A PM2.5 prediction method and system based on Copula entropy correlation analysis

Abstract

The present application relates to the technical field of air pollution prevention and control, and specifically to a PM2.5 prediction method and system based on Copula entropy correlation analysis, the method comprising obtaining historical data of atmospheric pollutants and historical data of meteorological factors, sorting the influencing factors in the historical data of atmospheric pollutants and historical data of meteorological factors based on Copula entropy correlation analysis, obtaining the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting, sorting the influencing factors in the historical data of atmospheric pollutants and historical data of meteorological factors based on Copula entropy correlation analysis, obtaining the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting, determining the parameters in the LSTM model based on AGA, predicting pollutants based on the LSTM model, and obtaining PM2.5 prediction results. CE is used to realize the correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants, and realize pollutant prediction.

Description

Technical Field

[0001] The present application relates to the technical field of air pollution prevention and control, and in particular to a PM2.5 prediction method and system based on Copula entropy correlation analysis. Background Art

[0002] With the deepening of my country's urbanization process, the improvement of productivity, and the transformation of production patterns, the types of air pollution have also undergone significant changes in this process. In addition to the pollution of traditional pollutants such as suspended particulate matter and SO2, which are still severe, the harmfulness of pollutants such as O3 and nitrogen oxides has gradually manifested, and the scope of influence has gradually expanded. In order to better prevent and control air pollution and effectively take preventive measures, it is very necessary to accurately predict the concentration of urban air pollutants, which has positive guidance and reference significance for the scientific decision-making of industrial structure and the formulation of environmental pollution prevention and control measures.

[0003] However, in the actual production, life and scientific research, we often encounter situations where multiple factors affect a thing or phenomenon at the same time. Since there is generally a certain correlation between multiple variables, whether strong or weak, the information overlaps to a certain extent, hindering the in-depth analysis of things or phenomena. Correlation analysis is a statistical method to analyze whether two or more things or phenomena are correlated and the strength of their correlation.

[0004] Among them, the measurement of correlation was proposed and studied in the early days of statistics. The most widely used measurement is the Pearson Correlation Coefficient (PCC). However, PCC is limited to the linear Gaussian case. Therefore, it is often powerless in the face of complex natural phenomena full of nonlinearity and its application scope is very limited. If it is applied without considering the premise, the conclusion drawn is also unreliable. Copula Entropy (CE) is a mathematical concept used to measure statistical independence relationships strictly defined by Ma Jian et al. [2] in 2008. CE is an ideal statistical independence measure. As a correlation analysis tool, it is suitable for the analysis of any nonlinear and non-Gaussian correlation and does not require any assumptions. Therefore, it is a good method for causal discovery.

[0005] Given that the main factors affecting changes in atmospheric pollutant concentrations include not only the mutual transformation between atmospheric pollutants caused by chemical reactions, but also the impact of different meteorological factors and geographical features on the atmospheric environment. Since regional geographical features are relatively stable, it is necessary to focus on the correlation between atmospheric pollutants and the correlation between atmospheric pollutants and meteorological factors. The generation and flow of atmospheric pollutants are closely related to the local climate environment. Different meteorological conditions also have a great impact on the diffusion and transport of pollutants. Summary of the invention

[0006] In order to solve the problem of predicting and analyzing the impact of meteorological conditions on the diffusion and transport of pollutants, the present invention provides a PM2.5 prediction method and system based on Copula entropy correlation analysis. By analyzing the correlation of multiple meteorological factors and multiple atmospheric pollutants, the correlation of the influencing factors is sorted according to the size of the correlation coefficient, thereby determining the contribution rate of meteorological factors to the pollutant index and realizing pollutant prediction.

[0007] To achieve the above objectives, the embodiments of the present invention provide the following technical solutions:

[0008] In a first aspect, in an embodiment provided by the present invention, a PM2.5 prediction method based on Copula entropy correlation analysis is provided, and the PM2.5 prediction method comprises the following steps:

[0009] Obtain historical data on air pollutants and meteorological factors;

[0010] Based on the Copula entropy correlation analysis, the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors are sorted to obtain the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting;

[0011] The Copula entropy correlation analysis is used again to normalize the correlation coefficient between the main influencing factors and PM2.5, and the correlation coefficient is used as the weight of each main influencing factor and brought into the constructed AGA-LSTM prediction model;

[0012] The parameters in the LSTM model are determined based on AGA, and the pollutants are predicted based on the LSTM model to obtain PM2.5 prediction results.

[0013] As a further solution of the present invention, the historical data of atmospheric pollutants include 5 kinds of atmospheric pollutants, namely PM10, NO2, SO2, O3, and CO; the historical data of meteorological factors include 8 kinds of meteorological factors, namely air temperature, perceived temperature, air pressure, humidity, rainfall, wind direction, wind force, and wind speed.

[0014] As a further solution of the present invention, the Copula entropy (CE) is based on Copula theory. The multivariate probability density function in the Copula theory characterizes the product of the number of edges and the Copula density function. The Copula density function characterizes the dependency structure between random variables. The Copula entropy is a non-parametric method for estimating CE or MI from data. The non-parametric method includes: estimating the empirical Copula density function and estimating CE.

[0015] As a further solution of the present invention, the MI is mutual information, which is used to reflect the size of the shared information between different variables. The method for calculating the mutual information between two random variables X and Y is:

[0016] Calculate the mutual information between random variables X and Y according to the formula:

[0017]

[0018] Among them, x i and i are samples of random variables X and Y respectively, and N is the sample size.

[0019] As a further solution of the present invention, estimating the empirical Copula density function includes:

[0020] Given a random variable X = {x1,…,x N} T The generated independent and identically distributed data samples {X1,…,X T};

[0021] The estimated empirical Copula density function is:

[0022] Where χ represents the indicator function when i = 1,…,N; assuming u = [F1,…,F N ] derive a new sample set {u1,…,u T}Data c(u) as the empirical Copula density function.

[0023] As a further solution of the present invention, the Copula entropy correlation analysis is based on sorting the influencing factors in the historical data of air pollutants and the historical data of meteorological factors, including:

[0024] PM2.5 was compared with 8 meteorological factors and 5 air pollutants respectively by using the non-parametric estimation CE method to rank the influencing factors;

[0025] According to the correlation with PM2.5, the first four influencing factors were taken as the main influencing factors, and 8 influencing factors were obtained.

[0026] As a further solution of the present invention, the constructed AGA-LSTM prediction model is constructed by integrating AGA and LSTM to predict PM2.5. The construction method of the AGA-LSTM prediction model includes the following steps:

[0027] Step S101, binary encoding the number of hidden layer neurons, number of training times, and learning rate of the parameters in the LSTM;

[0028] Step S102: Generate an initial population N based on the binary encoding of the parameters in the LSTM, where N is an even number;

[0029] Step S103: Establish an LSTM model, train and predict the data of the training set and the test set, and use the predicted mean square error as the AGA fitness value f i ;

[0030] Step S104: select N individuals according to the roulette rule and calculate f avg and f max ;

[0031] Step S105: randomly pair the individuals in the group into pairs, forming a total of N / 2 pairs, and calculate the adaptive crossover probability of each pair of individuals according to the adaptive formula Randomly generate R(0,1), if R<P c , then the crossover operation is performed on the pair of chromosomes;

[0032] Step S106: For all individuals N in the population, calculate the adaptive mutation probability according to the adaptive mutation formula If R<P m , then the chromosome is crossover operated;

[0033] Step S107, calculating the fitness of new individuals generated by crossover and mutation, and the new individuals and the parent generation together constitute a new population;

[0034] Step S108, determine whether the termination condition is met, if so, terminate and return to the optimal parameters, otherwise execute step S104;

[0035] Step S109: construct an LSTM network model using the optimal parameters obtained by AGA; train the model to output the prediction results.

[0036] As a further solution of the present invention, the parameters in the LSTM model are determined based on AGA, including determining three parameters of the number of hidden layer neurons, the number of training times, and the learning rate in the LSTM model through AGA.

[0037] As a further solution of the present invention, when establishing the LSTM model, the mean square error predicted by the LSTM model is used as the AGA fitness value, and the parameters of the LSTM model are adjusted based on the AGA fitness value.

[0038] In a second aspect, in another embodiment provided by the present invention, a PM2.5 prediction system based on Copula entropy correlation analysis is provided, wherein the PM2.5 prediction system based on Copula entropy correlation analysis is used to perform the above-mentioned PM2.5 prediction method based on Copula entropy correlation analysis; the PM2.5 prediction system based on Copula entropy correlation analysis includes:

[0039] A data acquisition module is used to acquire historical data of atmospheric pollutants and meteorological factors;

[0040] A correlation analysis module, for sorting the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors based on Copula entropy correlation analysis, and obtaining the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting;

[0041] A weight generation module is used to normalize the correlation coefficient between the main influencing factors and PM2.5 by using Copula entropy correlation analysis again, and bring it into the constructed AGA-LSTM prediction model as the weight of each main influencing factor;

[0042] The result prediction module is used to determine the parameters in the LSTM model based on AGA, predict pollutants based on the LSTM model, and obtain PM2.5 prediction results.

[0043] In addition, to achieve the above-mentioned purpose, the present application also provides a computer device, which includes a processor, a memory, and a PM2.5 prediction program based on Copula entropy correlation analysis stored in the memory and executable by the processor, wherein when the PM2.5 prediction program based on Copula entropy correlation analysis is executed by the processor, the steps of the PM2.5 prediction method based on Copula entropy correlation analysis as described above are implemented.

[0044] In addition, to achieve the above-mentioned purpose, the present application also provides a computer-readable storage medium, on which a PM2.5 prediction program based on Copula entropy correlation analysis is stored, wherein when the PM2.5 prediction program based on Copula entropy correlation analysis is executed by a processor, the steps of the PM2.5 prediction method based on Copula entropy correlation analysis as described above are implemented.

[0045] The technical solution provided by the present invention has the following beneficial effects:

[0046] The present application provides a PM2.5 prediction method and system based on Copula entropy correlation analysis, which uses CE to perform correlation analysis on multiple meteorological factors and multiple atmospheric pollutants, and sorts the correlation of the influencing factors according to the size of the correlation coefficient, so as to determine the contribution rate of meteorological factors to the pollutant index. The top N factors with the largest correlation with PM2.5 are taken as the main influencing factors; at this time, CE is used again to analyze the correlation between multiple main influencing factors and PM2.5, and the correlation coefficient is normalized as the weight of each main influencing factor, which is brought into the prediction model to realize pollutant prediction. The present application uses CE to realize the correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants.

[0047] These and other aspects of the present application will be more concise and understandable in the following description of the embodiments. It should be understood that the above general description and the following detailed description are only exemplary and explanatory and cannot limit the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or the prior art description. Obviously, the drawings described below are only some embodiments of the present application. In the drawings:

[0049] Figure 1 This is a flow chart of the PM2.5 prediction method based on Copula entropy correlation analysis in this application;

[0050] Figure 2 A flowchart of constructing an AGA-LSTM prediction model in the PM2.5 prediction method based on Copula entropy correlation analysis in this application;

[0051] Figure 3 This is a schematic diagram of the CE between various meteorological factors and PM2.5 in the PM2.5 prediction method based on Copula entropy correlation analysis in this application;

[0052] Figure 4 This is a schematic diagram of the CE of each pollutant factor and PM2.5 in the PM2.5 prediction method based on Copula entropy correlation analysis in this application;

[0053] Figure 5 It is a time series characteristic diagram of PM2.5 concentration at different times of each feature in the PM2.5 prediction method based on Copula entropy correlation analysis of this application;

[0054] Figure 6 This is a flow chart of the PM2.5 prediction method based on Copula entropy correlation analysis in this application;

[0055] Figure 7 This is the system block diagram of the PM2.5 prediction system based on Copula entropy correlation analysis in this application.

[0056] The purpose, features and advantages of this application will be further described in conjunction with the embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0057] Below, the present application is further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new embodiment.

[0058] It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0059] 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.

[0060] 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.

[0061] 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.

[0062] The embodiments of the present application provide a PM2.5 prediction method and system based on Copula entropy correlation analysis, which uses CE to implement correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants.

[0063] In some embodiments, the PM2.5 prediction method based on Copula entropy correlation analysis can be applied to a computer device, which may be a PC, a portable computer, a mobile terminal, or other device with display and processing functions, but is certainly not limited thereto.

[0064] Please refer to Figure 1 , Figure 1 The flowchart of the PM2.5 prediction method based on Copula entropy correlation analysis of the present application is shown in FIG. In an embodiment of the present application, the PM2.5 prediction method based on Copula entropy correlation analysis includes the following steps S10 to S30:

[0065] Step S10, obtaining historical data of air pollutants and historical data of meteorological factors.

[0066] In some embodiments, the historical data of atmospheric pollutants include five kinds of atmospheric pollutants, namely PM10, NO2, SO2, O3, and CO; the historical data of meteorological factors include eight kinds of meteorological factors, namely air temperature, perceived temperature, air pressure, humidity, rainfall, wind direction, wind force, and wind speed.

[0067] Step S20, based on Copula entropy correlation analysis, sorting the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors, and obtaining the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting.

[0068] Step S30, again using Copula entropy correlation analysis to normalize the correlation coefficient between the main influencing factors and PM2.5, as the weight of each main influencing factor, and bring it into the constructed AGA-LSTM prediction model.

[0069] Step S40, determining parameters in the LSTM model based on AGA, predicting pollutants based on the LSTM model, and obtaining PM2.5 prediction results.

[0070] The embodiment of the present application provides a PM2.5 prediction method based on Copula entropy correlation analysis, considering that the main factors affecting the change of atmospheric pollutant concentration include not only the mutual transformation caused by chemical reactions between various atmospheric pollutants, but also the influence of different meteorological factors and geographical features on the atmospheric environment. Since the regional geographical characteristics are relatively stable. This application focuses on the correlation between atmospheric pollutants and the correlation between atmospheric pollutants and meteorological factors. The generation and flow of atmospheric pollutants have a great relationship with the local climate environment. Different meteorological conditions also have a great influence on the diffusion and transportation of pollutants.

[0071] When verifying the relationship between the generation and flow of atmospheric pollutants and the local climate environment, the correlation between atmospheric pollutants and meteorological conditions in Zhumadian City was discussed, and it was found that different seasons and different meteorological conditions have a certain impact on the concentration of atmospheric pollution, among which atmospheric pollutants are negatively correlated with temperature and precipitation, and positively correlated with wind speed.

[0072] Since the pollutant factors that affect PM2.5 interact with each other and show strong nonlinear characteristics, the use of traditional analysis methods suitable for linear correlation will affect the accuracy of prediction. Therefore, this application uses CE to realize the correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants.

[0073] Based on the above Figure 1 In the embodiments shown, in some embodiments of the present application, the Copula entropy (CE) is based on the Copula theory, in which the multivariate probability density function represents the product of the number of edges and the Copula density function, and the Copula density function represents the dependency structure between random variables. The Copula entropy is a non-parametric method for estimating CE or MI from data. The non-parametric method includes: estimating the empirical Copula density function and estimating CE.

[0074] In the embodiments of the present application, mutual information (MI) as a mutual information indicator describing the mutual connection between different variables originates from information theory, which reflects the size of the shared information between different variables, that is, the larger the mutual information, the stronger the correlation between the two, and vice versa.

[0075] Among them, the MI is the mutual information, which is used to reflect the size of the shared information between different variables. The method for calculating the mutual information between two random variables X and Y is:

[0076] According to formula (1), the mutual information between random variables X and Y is calculated as:

[0077]

[0078] Among them, x i and i are samples of random variables X and Y respectively, and N is the sample size.

[0079] It is not difficult to find from formula (1) that when X and Y are unrelated, the MI value is close to 0, and when X and Y present a functional relationship, the MI value will be close to positive infinity. Compared with other similarity indicators, the mutual information index can not only reflect the nonlinear correlation between variables, but also any reversible transformation of random variables does not change the size of the mutual information.

[0080] In the embodiments of the present application, since the mutual information index has a strong information mining capability, but in actual research, the joint distribution between different random variables is often skewed and inhomogeneous, it is difficult to find a suitable distribution type to fit it.

[0081] To solve this problem, the present application introduces Copula theory. Copula theory is to represent multivariate dependency using Copula function. Among them, Copula entropy (CE) is based on Copula theory, and the core of Copula theory is Sklar theorem. The Sklar theorem points out that the multivariate probability density function can be expressed as the product of its number of edges and the Copula density function, and the Copula density function represents the dependency structure between random variables. This representation separates the dependency structure, i.e., the Copula function, and has the properties of the edge of the individual variable, so that no matter the joint distribution or the marginal distribution, only the dependency structure can be processed.

[0082] In the embodiment of the present application, Copula entropy is defined as follows: Let X be a random variable with marginal distribution u and associated density c(u), and the CE of X is defined as:

[0083] H c (X) = -∫ u c(u)logc(u)du

[0084] In information theory, mutual information and entropy are two different concepts. But Ma and Sun proved that they are essentially the same, and MI is also a kind of entropy, with the following relationship:

[0085] The MI of a random variable is equivalent to a negative CE, that is:

[0086] I(X)=-H c (X)

[0087] This application proposes a simple nonparametric method for estimating CE or MI from data, which consists of only two steps:

[0088] Step 11, estimating the empirical Copula density function (ECD);

[0089] Step 12: Estimate CE.

[0090] For step 11, given the random variable X = {x1,…,x N} T The generated independent and identically distributed data samples {X1,…,X T}, the ECD can be easily estimated as follows:

[0091]

[0092] Where χ represents the indicator function when i = 1,…,N, assuming u = [F1,…,F N ] can derive a new sample set {u1,…,u T} as ECD data c(u).

[0093] In an embodiment of the present application, once the ECD is estimated, step 12 is essentially an entropy estimation problem, and many existing methods can solve this problem, using the KNN method. Therefore, using rank statistics and the KNN method, a non-parametric method for estimating CE can be derived, which can be applied to any situation without making assumptions about the underlying system. Based on Copula theory, the present application uses the KNN method to rank the influencing factors through a two-step method of non-parametric estimation of CE.

[0094] In some embodiments of the present application, the Copula entropy correlation analysis is used to sort the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors, including:

[0095] PM2.5 was compared with 8 meteorological factors and 5 air pollutants respectively by using the non-parametric estimation CE method to rank the influencing factors;

[0096] According to the correlation with PM2.5, the first four influencing factors were taken as the main influencing factors, and 8 influencing factors were obtained.

[0097] In some embodiments of the present application, the constructed AGA-LSTM prediction model is an AGA-LSTM prediction model constructed by integrating AGA and LSTM, which is used to predict PM2.5.

[0098] Among them, LSTM subtly controls the combination of short-term memory and long-term memory through the "gate structure", and solves the problem of gradient disappearance to a certain extent. Compared with recurrent neural networks, it has better results in the analysis of time series data. However, in the process of building the LSTM model, it is necessary to first assume some model parameters, such as the number of hidden layer neurons, the number of training times, the learning rate, etc. The selection of these parameters will also affect the prediction accuracy of the model. This application adopts a heuristic search algorithm-genetic algorithm to determine the model parameters by searching for the global optimal solution through the algorithm, thereby improving the prediction accuracy of the model.

[0099] Genetic algorithm is a kind of random search algorithm that draws on the natural selection and natural genetic mechanism of the biological world. It is very suitable for dealing with complex and nonlinear optimization problems that are difficult to solve by traditional search methods. The crossover probability and mutation probability of genetic algorithms are key parameters that affect the behavior and performance of genetic algorithms, and directly affect the convergence of the algorithm. Adaptive genetic algorithms (AGA) enable the crossover probability and mutation probability to change automatically with fitness. This application integrates AGA with LSTM to construct an AGA-LSTM prediction model to realize the prediction of PM2.5.

[0100] See also Figure 2 and Figure 6 As shown, the method for constructing the AGA-LSTM prediction model includes the following steps:

[0101] Step S101, binary encoding the number of hidden layer neurons, number of training times, and learning rate of the parameters in the LSTM;

[0102] Step S102: Generate an initial population N based on the binary encoding of the parameters in the LSTM, where N is an even number;

[0103] Step S103: Establish an LSTM model, train and predict the data of the training set and the test set, and use the predicted mean square error as the AGA fitness value f i ;

[0104] Step S104: select N individuals according to the roulette rule and calculate f avg and f max ;

[0105] Step S105: randomly pair the individuals in the group into pairs, forming a total of N / 2 pairs, and calculate the adaptive crossover probability of each pair of individuals according to the adaptive formula Randomly generate R(0,1), if R<P c , then the crossover operation is performed on the pair of chromosomes;

[0106] Step S106: For all individuals N in the population, calculate the adaptive mutation probability according to the adaptive mutation formula If R<P m , then the chromosome is crossover operated;

[0107] Step S107, calculating the fitness of new individuals generated by crossover and mutation, and the new individuals and the parent generation together constitute a new population;

[0108] Step S108, determine whether the termination condition is met, if so, terminate and return to the optimal parameters, otherwise execute step S104;

[0109] Step S109: construct an LSTM network model using the optimal parameters obtained by AGA; train the model to output the prediction results.

[0110] The parameters in the LSTM model are determined based on AGA, including the number of hidden layer neurons, the number of training times, and the learning rate in the LSTM model.

[0111] In an embodiment of the present application, when establishing an LSTM model, the mean square error predicted by the LSTM model is used as the AGA fitness value, and the parameters of the LSTM model are adjusted based on the AGA fitness value.

[0112] When the PM2.5 prediction method based on Copula entropy correlation analysis is used to perform correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants, during the experimental analysis and result verification stage, the experimental data of this application is taken from the environmental cloud of Nanjing Yunchuang Big Data Technology Co., Ltd. from January 1, 2016 to December 31, 2016, including Beijing's hourly meteorological data (weather conditions, temperature, perceived temperature, air pressure, humidity, rainfall, wind direction, wind force, wind speed) and hourly air quality monitoring data (PM2.5, PM10, NO2, SO2, O3, CO).

[0113] Among them, the air quality monitoring data refers to the hourly monitoring data of 12 monitoring points. In order to make the meteorological data consistent with the air quality monitoring data, the data of a certain monitoring point is uniformly taken, so there are 8760 data records for each. However, due to some uncontrollable factors, there are missing data, and this part of the data is deleted. Finally, 6430 data records are used. The Chinese descriptions of weather conditions, wind direction, and wind force are coded, as shown in Tables 1-3.

[0114] Table 1 Weather condition codes

[0115]

[0116]

[0117] Table 2 Wind direction coding

[0118]

[0119]

[0120] Table 3 Wind force code

[0121]

[0122] In the experimental result analysis stage, when conducting the influencing factor selection experiment, the hardware environment of this application experiment is Intel(R) Core(TM) i7-8565U CPU 1.80GHz, 8GB memory, operating system is Windows 10, and the programming tool of this experiment uses Python3.7.8.

[0123] In order to analyze the correlation between PM2.5 and meteorological factors and other air pollutants, and further select influencing factors, this experiment compares PM2.5 with 8 meteorological factors (air temperature, perceived temperature, air pressure, humidity, rainfall, wind direction, wind force, wind speed) and 5 air pollutants (PM10, NO2, SO2, O3, CO) respectively, and sorts the influencing factors through the non-parametric estimation CE method, and takes the top 4 with the largest correlation with PM2.5 as the main influencing factors, thus obtaining 8 influencing factors; again use the method of this application to analyze the correlation between the 8 main influencing factors and PM2.5, normalize the estimated CE value, and bring it into the prediction model as the weight of each main influencing factor to achieve pollutant prediction. The experimental results are as follows: Figure 3 and Figure 4 As shown. Figure 3 The horizontal axes 1-8 represent the air temperature (TMP), perceived temperature (FEELST), air pressure (PRES), relative humidity (HUM), precipitation (RAIN), wind direction (WDIR), wind force (WSC) and wind speed (WSPD), respectively. Figure 4 The horizontal axes 1-5 represent PM10, NO2, SO2, O3, and CO respectively.

[0124] Depend on Figure 3 It can be deduced that, based on the calculation of the CE results of various meteorological factors and PM2.5, the top four factors with strong correlation are PRES, TMP, HUM and WDIR. Figure 4 The CE results of each pollutant factor and PM2.5 were calculated, and the top four factors with the strongest correlation were PM10, NO2, SO2 and O3. It was also found that the correlation between PM10, NO2, SO2 and O3 and PM2.5 was stronger than that of the meteorological factors.

[0125] Therefore, eight influencing factors with strong correlation with PM2.5, namely PM10, NO2, SO2, O3, PRES, TMP, HUM, and WDIR, were selected; the weight ω of each influencing factor was calculated by calculating the CE results of the eight influencing factors and PM2.5 concentration. i . The 8 influencing factor attribute data x i Combined weight ω i , input into the model for prediction.

[0126] Based on the influencing factor selection experiment, 8 influencing factors were selected using the CE method. The new data set has 10 dimensions, namely time, PM10, NO2, SO2, O3, PRES, TMP, HUM, WDIR, and PM2.5. Therefore, the data set contains 64,300 data records in total. By drawing the time series feature diagrams of each feature at different times, the changing rules of each feature at different times are further analyzed. Figure 5As shown. Figure 5 It can be seen that each feature of the data set has a certain periodicity.

[0127] In the experiment of this application, formula (1) is used to normalize the input data, and the numerical values ​​are mapped to decimals between (0, 1). Finally, the data set is processed into the data format required for supervised learning and input into the prediction model.

[0128]

[0129] The format of each supervised learning data is: (x1ω1,x2ω2,x3ω3,…,x 10 ), where the first 9 dimensions are the training data at time t-1, and the last dimension is the label, i.e., the PM2.5 concentration at time t. The length of each input sequence is 24. To predict the PM2.5 concentration on the second day, the data from the previous 7 days are used as the model input data for prediction, and the PM2.5 concentration data on the first day after 7 days is used as the prediction result output of the model. Therefore, the input data of the prediction model is 7*24*10, and the output data is 24*1. The data is divided into two parts for experiments, with 80% as the training set and 20% as the test set. The three parameters of the number of hidden layer neurons, the number of training times, and the learning rate in the LSTM model are determined by AGA. The relevant parameter settings of the prediction model are shown in Table 4.

[0130] Table 4 Prediction model related parameter settings

[0131]

[0132]

[0133] In order to verify the feasibility and accuracy of the method proposed in this application, the experiment designed four prediction models for comparative tests, namely, all attributes directly input into the LSTM prediction model (LSTM), the CE-based LSTM prediction model (CE+LSTM), the method proposed in this application (CE+AGA-LSTM) and the CE-based RNN prediction model (CE+RNN).

[0134] Result analysis:

[0135] To evaluate the performance of the prediction model, five samples (five 7-day time series data) were randomly selected as input data to predict PM2.5 concentrations, and the mean absolute error (MAE) and root mean square error (RMSE) were used to measure the performance. The two indicators can be obtained by formulas (2) and (3). The experimental comparison results are shown in Tables 5 and 6.

[0136]

[0137]

[0138] in, and yi represent the predicted value and true value of PM2.5 concentration, respectively.

[0139] Table 5 Average MAE of 5 experiments

[0140]

[0141]

[0142] Table 6 Average RMSEE of 5 experiments

[0143]

[0144]

[0145] Table 5 shows the average MAE of the four models in 5 sample sets. Each column value represents the MAE of the PM2.5 concentration of the model in sample sets 1-5. The results show that the MAE of the method proposed in this application (CE+AGA-LSTM) is the smallest. At the same time, it is found that the MAE results of the LSTM prediction model (LSTM) directly inputting all attributes are comparable to the LSTM prediction model based on CE (CE+LSTM), while the MAE value of the RNN prediction model based on CE (CE+RNN) is the largest.

[0146] Therefore, the PM2.5 prediction method based on Copula entropy correlation analysis of the present application, the input data includes historical atmospheric pollutant data, historical meteorological data and meteorological data for the next 7 days.

[0147] Through the CE correlation analysis results between 8 meteorological factors and 5 air pollutants and PM2.5, we can obtain the characteristic analysis of air pollutants in the one and a half years from January 2017 to June 2018.

[0148] In addition, an embodiment of the present application also provides a PM2.5 prediction system based on Copula entropy correlation analysis.

[0149] Reference Figure 7 , Figure 7 This is a system block diagram of the PM2.5 prediction system based on Copula entropy correlation analysis of this application. In an embodiment of this application, the PM2.5 prediction system based on Copula entropy correlation analysis includes:

[0150] The data acquisition module 10 is used to acquire historical data of air pollutants and historical data of meteorological factors;

[0151] The correlation analysis module 20 is used to sort the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors based on Copula entropy correlation analysis, and obtain the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting;

[0152] The weight generation module 30 is used to normalize the correlation coefficient between the main influencing factors and PM2.5 by using the Copula entropy correlation analysis again, and bring it into the constructed AGA-LSTM prediction model as the weight of each main influencing factor;

[0153] The result prediction module 40 is used to determine the parameters in the LSTM model based on AGA, predict pollutants based on the LSTM model, and obtain PM2.5 prediction results.

[0154] Among them, each module in the above-mentioned PM2.5 prediction system based on Copula entropy correlation analysis corresponds to each step in the above-mentioned PM2.5 prediction method embodiment based on Copula entropy correlation analysis, and its functions and implementation processes will not be repeated here one by one.

[0155] The PM2.5 prediction method and device based on Copula entropy correlation analysis of the present application can be implemented in the form of a computer program, which can be run on a computer device.

[0156] The computer device includes a processor and a memory connected via a system bus, wherein the memory may include a non-volatile storage medium and an internal memory.

[0157] The processor is used to provide computing and control capabilities and support the operation of the entire computer equipment.

[0158] 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 PM2.5 prediction method based on Copula entropy correlation analysis.

[0159] 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.

[0160] Among them, the processor is used to run the computer program stored in the memory to implement various embodiments of the PM2.5 prediction method based on Copula entropy correlation analysis of the present application, which will not be repeated here.

[0161] In addition, an embodiment of the present application also provides a computer-readable storage medium.

[0162] The computer-readable storage medium of the present application stores a PM2.5 prediction program based on Copula entropy correlation analysis, wherein when the PM2.5 prediction program based on Copula entropy correlation analysis is executed by a processor, the steps of the PM2.5 prediction method based on Copula entropy correlation analysis as described above are implemented.

[0163] Among them, the method implemented when the PM2.5 prediction program based on Copula entropy correlation analysis is executed can refer to the various embodiments of the PM2.5 prediction method based on Copula entropy correlation analysis of the present application, and will not be repeated here.

[0164] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article or system including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or system. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or system including the element.

[0165] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0166] The present application can be used in many general or special computer system environments or configurations. For example: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0167] The present application provides a PM2.5 prediction method and system based on Copula entropy correlation analysis, which uses CE to perform correlation analysis on multiple meteorological factors and multiple atmospheric pollutants, and sorts the correlation of the influencing factors according to the size of the correlation coefficient, so as to determine the contribution rate of meteorological factors to the pollutant index. The top N factors with the largest correlation with PM2.5 are taken as the main influencing factors; at this time, CE is used again to analyze the correlation between multiple main influencing factors and PM2.5, and the correlation coefficient is normalized as the weight of each main influencing factor, which is brought into the prediction model to realize pollutant prediction. The present application uses CE to realize the correlation analysis between PM2.5 and meteorological factors and atmospheric pollutants.

[0168] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A PM2.5 prediction method based on Copula entropy correlation analysis, characterized in that: The PM2.5 prediction method comprises the following steps: Obtain historical data on air pollutants and meteorological factors; Based on the Copula entropy correlation analysis, the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors are sorted to obtain the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting; The Copula entropy correlation analysis is used again to normalize the correlation coefficient between the main influencing factors and PM2.5, and the correlation coefficient is used as the weight of each main influencing factor and brought into the constructed AGA-LSTM prediction model; Determine parameters in the LSTM model based on AGA, predict pollutants based on the LSTM model, and obtain PM2.5 prediction results; Wherein, the Copula entropy is based on the Copula theory, in which the multivariate probability density function represents the product of the number of edges and the Copula density function, and the Copula density function represents the dependency structure between random variables. The Copula entropy is a non-parametric method for estimating CE or MI from data, and the non-parametric method includes: estimating the empirical Copula density function and estimating the Copula entropy; The constructed AGA-LSTM prediction model is constructed by integrating AGA and LSTM to predict PM2.

5. The construction method of the AGA-LSTM prediction model includes the following steps: Step S101, binary encoding the number of hidden layer neurons, number of training times, and learning rate of the parameters in the LSTM; Step S102: Generate an initial population N based on the binary encoding of the parameters in the LSTM, where N is an even number; Step S103: Establish an LSTM model, train and predict the data of the training set and the test set, and use the predicted mean square error as the AGA fitness value f i ; Step S104: select N individuals according to the roulette rule and calculate f avg and f max ; Step S105: randomly pair the individuals in the group into pairs, forming a total of N / 2 pairs, and calculate the adaptive crossover probability of each pair of individuals according to the adaptive formula Randomly generate R(0,1), if R<P c , then the crossover operation is performed on the pair of chromosomes; Step S106: For all individuals N in the population, calculate the adaptive mutation probability according to the adaptive mutation formula If R<P m , then the chromosome is crossover operated; Step S107, calculating the fitness of new individuals generated by crossover and mutation, and the new individuals and the parent generation together constitute a new population; Step S108, determine whether the termination condition is met, if so, terminate and return to the optimal parameters, otherwise execute step S104; Step S109: construct an LSTM network model using the optimal parameters obtained by AGA; train the model to output the prediction results.

2. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 1 is characterized in that: The historical data of atmospheric pollutants include five kinds of atmospheric pollutants, namely PM10, NO2, SO2, O3, and CO; the historical data of meteorological factors include eight kinds of meteorological factors, namely air temperature, perceived temperature, air pressure, humidity, rainfall, wind direction, wind force, and wind speed.

3. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 2 is characterized in that: The MI is the mutual information, which is used to reflect the size of the shared information between different variables. The method for calculating the mutual information between two random variables X and Y is: Calculate the mutual information between random variables X and Y according to the formula: Among them, x i and i are samples of random variables X and Y respectively, and N is the sample size.

4. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 3 is characterized in that: Estimate the empirical Copula density function, including: Given a random variable X = {x1,L,x N } T The generated independent and identically distributed data samples {X1,L,X T }; The estimated empirical Copula density function is: Where χ represents the indicator function when i = 1, L, N; assuming u = [F1, L, F N ] derive a new sample set {u1,L,u T }Data c(u) as the empirical Copula density function.

5. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 4 is characterized in that: The Copula entropy correlation analysis is based on sorting the influencing factors in the historical data of air pollutants and the historical data of meteorological factors, including: PM2.5 was compared with 8 meteorological factors and 5 air pollutants respectively by using the non-parametric estimation CE method to rank the influencing factors; According to the correlation with PM2.5, the first four influencing factors were taken as the main influencing factors, and 8 influencing factors were obtained.

6. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 5, characterized in that: The parameters in the LSTM model are determined based on AGA, including the number of hidden layer neurons, the number of training times, and the learning rate in the LSTM model.

7. The PM2.5 prediction method based on Copula entropy correlation analysis according to claim 5, characterized in that: When establishing the LSTM model, the mean square error predicted by the LSTM model is used as the AGA fitness value, and the parameters of the LSTM model are adjusted based on the AGA fitness value.

8. A PM2.5 prediction system based on Copula entropy correlation analysis, characterized in that: The PM2.5 prediction system is used to perform the steps of the PM2.5 prediction method based on Copula entropy correlation analysis according to any one of claims 1 to 7, and the PM2.5 prediction system comprises: A data acquisition module is used to acquire historical data of atmospheric pollutants and meteorological factors; A correlation analysis module, for sorting the influencing factors in the historical data of atmospheric pollutants and the historical data of meteorological factors based on Copula entropy correlation analysis, and obtaining the influencing factors of the main influencing factors with greater correlation with PM2.5 after sorting; A weight generation module is used to normalize the correlation coefficient between the main influencing factors and PM2.5 by using Copula entropy correlation analysis again, and bring it into the constructed AGA-LSTM prediction model as the weight of each main influencing factor; The result prediction module is used to determine the parameters in the LSTM model based on AGA, predict pollutants based on the LSTM model, and obtain PM2.5 prediction results.

Images