Search

CN-116050641-B - Port TSP concentration short-time prediction method, device and storage medium

CN116050641BCN 116050641 BCN116050641 BCN 116050641BCN-116050641-B

Abstract

The invention discloses a short-time prediction method, a device and a storage medium for port TSP concentration, wherein the short-time prediction method comprises the steps of acquiring historical monitoring data of the TSP concentration and cleaning the concentration data; the method comprises the steps of dividing the washed TSP concentration data into N groups of time series data, wherein the former N-1 groups of time series data have periodic variation characteristics, the nth group of time series data have random fluctuation characteristics, respectively constructing N-1 periodic time series prediction models and 1 random time series prediction model aiming at the N groups of time series data, and finally summarizing the prediction results of the N time series prediction models to obtain the predicted port TSP concentration. The method can solve the problem of prediction accuracy reduction caused by non-stationary and random characteristics of TSP concentration data displayed along with time change, improves the reliability of port TSP concentration short-time prediction results, and has important significance for port atmospheric pollution prevention and control.

Inventors

  • SHEN JINXING
  • LIU QINXIN
  • LIU MENGMENG
  • GU LE
  • FENG XUEJUN

Assignees

  • 河海大学

Dates

Publication Date
20260505
Application Date
20230214

Claims (9)

  1. 1. A harbor TSP concentration forecasting method considering the influence of dust suppression measures, characterized by comprising: Acquiring harbour presence TSP concentration data set monitored over an hour And weather factor data set under influence of dust suppression measures ; For TSP concentration data set And meteorological factor data set Cleaning, removing abnormal value, and adjusting data time interval to obtain TSP concentration data set for analysis And meteorological factor data set ; Constructing a characteristic meteorological factor screening model and analyzing a TSP concentration data set And meteorological factor data set Screening out characteristic weather factor set capable of influencing TSP concentration ; Building a multi-input single-output TSP concentration prediction model according to a TSP concentration data set Characteristic meteorological factor dataset Obtaining the concentration forecast result of port TSP ; The method for constructing the TSP concentration forecasting model with multiple inputs and single outputs comprises the following steps: data conversion of the washed TSP concentration data set And characteristic meteorological factor data set The conversion to a series of lag time is As training samples and output labels: First layer data convolution for the first layer Training samples Setting input data for a first layer convolution Is that Setting the total number of the first layer neurons as Let the first The weight matrix of each neuron is The deviation matrix is Then (1) Output values of individual neurons The method comprises the following steps: Wherein, the Is the sum of Hadamard products of the elements of the two matrices; Input data of the first layer Data result output after convolution The method comprises the following steps: Second layer data recursion for the first layer Training samples Output data after first layer convolution Input data recursive as a second layer data I.e. Setting the number of neurons of the second layer as Input data Through the first The output data of the data recursion after the status update of each neuron is that The final second layer data recursively output results are: In the formula, As a hyperbolic tangent function; is a sigmoid growth curve function; 、 、 And Respectively setting state weight values; 、 、 And Respectively setting the deviation matrix values; Is the first Training samples Output data obtained after data recursion, when Time of day Is a unit matrix; third layer weight optimization of outputting data of the second layer Input data optimized as third tier weights I.e. Setting input data And outputting the label Is a correlation score of (2) The method comprises the following steps: In the formula, Is a weight matrix; Is a deviation matrix; relevance score for each input data By normalizing an exponential function Calculating to obtain input data Weight index of (2) The method comprises the following steps: Based on input data And weight index Weighted average summation is carried out to obtain the final TSP concentration forecast result : 。
  2. 2. The method for forecasting the TSP concentration of a port taking account of the influence of dust suppression measures as recited in claim 1, wherein a characteristic weather factor characteristic screening model is constructed to analyze a TSP concentration dataset And meteorological factor data set Screening out characteristic weather factors which can influence TSP concentration In the step, characteristic meteorological factors are constructed The method for screening the model comprises the following steps: calculating linear correlation coefficient according to the washed TSP concentration data set And (3) with Individual meteorological monitoring data sets Calculating TSP concentration data set And the first Monitoring data set of individual meteorological factors Linear correlation coefficient between : Wherein x t is the time of day Monitored TSP concentration, f t,k is time of day Monitored first Data of individual meteorological factors; respectively calculating TSP concentration data and all Linear correlation coefficients between meteorological factors to obtain vector ; Setting a linear correlation determination threshold If (3) Then consider the first The weather factor has a remarkable linear correlation with the change of the TSP concentration, thereby obtaining the product meeting the condition G is less than or equal to K, which is a characteristic meteorological factor; nonlinear correlation coefficient calculation based on the washed TSP concentration data set And a weather monitoring dataset Calculating TSP concentration data And the first Monitoring data of individual meteorological factors Nonlinear correlation coefficient between : In the formula, For the moment of time Monitored TSP concentration And meteorological factors Is (are) monitored data Is smaller of (a); For a certain moment Monitored TSP concentration And meteorological factors Is (are) monitored data Is a joint probability of (2); For TSP concentration data Is used for the edge probability of (1), Is a meteorological factor Is (are) monitored data Edge probability of (2); respectively calculating TSP concentration data and all Nonlinear correlation coefficients among meteorological factors to obtain vectors ; Setting a nonlinear correlation determination threshold If (3) Then consider the first The weather factor has significant nonlinear correlation with the change of TSP concentration, thereby obtaining the condition E is less than or equal to K, which is a characteristic meteorological factor; characteristic weather factor screening, to be linearly related Individual characteristic weather factors and non-linearities The individual characteristic meteorological factors are combined into The following characteristic meteorological factors: screening index calculation formula according to characteristic meteorological factors Determining meteorological factor screening index ; The calculation formula of (2) is as follows: In the formula, Is a control coefficient; in order for the calculation formula A minimum value screening index; If it is Removing the weather factor without significant influence on the change in TSP concentration, otherwise, considering the factor as a characteristic weather factor, and passing Determination of the final influencing TSP concentration The characteristic weather factors are Wherein 。
  3. 3. The port TSP concentration forecast method considering the influence of dust suppression measures as recited in claim 2, wherein at a certain point in time Monitored TSP concentration And meteorological factors Is (are) monitored data Joint probabilities of (a) TSP concentration data Edge probability of (a) Weather factors Is (are) monitored data Edge probability of (a) Is determined by the concentration of TSP And meteorological factors Is of the detection data of (a) In the distribution range of Cartesian coordinate axes, the Cartesian coordinate axes are divided into The number of the grids is one, Is that And While occupying the number of grids divided by the total number of grids, Is that The number of occupied cells divided by the total number of cells, Is that The number of occupied cells divided by the total number of cells.
  4. 4. The port TSP concentration forecast method considering influence of dust suppression measures as recited in claim 1, wherein the port TSP concentration forecast method includes the steps of cleaning Individual TSP concentration data And characteristic meteorological factors The conversion to a series of lag time is The conversion mode of taking the two-dimensional matrix of the (2) as a training sample and outputting a label is as follows: In the formula, Is the first Training samples; Is the first Outputting labels; =1,2,...,T-p。
  5. 5. the method for forecasting the concentration of port TSP taking into account the influence of dust suppression measures as recited in any one of claims 1 to 4, wherein the obtained historical monitoring data of the concentration of port TSP and weather factors under the influence of dust suppression measures comprises: Acquiring port TSP concentration data, namely acquiring the port TSP concentration data in a time range according to a port TSP online monitoring system Port monitoring obtained internally The individual TSP concentration data form a data set with time series characteristics ; Port meteorological factor data acquisition in time range Can be monitored internally Data set with time sequence characteristics formed by individual meteorological factors The method comprises the following steps: Wherein f m,k is the first First of all meteorological factors The number of data to be monitored is determined, , 。
  6. 6. The method for forecasting the TSP concentration in a port taking into account the influence of dust suppression measures as recited in claim 5, wherein the method for cleaning the acquired historical monitoring data of the TSP concentration and weather factors is as follows: abnormal data clearing if at Data set of individual TSP concentrations In (1) Monitoring data at various time points Satisfy the following requirements Or (b) Will then As a result of the clearing of the abnormal data, To (3) pair In the meteorological monitoring data, the first Of meteorological factors Individual monitoring data In (1) Monitoring data at various time points Satisfy the following requirements Or (b) Will then As a result of the clearing of the abnormal data, Wherein , , , The method comprises the following steps: Time interval adjustment, namely adjusting the acquired TSP concentration data set and the meteorological factor data set at any moment Concentration data of (a) for time of day And Between (a) and (b) Arithmetic average is carried out on the TSP concentration data to obtain time TSP concentration of (2) For the time of day And Between (a) and (b) First one Arithmetically averaging the monitoring data of the seed meteorological factors to obtain Time of day (time) The monitoring data of the seed meteorological factors are ; Data cleaning results of harbour Individual TSP concentration data After washing, a time series characteristic is obtained at intervals of 1 hour Data set of individual TSP concentrations ; Monitoring data set of individual meteorological factors Cleaning to obtain Monitoring data set of individual meteorological factors 。
  7. 7. The port TSP concentration forecasting method considering the influence of dust suppression measures as claimed in claim 1, wherein the weight matrix and the deviation matrix of the three-layer multi-input single-output forecasting model are constructed according to the actually measured TSP concentration data And characteristic meteorological factor data And determining optimal parameters by an iterative updating method.
  8. 8. A harbor TSP concentration forecasting device considering the influence of dust suppression measures, characterized by comprising a processor and a memory, wherein the memory stores a program or instructions that are loaded and executed by the processor to implement the steps of the harbor TSP concentration forecasting method under the influence of dust suppression measures as claimed in any one of claims 1 to 7.
  9. 9. A computer readable storage medium, wherein a program or instructions is stored on the readable storage medium, which when executed by a processor, implements the steps of the harbor TSP concentration forecast method under the influence of dust suppression measures as claimed in any one of claims 1 to 7.

Description

Port TSP concentration short-time prediction method, device and storage medium Technical Field The invention belongs to the technical field of atmospheric pollution control in the field of environmental engineering, and particularly relates to a calculation method for port TSP concentration prediction. Background Ports are key nodes of the global logistics supply chain, supporting regional commerce and high-speed development of economies. However, a great amount of particulate matters generated in the process of loading, unloading and piling dry bulk cargoes such as coal, ore and the like in a port can have a significant negative effect on the air quality in the surrounding area, and port workers exposed to the high-concentration particulate matter environment for a long time can increase the probability of suffering from respiratory diseases and the risk of suffering from heart diseases. Investigation of various particulate matter concentrations in ambient air in port surrounding areas has shown that the concentration value of total suspended particulate matter (TSP) is generally out of standard. Particularly, partial old harbor areas contribute more than 50% of the total TSP pollution amount of the city, and the serious pollution problem is formed, so that the method has become one of the main challenges for restricting the green sustainable development of the harbor. Therefore, according to the TSP monitoring system laid in the port, the time sequence of the TSP concentration is acquired, the distribution condition of the port atmospheric pollutant concentration is accurately and efficiently predicted, and the targeted treatment measure of the port TSP pollution is determined, so that the method is one of the primary tasks of building the intelligent green world first-class port. Earlier researches of the inventor find that the time sequence change of the port TSP concentration has complex nonlinear chaos characteristics, and it is difficult to accurately predict future short-time concentration data directly through a common time sequence prediction model (such as an ARIMA model). Therefore, the method can be used for constructing time sequence non-stable fluctuation information of the port TSP concentration data, decomposing the TSP concentration into periodic stable data and random non-stable data, and respectively constructing a prediction model, so that the accuracy and reliability of port TSP concentration short-time prediction can be remarkably improved. Disclosure of Invention The method aims to solve the technical problems that the prediction accuracy is reduced due to the non-stable and random characteristics of TSP concentration data which are displayed along with the change of time, so that the reliability of port TSP concentration short-time prediction results is improved, port atmospheric pollution control work is supported, and intelligent green sustainable development of ports is realized. In order to solve the technical problems, the invention adopts the following technical scheme: the invention firstly provides a short-time prediction method for port TSP concentration, which comprises the following steps: Acquiring historical monitoring data of port TSP concentration; Decomposing the acquired historical monitoring data to obtain N groups of time series data, wherein the former N-1 groups of time series data F i (t) have periodic variation characteristics, and the nth group of time series data R N (t) have random fluctuation characteristics; n short-time prediction models are respectively constructed for N groups of time series data, wherein N-1 periodic time series prediction models P iF (t) are constructed by using the time series data F i (t) of the previous N-1 groups of periodic variation characteristics, and 1 random time series prediction model is constructed by using the time series data R N (t) of the nth group of random fluctuation characteristics Summarizing the predicted results of the N-1 periodic time sequence predicted models and the predicted results of the 1 random time sequence predicted models to finally obtain the predicted concentration of the port TSP. The invention also provides a short-time predicting device for the port TSP concentration, which comprises a processor and a memory, wherein the memory stores a program or an instruction, and the program or the instruction is loaded and executed by the processor to realize the steps of the short-time predicting method for the port TSP concentration. The present invention also provides a computer readable storage medium having stored thereon a program or instructions which when executed by a processor performs the step of short-term prediction of the TSP concentration of the port. The technical scheme adopted by the invention has the following beneficial effects: (1) Compared with the traditional scheme, the method is constructed when the TSP concentration of the port is analyzed, the TSP concentration of the time serie