal and triggers a trend warning signal. This trend detection enables the system to capture the potential risk of rapid concentration increases.

[0177] 4. Warning information output unit:

[0178] In this embodiment, the warning information output unit is used to display the warning signal and abnormal monitoring results to the user interface, and output the current concentration value, change rate and future concentration prediction curve, so that the user can make judgments and respond according to the real-time information provided by the system. The unit integrates various key data into intuitive information to assist users in analyzing the spread of pollution sources and possible risk levels.

[0179] The dynamic monitoring and early warning module realizes real-time monitoring and abnormal early warning of VOCs concentration through the coordinated work of the above units, effectively helping the system to respond in time when abnormal concentration is detected.

[0180] The industrial source VOCs distribution analysis system based on laser radar of the present invention realizes a complete monitoring process from data collection to concentration prediction and early warning through the collaborative work of multiple modules. The following is the operation process of the present invention:

[0181] 1. Data collection and preprocessing

[0182] The data acquisition module uses laser radar to collect VOCs distribution data in industrial source areas in a specific band (such as ultraviolet or infrared). The laser emitting unit emits a laser signal in a specific band. When passing through the VOCs area, part of the light is absorbed or scattered, and the remaining signal is returned to the laser receiving unit.

[0183] Preprocessing: The data acquisition module performs signal normalization and noise filtering on the received laser signal. Signal normalization removes the influence of environmental changes, and noise filtering (wavelet denoising and Kalman filtering) further improves the signal quality and ensures data consistency and validity.

[0184] 2. Spectral feature extraction

[0185] The spectrum extraction module performs Fourier transform on the preprocessed laser signal, converts the time domain signal into a frequency domain signal, and extracts the characteristic absorption peaks of VOCs.

[0186] Characteristic peak identification: The characteristic absorption peaks of VOCs are identified in the frequency domain signal, and the absorption intensity of specific bands is obtained by integral calculation. These absorption intensities are used as input parameters for subsequent concentration models.

[0187] 3. Preliminary Modeling

[0188] The preliminary modeling module builds a preliminary model of VOCs concentration based on spectral feature data. The absorption intensity of different bands is associated with VOCs concentration using a multiple linear regression method.

[0189] Regression coefficient calculation: The regression coefficient was estimated by the least squares method to minimize the model prediction error and obtain a preliminary estimate of the VOCs concentration.

[0190] 4. Inversion Optimization

[0191] The inverse optimization module further optimizes the preliminary modeling results to improve the accuracy of VOCs concentration prediction.

[0192] Atmospheric transmission models are used to describe the multiple scattering and absorption characteristics of laser signals in industrial environments.

[0193] A nonlinear inversion algorithm (Levenberg-Marquardt algorithm) is used to minimize the error between the observed and predicted values ​​and further optimize the model parameters.

[0194] The genetic algorithm is used to globally optimize the model parameters to ensure the adaptability and stability of the model in complex environments.

[0195] 5. Intelligent analysis

[0196] The intelligent analysis module uses deep neural networks to perform nonlinear analysis on spectral data to further improve the prediction accuracy of VOCs concentration.

[0197] Model training: The model is trained using historical data and the mean square error is used as the loss function to optimize the model parameters.

[0198] Feature selection: The recursive feature elimination method is used to screen the characteristic bands that contribute greatly to concentration prediction, reduce the dimension of input data, and improve the computational efficiency and accuracy of the model.

[0199] 6. Visualization

[0200] The visualization module generates a three-dimensional heat map of VOCs concentration, intuitively showing the distribution of pollutants in space.

[0201] Data interpolation: The Kriging interpolation method is used to smooth the VOCs concentration data to generate a continuous spatial concentration distribution.

[0202] Triangulation and rendering: Delaunay triangulation and color mapping are used to map VOCs concentration values ​​to different colors to generate heat maps, providing users with intuitive pollution distribution information.

[0203] 7. Dynamic monitoring and early warning

[0204] The dynamic monitoring and early warning module is used to monitor the changes in VOCs concentration in real time and issue early warning signals when abnormal conditions are detected.

[0205] Time series analysis: ARIMA model is used to analyze historical data and predict future concentration trends.

[0206] Threshold judgment and trend detection: When the monitored concentration value exceeds the safety threshold or rises rapidly, an early warning signal is triggered.

[0207] Early warning output: Output current concentration, change rate and future concentration trend to assist users in responding to potential pollution risks in real time.

[0208] Through the above steps, the operation process of the present invention covers the whole process from data collection, feature extraction, modeling optimization, intelligent analysis, visualization to real-time monitoring and early warning. The modules work together to finally achieve high-precision monitoring and early warning of VOCs concentrations in industrial sources, ensuring that users can effectively manage VOCs pollution in industrial environments.

[0209] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The industrial source VOCs distribution analysis system based on laser radar is characterized by: include: The data acquisition module is used to collect VOCs distribution data in industrial source areas in a specific band through lidar and perform preprocessing; Spectral extraction module, used to extract spectral information related to VOCs characteristics from the processed lidar signal; A preliminary modeling module, used to establish a preliminary model of VOCs concentration based on spectral information; Inversion optimization module, used to optimize the preliminary model of VOCs concentration through inversion algorithm to obtain accurate VOCs concentration distribution; The intelligent analysis module is used to perform machine learning and deep learning analysis on VOCs concentration distribution to improve the accuracy of concentration prediction.

2. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The data acquisition module comprises: A laser emitting unit, used for emitting a laser signal of a specific wavelength band, wherein the wavelength band is an ultraviolet wavelength band of 266nm or 355nm or an infrared wavelength band of 1064nm; A laser receiving unit is used to receive an echo signal of a laser signal emitted by a laser emitting unit after being scattered or absorbed in a VOCs area, and generate corresponding spectral data; The preprocessing unit is used to perform normalization processing and noise filtering processing on the laser echo signal.

3. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The spectrum extraction module comprises: A Fourier transform unit, used to perform Fourier transform on the normalized lidar signal to extract the spectral characteristic signal of VOCs; A characteristic peak identification unit, used to identify characteristic absorption peaks of VOCs from the frequency domain signal after Fourier transformation, and generate characteristic spectrum data related to VOCs; The signal intensity calculation unit is used to perform integral calculation on the characteristic absorption peak to obtain the absorption intensity I in a specific band. i As the input parameter of the VOCs concentration model, the calculation formula of the absorption intensity is: Among them, S(f) is the frequency domain signal, which is used to represent the absorption characteristics at different frequencies, I i is the absorption intensity of the i-th band, and f1 and f2 are the start and end frequencies of the characteristic absorption peak.

4. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The preliminary modeling module includes: The regression analysis unit is used to establish a multivariate linear regression model of VOCs concentration based on the characteristic spectrum data generated by the spectrum extraction module to achieve a preliminary prediction of VOCs concentration. The expression of the model is: Among them, C is the initial concentration prediction value of VOCs, I i is the absorption intensity of the i-th band, β0 and β i are the intercept and regression coefficient of the regression model respectively, ∈ is the error term; Regression coefficient calculation unit, used to estimate the coefficient β of the regression model based on the least squares method i , so that the prediction error of the regression model for VOCs concentration is minimized; The error evaluation unit is used to calculate the mean square error of the regression model prediction results to evaluate the accuracy of the preliminary modeling.

5. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The inversion optimization module includes: The atmospheric transmission model unit is used to establish an improved multiple scattering model to describe the propagation characteristics of the laser in the industrial source environment and perform a preliminary inversion calculation of the VOCs concentration. The atmospheric transmission model expression is: I(x)=I0e -ax +∫0 x k(x′)I(x′)dx′ Where I(x) represents the intensity of the laser at a distance of x, I0 is the initial laser intensity, α is the atmospheric absorption coefficient, and k(x′) is the scattering coefficient. This model optimizes the light transmission path based on the consideration of the atmospheric multiple scattering effect. The nonlinear inversion unit is used to optimize the preliminary inversion results of the atmospheric transport model based on the nonlinear inversion algorithm to improve the calculation accuracy of the VOCs concentration. The nonlinear inversion algorithm adopts the Levenberg-Marquardt optimization method, and its objective function is: Among them, C i is the actual observed VOCs concentration, f(I i ; θ) is the predicted concentration calculated according to the laser intensity, and θ is the model parameter to be optimized.

6. The industrial source VOCs distribution analysis system based on laser radar according to claim 5 is characterized in that: The inverse optimization module also includes: The parameter optimization unit is used to globally optimize the parameters in the inversion model based on the genetic algorithm to further improve the adaptability and accuracy of the model in various industrial environments. The specific implementation of the genetic algorithm includes the following steps: Initialize the population and randomly generate a set of initial parameter sets θ; Fitness evaluation, calculating the fitness value of each set of parameters based on the objective function; Selection and crossover: select parameters according to fitness and perform crossover operations to generate new parameter groups; Mutation, mutating the newly generated parameter group to increase the diversity of solutions; Iterate and update, repeating the above steps until the optimization conditions are met.

7. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The intelligent analysis module includes: A deep learning model unit is used to perform nonlinear inversion analysis based on the VOCs spectral data generated by the spectral extraction module. The deep learning model unit adopts a deep neural network structure, and the input is the spectral data feature X = {I1, I2, ..., I n }, the output is the predicted value C of VOCs concentration, and the mathematical expression of its neural network is: C=f(W n ·σ(…σ(W2·σ(W1·X+b1)+b2)…)+b n ) Among them, W i and b i are the weight and bias of the i-th layer respectively, and σ is the activation function, which can learn and fit the nonlinear features in the spectral data, thereby improving the accuracy of concentration prediction; Model training unit, used to train the deep neural network model based on the historical data of VOCs, and optimize the weights and biases of the network by minimizing the loss function between the predicted concentration and the actual concentration; The optimizer unit is used to optimize the value of the loss function using the gradient descent method during model training; The feature selection unit is used to screen out the characteristic bands that contribute more to the prediction of VOCs concentration from the spectral data, and select the optimal feature subset through the recursive feature elimination method.

8. The industrial source VOCs distribution analysis system based on laser radar according to claim 1 is characterized in that: The system also includes a visualization module for generating a three-dimensional thermal distribution map of VOCs.

9. The industrial source VOCs distribution analysis system based on laser radar according to claim 1, characterized in that: The system also includes a dynamic monitoring and early warning module for real-time monitoring of changes in VOCs concentration and abnormal early warning.

10. The industrial source VOCs distribution analysis system based on laser radar according to claim 9, characterized in that: The dynamic monitoring and early warning module includes: Time series analysis unit, used to establish a time series model based on historical VOCs concentration data to predict the changing trend of VOCs concentration in the future; A concentration threshold judgment unit is used to compare the VOCs concentration value monitored in real time with a preset safety threshold, and trigger an early warning signal when it is detected that the concentration value exceeds the threshold; The trend anomaly detection unit is used to identify the abnormal growth trend of VOCs concentration and determine whether there is an abnormality by calculating the concentration change rate; The warning information output unit is used to output warning signals and abnormal monitoring results to the user interface, and provide the diffusion path and trend information of VOCs concentration, including the current concentration value, change rate and predicted concentration curve, so that users can respond and process in time.

Images