CN107943928B - A kind of ozone concentration prediction technique and system based on space-time data statistical learning - Google Patents
A kind of ozone concentration prediction technique and system based on space-time data statistical learning Download PDFInfo
- Publication number
- CN107943928B CN107943928B CN201711168594.7A CN201711168594A CN107943928B CN 107943928 B CN107943928 B CN 107943928B CN 201711168594 A CN201711168594 A CN 201711168594A CN 107943928 B CN107943928 B CN 107943928B
- Authority
- CN
- China
- Prior art keywords
- environmental monitoring
- historical
- ozone concentration
- monitoring website
- measured
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- CBENFWSGALASAD-UHFFFAOYSA-N Ozone Chemical compound [O-][O+]=O CBENFWSGALASAD-UHFFFAOYSA-N 0.000 title claims abstract description 121
- 238000000034 method Methods 0.000 title claims abstract description 53
- 238000012544 monitoring process Methods 0.000 claims abstract description 162
- 230000007613 environmental effect Effects 0.000 claims abstract description 133
- 238000012549 training Methods 0.000 claims abstract description 42
- 238000012360 testing method Methods 0.000 claims abstract description 33
- 238000012795 verification Methods 0.000 claims abstract description 26
- 238000002790 cross-validation Methods 0.000 claims abstract description 14
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 11
- 239000003344 environmental pollutant Substances 0.000 claims description 28
- 231100000719 pollutant Toxicity 0.000 claims description 28
- MWUXSHHQAYIFBG-UHFFFAOYSA-N Nitric oxide Chemical compound O=[N] MWUXSHHQAYIFBG-UHFFFAOYSA-N 0.000 claims description 26
- 238000005259 measurement Methods 0.000 claims description 15
- 239000006185 dispersion Substances 0.000 claims description 14
- 230000001235 sensitizing effect Effects 0.000 claims description 8
- 230000002123 temporal effect Effects 0.000 claims description 6
- 235000013399 edible fruits Nutrition 0.000 claims description 2
- 230000035945 sensitivity Effects 0.000 claims description 2
- 238000013179 statistical model Methods 0.000 abstract description 18
- 238000004458 analytical method Methods 0.000 abstract description 9
- 238000000605 extraction Methods 0.000 abstract description 6
- 239000003570 air Substances 0.000 description 19
- 238000003066 decision tree Methods 0.000 description 7
- 230000007246 mechanism Effects 0.000 description 7
- 238000011156 evaluation Methods 0.000 description 6
- 230000008859 change Effects 0.000 description 5
- 238000009792 diffusion process Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 241001269238 Data Species 0.000 description 3
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 3
- 239000007789 gas Substances 0.000 description 3
- 239000001301 oxygen Substances 0.000 description 3
- 229910052760 oxygen Inorganic materials 0.000 description 3
- 238000005192 partition Methods 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- GQPLMRYTRLFLPF-UHFFFAOYSA-N Nitrous Oxide Chemical compound [O-][N+]#N GQPLMRYTRLFLPF-UHFFFAOYSA-N 0.000 description 2
- 239000000809 air pollutant Substances 0.000 description 2
- 231100001243 air pollutant Toxicity 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 238000001311 chemical methods and process Methods 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 230000008520 organization Effects 0.000 description 2
- 238000004393 prognosis Methods 0.000 description 2
- 238000000611 regression analysis Methods 0.000 description 2
- 241000208340 Araliaceae Species 0.000 description 1
- UGFAIRIUMAVXCW-UHFFFAOYSA-N Carbon monoxide Chemical compound [O+]#[C-] UGFAIRIUMAVXCW-UHFFFAOYSA-N 0.000 description 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 1
- 235000003140 Panax quinquefolius Nutrition 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 239000012080 ambient air Substances 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 229910002091 carbon monoxide Inorganic materials 0.000 description 1
- 238000010219 correlation analysis Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 235000008434 ginseng Nutrition 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 239000008188 pellet Substances 0.000 description 1
- 239000002243 precursor Substances 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000003672 processing method Methods 0.000 description 1
- 238000005086 pumping Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000005096 rolling process Methods 0.000 description 1
- 238000010187 selection method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/20—Information retrieval; Database structures therefor; File system structures therefor of structured data, e.g. relational data
- G06F16/24—Querying
- G06F16/245—Query processing
- G06F16/2458—Special types of queries, e.g. statistical queries, fuzzy queries or distributed queries
- G06F16/2462—Approximate or statistical queries
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Probability & Statistics with Applications (AREA)
- Marketing (AREA)
- Development Economics (AREA)
- General Business, Economics & Management (AREA)
- Fuzzy Systems (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Educational Administration (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Engineering & Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides a kind of ozone concentration prediction technique and system based on space-time data statistical learning, including：Obtain corresponding fixed reference feature of each environmental monitoring website each historical juncture and reference result；Obtain training verification collection and test set；Using statistical learning algorithm, model parameter selection and training are carried out by ten folding cross validations on training verification collection, obtains prediction model, and tested on test set；According to environmental monitoring website current time to be measured corresponding target signature, prediction model and environmental monitoring website current time to be measured corresponding ozone concentration, the ozone concentration at environmental monitoring website current predictive moment to be measured is obtained.The present invention passes through space-time data and statistical learning, by ozone concentration, meteorological condition and point source emission introduced feature extraction process, precisely portray the raising reason of ozone concentration, incremental analysis is used simultaneously, compared to statistical model before, accuracy has significant increase, and versatile, is suitable for different environmental monitoring websites.
Description
Technical field
The present invention relates to computer data analysis technical fields, and space-time data statistics is based on more particularly, to one kind
The ozone concentration prediction technique and system of habit.
Background technology
Currently, with the development of modern society, industrialization and urbanization is constantly deepened, and air quality problems can not more neglect
Depending on.It, all can be right when the concentration of the atmosphere pollutions such as pellet, nitrogen oxides, carbon monoxide, ozone is more than a certain range
The health of the mankind generates harm.And the concentration of atmosphere pollution is usually influenced by factors, the shiftings such as vehicle, ship, aircraft
The exhaust emissions in dynamic source can impact the concentration of atmosphere pollution, and the point sources such as industrial gas emission are to the dense of atmosphere pollution
Degree impacts, and the factors such as point source, the meteorology on monitoring station periphery, landform can also play the generation of pollutant, diffusion and accumulation
Different role.
Therefore, how by these data carry out integrate and comprehensive analysis, accurately forecast certain time within the scope of air
Pollutant concentration even takes adjustment or limitation point source, moving source row accordingly to carry out giving warning in advance for high density pollution
The measure put is extremely important and significant.
The method of existing pollutant or Air Quality Forecast is broadly divided into two classes：Prediction based on numerical model
With the prediction based on statistical model.
Pollutant prediction based on numerical model is existing main stream approach, and such method mainly utilizes pollution
Gentle two class of the image data input of source data is discharged, by the simulation to physics, chemical process, carries out the pre- of pollutant
It surveys.Although such method is more accurate to the simulation of physics, chemical process, in its input data, emission inventory and meteorology
Mode data may have certain error.Especially emission inventory, mostly according to the whole emissions data of provinces and cities and geographical position
The discharge capacity analysis estimation that distribution carries out each region is set, this estimation relies on experience and carries out mostly, and accuracy is extremely low.Input data
Inaccuracy, cause the prediction result of numerical model usually to greatly differ from each other with true monitor value.In addition, the calculating of numerical model
Generally require a large amount of computing resource and longer calculating time, it is difficult to the event procedure of short-term burst carry out simulation and in time
Forecast.
It is not allowed for inputs such as the above emission inventory, meteorologies and is difficult to cope at 2 points that emergency case is forecast in time not
Foot, the Predict Model of Air Pollutant Density based on statistical model come into being.
Prediction based on statistical model is mainly by the multi-sources space-time such as air quality monitoring data, weather data
Data carry out comprehensive analysis, a model are fitted using the correlation technique of machine learning, to be predicted.Statistical model
Advantage is mainly manifested in following two points：First, mode input is truthful data, the inaccurate possibility of mode input data drops significantly
It is low；Second is that the speed of statistical model output result is quickly, being easy should be to emergency case and progress rolling forecast.
However existing the ozone concentration prediction technique based on statistical model most of only consideration monitoring data and meteorological data
Regression analysis is carried out, does not consider that the factor for causing pollutant to change, prediction result are also and not fully up to expectations.
Invention content
The present invention provides a kind of one kind for overcoming the above problem or solving the above problems at least partly and is based on space-time number
The ozone concentration prediction technique and system learnt according to statistics.
According to an aspect of the present invention, a kind of ozone concentration prediction technique is provided, this method includes：S1, acquisition are each
Environmental monitoring website each historical juncture corresponding fixed reference feature and each environmental monitoring website each historical juncture correspond to
Reference result, for any historical juncture of any environment monitoring station, any historical juncture corresponding fixed reference feature
It is pre- including any historical juncture corresponding ozone concentration, any historical juncture corresponding point source discharge characteristics and history
Survey the meteorological condition feature at moment, any historical juncture, corresponding reference result was that the historical forecast moment is corresponding smelly
The difference of oxygen concentration and corresponding ozone concentration of any historical juncture, the historical forecast moment is in any historical juncture
Later；S2, training verification collection and test set are obtained, when each by each environmental monitoring website history of the training verification collection
It is default according to first to carve corresponding fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture
Proportioning is constituted, and the test set is by each environmental monitoring website each historical juncture corresponding fixed reference feature and described each
Environmental monitoring website each historical juncture corresponding reference result is constituted according to the second default proportioning；S3, it is calculated using statistical learning
Method carries out model parameter selection and model training by ten folding cross validations on the training verification collection, obtains prediction model,
Model measurement is carried out to the prediction model by the test set, the prediction model is carried out using the result of model measurement
Comparison and evaluation；S4, according to environmental monitoring website current time to be measured corresponding target signature and the prediction model, obtain institute
Environmental monitoring website current time to be measured corresponding objective result is stated, is corresponded to according to the environmental monitoring website current time to be measured
Objective result and the environmental monitoring website current time to be measured corresponding ozone concentration, obtain the environmental monitoring station to be measured
The ozone concentration at point current predictive moment, the target signature includes the current time corresponding ozone concentration, described current
The meteorological condition feature of moment corresponding point source discharge characteristics and the environmental monitoring website current predictive moment to be measured, the mesh
Mark the difference that result is the current predictive moment corresponding ozone concentration and the current time corresponding ozone concentration.
Preferably, corresponding point source discharge characteristics of any historical juncture of any environment monitoring station described in step S1 is logical
Cross following steps acquisition：S11, obtain sensitizing range, the sensitizing range include with where any environment monitoring station
All cities that city borders on；S12, the sensitizing range is divided into several identical rectangles, obtains each rectangle
Corresponding source emission subcharacter obtains the point source discharge characteristics according to the point source emission subcharacter of each rectangle.
Preferably, corresponding source emission subcharacter of any rectangle includes the corresponding mechanistic features of several historical periods,
Mechanistic features corresponding for any historical period, if point source number is 0 in any rectangle, by any historical period
Corresponding mechanistic features are set as 0, and the mechanistic features include Pollutant Dispersion Law feature and dispersion of pollutants temporal characteristics,
One historical period is a period of time before the historical juncture.
Preferably, if point source number is more than 0 in any rectangle, and in the first future time period in any rectangle
The relative direction of wind direction and the environmental monitoring website to be measured not within a preset range, by the corresponding machine of any historical period
Reason feature is set as 0, and first future time period is a period of time after any historical period.
Preferably, if point source number is more than 0 in any rectangle, and in the first future time period in any rectangle
The relative direction of wind direction and the environmental monitoring website to be measured calculates any historical period and corresponds in the preset range
Mechanistic features.
Preferably, according to the mean wind speed of sensitive wind in first future time period in any rectangle and described
The summation of all point sources nitrogen oxide emission in any historical time section in any rectangle obtains the pollutant and expands
Law characteristic is dissipated, the wind direction of the sensitivity wind is with the relative direction of the environmental monitoring website to be measured in the preset range.
Preferably, according to any rectangle at a distance from the environmental monitoring website to be measured, institute in any rectangle
State the mean wind speed of sensitive wind described in the first future time period, any historical period and between any historical juncture
Time difference obtains the dispersion of pollutants temporal characteristics.
Preferably, any rectangle is obtained at a distance from the environmental monitoring website to be measured by following steps：It obtains
The longitude and latitude of any rectangular centre point longitude and latitude and the environmental monitoring website to be measured；Pass through Gauss Kru&4&ger projection's public affairs
The longitude and latitude of any rectangular centre point is converted to the first rectangular co-ordinate in rectangular coordinate system, passes through Gauss-Ke Lv by formula
The longitude and latitude of the environmental monitoring website to be measured is converted to the second right angle in the rectangular coordinate system and sat by lattice projection formula
Mark；According to first rectangular co-ordinate and second rectangular co-ordinate, any rectangle and the environmental monitoring to be measured are obtained
The distance of website.
Preferably, the meteorological condition feature at any historical forecast moment of environmental monitoring website described in step S1 is specifically wrapped
It includes：Humidity average value, humidity maximum value, humidity minimum value, wind speed average value, the wind speed of three periods of the environmental monitoring website
Maximum value and wind speed minimum value, three periods be the historical forecast moment before 4 hours to the historical forecast moment,
Before the historical forecast moment before 6 hours to the historical forecast moment, the historical forecast moment 24 hours to described
The historical forecast moment.
According to another aspect of the present invention, a kind of ozone concentration forecasting system is provided, which includes：Historical data mould
Block, it is every for obtaining corresponding fixed reference feature of each environmental monitoring website each historical juncture and each environmental monitoring website
One historical juncture corresponding reference result, for any historical juncture of any environment monitoring station, any historical juncture
Corresponding fixed reference feature includes corresponding ozone concentration of any historical juncture, any historical juncture corresponding point source row
Put feature and the meteorological condition feature at historical forecast moment, any historical juncture, corresponding reference result was that the history is pre-
The difference of moment corresponding ozone concentration and corresponding ozone concentration of any historical juncture is surveyed, the historical forecast moment is in institute
After stating any historical juncture；Module is matched, for obtaining training verification collection and test set, the training verification collection is by described every
One environmental monitoring website each historical juncture corresponding fixed reference feature and each environmental monitoring website each historical juncture pair
The reference result answered is constituted according to the first default proportioning, and the test set is by each environmental monitoring website each historical juncture
Corresponding fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture are according to the second pre- establishing
Than constituting；Training test module, for utilize statistical learning algorithm, it is described training verification collection on by ten folding cross validations into
Row model parameter selects and model training, obtains prediction model, and model survey is carried out to the prediction model by the test set
Examination, is compared and is evaluated to the prediction model using the result of model measurement；Prediction module, for being supervised according to environment to be measured
Survey station point current time corresponding target signature and the prediction model obtain the environmental monitoring website current time pair to be measured
The objective result answered, according to the environmental monitoring website current time corresponding objective result to be measured and the environmental monitoring to be measured
Website current time corresponding ozone concentration obtains the ozone concentration at the environmental monitoring website current predictive moment to be measured, institute
It includes the current time corresponding ozone concentration, the current time corresponding point source discharge characteristics and described to state target signature
The meteorological condition feature at environmental monitoring website current predictive moment to be measured, the objective result correspond to for the current predictive moment
Ozone concentration and the current time corresponding ozone concentration difference.
The present invention proposes a kind of ozone concentration prediction technique and system based on space-time data statistical learning, passes through space-time number
According to and statistical learning, ozone concentration feature, meteorological condition feature and point source discharge characteristics are introduced into statistical model feature extraction
In the process, the raising reason of ozone concentration is more accurately portrayed, and uses incremental analysis, capturing point source etc. influences ozone concentration
Raised variable quantity weakens the invariants such as metastable natural source, the statistical model before comparing, and accuracy, which has, greatly to be carried
It rises.The present invention has very strong versatility, can be adapted for different environmental monitoring websites；The present invention can also capture different rings simultaneously
The otherness of border monitoring station carries out personalized prediction model to different environmental monitoring station's points and trains.
Description of the drawings
Fig. 1 is a kind of flow chart of the ozone concentration prediction technique based on space-time data statistical learning of the embodiment of the present invention；
Fig. 2 is a kind of stream of the ozone concentration prediction technique based on space-time data statistical learning of one embodiment of the present invention
Cheng Tu.
Specific implementation mode
With reference to the accompanying drawings and examples, the specific implementation mode of the present invention is described in further detail.Implement below
Example is not limited to the scope of the present invention for illustrating the present invention.
Only consider that monitoring data are gentle since the ozone concentration prediction technique in the prior art based on statistical model is most of
Image data carries out regression analysis, does not consider the factor such as point source for causing pollutant to change, mobile source emission etc.,
The Related Mechanism in atmosphere pollution generating process is not accounted for, prediction result is also and not fully up to expectations.
Based on this, the present invention proposes a kind of statistical model of mechanism driving, and gives a kind of based on space-time data system
Count the ozone concentration prediction technique of study.The particularity of ozone concentration prediction is, generates and evanishment is mainly in local
The better place of generation, the mainly progress in the form of its precursor such as nitrogen oxides when pollutant is spread, and air quality more holds
It is exceeded that easily there is a situation where ozone concentrations.Therefore it is dense ozone to be carried out by the multisource spatio-temporal data statistical learning method that mechanism drives
Degree prediction is necessary.
Fig. 1 is a kind of flow chart of the ozone concentration prediction technique based on space-time data statistical learning of the embodiment of the present invention,
This method includes：S1, corresponding fixed reference feature of each environmental monitoring website each historical juncture and each environment prison are obtained
Survey station point each historical juncture corresponding reference result, it is described any for any historical juncture of any environment monitoring station
Historical juncture corresponding fixed reference feature includes corresponding ozone concentration of any historical juncture, any historical juncture correspondence
Point source discharge characteristics and historical forecast moment meteorological condition feature, any historical juncture, corresponding reference result was institute
State the difference of historical forecast moment corresponding ozone concentration and corresponding ozone concentration of any historical juncture, the historical forecast
Moment is after any historical juncture；S2, training verification collection and test set are obtained, the training verification collection is by described each
Environmental monitoring website each historical juncture corresponding fixed reference feature and each environmental monitoring website each historical juncture correspond to
Reference result constituted according to the first default proportioning, the test set is by each environmental monitoring website each historical juncture pair
The fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture answered are according to the second default proportioning
It constitutes；S3, using statistical learning algorithm, on the training verification collection by ten folding cross validations carry out model parameter selection and
Model training obtains prediction model, carries out model measurement to the prediction model by the test set, utilizes model measurement
As a result the prediction model is compared and is evaluated；It is S4, special according to environmental monitoring website current time to be measured corresponding target
It seeks peace the prediction model, the environmental monitoring website current time to be measured corresponding objective result is obtained, according to described to be measured
Environmental monitoring website current time corresponding objective result and the environmental monitoring website current time to be measured corresponding ozone are dense
Degree obtains the ozone concentration at the environmental monitoring website current predictive moment to be measured, when the target signature includes described current
Carve corresponding ozone concentration, the current time corresponding point source discharge characteristics and the environmental monitoring website current predictive to be measured
The meteorological condition feature at moment, the objective result are the current predictive moment corresponding ozone concentration and the current time
The difference of corresponding ozone concentration.
Ozone concentration prediction technique proposed by the present invention based on space-time data statistical learning, can be to each environmental monitoring
Website individually establishes a statistical model, predicts the ozone concentration variable quantity after being carved 24 hours when each environmental monitoring website.
The purpose of the present invention is to propose to a kind of ozone concentration prediction techniques based on space-time data statistical learning, from ozone generating principle
Angle is set out, by carrying out multisource spatio-temporal datas such as air quality data, weather data, point source emissions datas across the time
The spatial Correlation Analysis of scale finds out its being associated between ozone concentration variation, and Related Mechanism is combined to carry out feature pumping
It takes, being associated between then statistical model fit characteristic being utilized to change with ozone concentration, to reach the mesh of ozone concentration prediction
's.
According to the generating principle of ozone, first to the air quality number in each environmental monitoring website the past period
Data parsing and pretreatment are carried out according to, space-time datas such as weather data, point source emissions data.For the prediction effect for ensureing final
Fruit, it is preferred to use at least continuous 1 year data.
Obtain each historical juncture corresponding fixed reference feature and corresponding reference during each environmental monitoring website is gone over 1 year
As a result, for wherein some environmental monitoring website, in past 1 year sometime, the moment corresponding fixed reference feature
For the moment corresponding ozone concentration, the moment corresponding point source discharge characteristics and 24 hours corresponding meteorologies after the moment
Condition flag.
Then, corresponding fixed reference feature of each environmental monitoring website each historical juncture and each environmental monitoring website is every
One historical juncture corresponding reference result is constituted according to the first default proportioning, obtains training verification collection, and the first default proportioning can be with
Determines according to actual conditions.By each environmental monitoring website each historical juncture corresponding fixed reference feature and each environmental monitoring station
Point each historical juncture corresponding reference result is constituted according to the second default proportioning, obtains test set.
Then, using statistical learning algorithms such as decision tree or XGBoost, it is not limited to decision tree herein or XGBoost is calculated
Method can also be other algorithms, and model parameter selection and model training are carried out by ten folding cross validations on training verification collection,
Prediction model is obtained, and model measurement is carried out to prediction model on test set, prediction model is compared with test result
And evaluation, then the parameter of prediction model is optimized or adjusted according to comparison and evaluation result.
It should be noted that decision tree be it is known it is various happen probability on the basis of, by constitute decision tree come
The desired value for seeking net present value (NPV) is more than or equal to zero probability, and assessment item risk judges the method for decision analysis of its feasibility, is
A kind of intuitive graphical method for using probability analysis.Since this decision branch is drawn as limb of the figure like one tree, therefore claim to determine
Plan tree.In machine learning, decision tree is a prediction model, and what it was represented is that one kind between object properties and object value is reflected
Penetrate relationship.
Ten folding cross validations, English name are called 10-fold cross-validation, are used for test model accuracy, are
Common test method and model parameter selection method.Data set is divided into ten parts, in turn will wherein 9 parts be used as training data, 1
Part is used as test data, is tested.Experiment can all obtain corresponding accuracy (or error rate) every time.10 results are just
The average value of true rate (or error rate) generally also needs to carry out multiple 10 folding cross validation (example as the estimation to arithmetic accuracy
Such as 10 10 folding cross validations), then its mean value is sought, as the estimation to model accuracy, to select preferably model parameter.
Finally, according to environmental monitoring website current time to be measured corresponding target signature and the prediction model, institute is obtained
Environmental monitoring website current time to be measured corresponding objective result is stated, is corresponded to according to the environmental monitoring website current time to be measured
Objective result and the environmental monitoring website current time to be measured corresponding ozone concentration, obtain the environmental monitoring station to be measured
The ozone concentration at point current predictive moment.
The embodiment of the present invention provides a kind of ozone concentration prediction technique based on space-time data statistical learning, is united by data
Model learning is counted, ozone concentration feature, meteorological condition feature and point source discharge characteristics are introduced to the mistake of statistical model feature extraction
Cheng Zhong more accurately portrays the raising reason of ozone concentration, and uses incremental analysis, and capturing point source etc. influences ozone concentration liter
High variable quantity weakens the invariants such as metastable natural source, and the statistical model before comparing, accuracy has significant increase.
The present invention has very strong versatility, can be adapted for different environmental monitoring websites；The present invention can also capture varying environment simultaneously
The otherness of monitoring station carries out personalized prediction model to different environmental monitoring station's points and trains.
In order to preferably illustrate the technical solution, with the ozone concentration forecasting problem of Xiamen City, Fujian Province environmental monitoring website
To represent.
First, corresponding fixed reference feature of each environmental monitoring website each historical juncture and each environmental monitoring are obtained
Website each historical juncture corresponding reference result, wherein the acquisition methods of fixed reference feature and reference result are all identical, by such as
Lower step obtains：
Air quality data is obtained first from the monitor value of environmental monitoring website and meteorological data is parsed, to air
It, only need to be by it according to time, environmental monitoring website number, pollutant when qualitative data and meteorological data are parsed
Format organization, monitoring record frequency herein is 1 hour/time.Table 1 is air quality data table, and table 2 is meteorological data
Table, as shown in Table 1 and Table 2, air quality data and meteorological data used herein are the monitor value of environmental monitoring website.
Table 1
Table 2
It should be noted that Characteristics of Air Quality can be extracted from air quality data, air quality data includes can
The data such as particle concentration, nitrous oxides concentration, carbonomonoxide concentration and ozone concentration are sucked, being all can be from environmental monitoring station
It is directly obtained in the monitor value of point.Meteorological condition feature can include temperature from meteorological condition extracting data, meteorological condition data
The data such as degree, humidity, air pressure, wind speed and direction.
Then point source discharge characteristics is obtained, it is contemplated that the generating principle of ozone is used only in point source exhaust emissions data
The conversion concentration of nitrogen oxides.Need to be hour grade by the processing of the point source exhaust emissions data of minute rank when carrying out data parsing
Not, so as to corresponding with air quality data and the monitoring time interval of meteorological data, and between the different point source monitoring times of solution
Every different problems.
Simply processing method is used herein, i.e., has the nitrogen oxides of record to convert concentration in per hour each point source
Average value converts the monitor value of concentration as this hour nitrogen oxides of the point source, then by data according to time, point source name, city
The format organization of concentration is converted in city, nitrogen oxides.Table 3 is that minute grade point source monitors initial data pattern, and table 4 is hour grade
Point source monitoring data pattern needs minute grade point source monitoring initial data pattern being converted to hour grade point source monitoring data sample
Formula.
Table 3
Table 4
Because the position of the dispersal direction that point source discharges pollutants and wind direction, point source and air quality website is related.Cause
When this data prediction, the position etc. of point source, environmental monitoring website is recorded in advance.Original point source, environmental monitoring website
Position recorded using longitude and latitude, for ease of calculating point source and environmental monitoring website spacing later from the meter with relative position
It calculates, carries out conversion of the longitude and latitude to Gaussian parabolic line using Gauss Kru&4&ger projection formula in advance herein, and by point source
It is recorded simultaneously with the longitude and latitude and rectangular co-ordinate of environmental monitoring website.
According to ozone generating principle and Pollutant Dispersion Law, needing to select one first may be to current environment monitoring station
The sensitizing range that the ozone concentration of point has an impact, the sensitizing range can border on city where the environmental monitoring website
Then this region is divided into the rectangle that the K length of side is M kilometers by all cities.To each rectangle, above-mentioned conversion is utilized
Gauss rectangular co-ordinate afterwards calculates the relative direction and distance of its central point and current environment monitoring station, as the grid to working as
The relative direction and distance of preceding environment monitoring station, and which point is recorded in the grid according to the location information of each point source includes
Source and the affiliated city of the grid, if point source belongs to multiple cities in the grid, then it is assumed that the grid belongs to the city more than number of point sources
City, if as many, therefrom randomly choosing a city as the affiliated city of the grid.The write-in of this grid partition information is deposited
Storage, convenient for directly invoking later.
It is such as directed to Xiamen City's website and carries out ozone prediction, point source range only considers the adjacent city in Xiamen City and Xiamen periphery
City, that is, Xiamen City, Zhangzhou City, Quanzhou City, Longyan totally four cities, it includes this four all point sources in city to look for one
Minimum rectangle, it is M kilometers of lattice to be divided into the K length of side.Record the above point source belonging positions information and grid partition information.
According to the number of Related Mechanism tectonic model fixed reference feature and reference result that ozone generating principle and pollutant are spread
According to right, prepared with statistical learning algorithm fitting for after.Per data to being organized as unit of hour, be divided into fixed reference feature and
Reference result two parts.
It is as follows that the fixed reference feature and reference result of any historical juncture t of some corresponding website calculates detailed process：When history
3 parts can be divided by carving the input feature vector of t, the ozone concentration feature of historical juncture t, historical juncture t based on pollutant diffusion machine
The point source discharge characteristics of reason and the meteorological condition feature of historical forecast moment t+24, specific calculation are as follows：
The ozone concentration y of historical juncture tt, analytically after air quality data in directly extract, historical juncture t
Ozone concentration characteristic dimension be 1.
The point source discharge characteristics of historical juncture t, the point source discharge characteristics based on pollutant diffusion mechanism need to consider environment
The point source of monitoring station geographic location and periphery carries out feature extraction using the K lattice handled well.
To each rectangle, the mechanistic features of 5 historical periods of rectangle historical juncture t are calculated, this 5 historical periods
For：When t-120 moment to t-97 moment, t-96 moment to t-73 moment, t-72 moment to t-49 moment, t-48 moment are to t-25
It carves and t-24 moment to t-1 moment, each historical period has 2 dimensional features, therefore, which shares 10 dimensional features.
If without point source in the rectangle, this 10 dimensional feature is 0.
If having point source in the rectangle, each historical period is individually calculated.To some historical period, first confirm that this is gone through
Predominant wind after the history period in 24 hours whether within the scope of the rectangle to positive and negative d ° of the environmental monitoring website true directions,
Such as this historical period of t-24 moment to t-1 moment considers t moment to the wind direction and wind speed at t+23 moment, wind direction and wind here
Speed is read by the weather data of the corresponding environmental monitoring website of the affiliated Urban Data of rectangle that front records and affiliated city
It takes.If the predominant wind after the historical period in 24 hours is not in the rectangle to positive and negative d ° of the environmental monitoring website true directions
In range, then 2 dimensional feature of historical period is 0.
If predominant wind after the historical period in 24 hours the rectangle to the environmental monitoring website true directions just
Within the scope of d ° negative, the mean wind speed within the scope of this wind direction in 24 hours thereafter can be calculatedWith in the period in the rectangle
The nitrogen oxides of all point sources converts concentration discharge summation s, in addition, can also read the rectangle to the environment by rectangular information
The actual distance d of monitoring station.
Each historical period corresponds to 2 dimensional features, this 2 dimensional feature one-dimensional representation Pollutant Dispersion Law feature is another
Dimension table shows dispersion of pollutants temporal characteristics.Wherein, Pollutant Dispersion Law is characterized asDispersion of pollutants temporal characteristicsIts
Middle tsepThe number of days interval for indicating the historical period and historical juncture, such as the t-24 moment to the number of days at t-1 moment and current time t
Interval is 1.The corresponding point source discharge characteristics dimension of the rectangle is 10K herein.
The meteorological condition feature of historical forecast moment t+24, main two kinds of numbers of humidity and wind speed for considering the historical forecast moment
According to.To two kinds of data of humidity and wind speed, considered respectively before the historical forecast moment in 4 hours, before the historical forecast moment in 6 hours and
Before the historical forecast moment in 24 hours three sections of periods maximum value, minimum value and average value, analytically after weather data
In directly extract.The dimension of the meteorological condition feature of historical forecast moment t+24 is 18 herein.
By historical juncture corresponding ozone concentration, any historical juncture corresponding point source discharge characteristics and historical forecast
Three kinds of features of meteorological condition feature of moment t+24 are spliced, the fixed reference feature of history of forming moment t, the dimension of the fixed reference feature
Degree is 10K+19.
The reference result of historical juncture t is historical forecast moment yt+24The variable quantity of ozone concentration compared to historical juncture t
Δ, i.e. Δ=yt+24-yt, wherein yt+24Indicate the ozone concentration of prediction time, ytIndicate the ozone concentration of historical juncture.
By the data pair of obtained fixed reference feature and reference result, according to 2:1 ratio is divided into training verification collection and test
Collection carries out model ginseng on training verification collection using statistical learning algorithms such as decision tree or XGBoost by ten folding cross validations
Number selection and model training, obtain prediction model, and carry out model measurement on test set, model comparison are carried out with test result
And evaluation.
Using obtained prediction model, the ozone concentration after 24 hours futures at current time can be predicted.Specifically
Process is as follows：
With reference to the fixed reference feature extracting method of historical juncture, target's feature-extraction is carried out to current time T.It should be noted that
Any is to need to use the weather prognosis number in the affiliated city of point source and the affiliated city of current environment monitoring station when extracting feature
According to the meteorological measured data after historical juncture t when being trained with weather prognosis data substitution model.
In the prediction model that the input training of obtained target signature is obtained, the ozone concentration for obtaining the current predictive moment becomes
Change amount Δ, current predictive moment are that current time is 24 hours following.
The result Δ of output is handled, by current time ozone concentration yTBecome with the ozone concentration of current time prediction
Change amount Δ is added, you can obtains the ozone concentration hour predicted value y at current predictive momentT+24, i.e. yT+24=yT+Δ.According to environment
Protection portion publication《Ambient air quality index (AQI) daily paper technical stipulation》It is smelly with following 24 hours of current time forecast
Oxygen concentration hour predicted value, you can corresponding ozone concentration prediction daily paper value and forecast ratings are calculated.
Fig. 2 is a kind of stream of the ozone concentration prediction technique based on space-time data statistical learning of one embodiment of the present invention
Cheng Tu, as shown in Fig. 2, this method comprises the following steps：
To the air quality data of each environmental monitoring website multiple historical junctures, weather data, point source emission number
According to progress data parsing and pretreatment, including point source rectangular partition.
According to Pollutant Dispersion Law, current time ozone concentration, gas such as humidity and wind speed in 24 hours after current time
Image data, point source emissions data and the point source rectangular information that generates when pretreatment in 5 days before current time, construction fixed reference feature and
Reference result.
By fixed reference feature and reference result according to 2:1 composition of proportions training verification collection and test set.
Using decision tree or XGBoost algorithms, model parameter choosing is carried out by ten folding cross validations on training verification collection
It selects and model training, obtains prediction model.
Model measurement is carried out to prediction model by test set, prediction model is compared using the result of model measurement
And evaluation is further processed according to comparison and evaluation result.
To the air quality data at environmental monitoring station's point current time to be measured, weather data, point source emissions data into
Row data parse and pretreatment, continue to use the point source division side when environmental monitoring website construction fixed reference feature and reference result to be measured
Method.
According to Pollutant Dispersion Law, current time ozone concentration, gas such as humidity and wind speed in 24 hours after current time
Image data, point source emissions data and the point source rectangular information that generates when pretreatment in 5 days before current time, construct target signature.
Using target signature as the input of prediction model, the ozone concentration of the prediction time of environmental monitoring website to be measured is obtained
Variable quantity.
According to ozone concentration variable quantity, output ozone hour prediction concentrations, ozone daily paper concentration and IAQI.
The present invention also provides a kind of ozone concentration forecasting system, which includes：Historical data module, it is each for obtaining
Environmental monitoring website each historical juncture corresponding fixed reference feature and each environmental monitoring website each historical juncture correspond to
Reference result, for any historical juncture of any environment monitoring station, any historical juncture corresponding fixed reference feature
It is pre- including any historical juncture corresponding ozone concentration, any historical juncture corresponding point source discharge characteristics and history
Survey the meteorological condition feature at moment, any historical juncture, corresponding reference result was that the historical forecast moment is corresponding smelly
The difference of oxygen concentration and corresponding ozone concentration of any historical juncture, the historical forecast moment is in any historical juncture
Later；Module is matched, for obtaining training verification collection and test set, the training verification collection is by each environmental monitoring website
Each historical juncture corresponding fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture are pressed
It is constituted according to the first default proportioning, the test set is by each environmental monitoring website each historical juncture corresponding fixed reference feature
Reference result corresponding with each environmental monitoring website each historical juncture is constituted according to the second default proportioning；Training test
Module carries out model parameter selection on the training verification collection for utilizing statistical learning algorithm by ten folding cross validations
And model training, prediction model is obtained, model measurement is carried out to the prediction model by the test set, utilizes model measurement
Result the prediction model is compared and is evaluated；Prediction module, for according to environmental monitoring website current time to be measured
Corresponding target signature and the prediction model obtain the environmental monitoring website current time to be measured corresponding objective result,
According to the environmental monitoring website current time corresponding objective result to be measured and the environmental monitoring website current time to be measured
Corresponding ozone concentration obtains the ozone concentration at the environmental monitoring website current predictive moment to be measured, the target signature packet
Include the current time corresponding ozone concentration, the current time corresponding point source discharge characteristics and the environmental monitoring to be measured
The meteorological condition feature at website current predictive moment, the objective result be the current predictive moment corresponding ozone concentration with
The difference of current time corresponding ozone concentration.
A kind of ozone concentration prediction technique and system based on space-time data statistical learning proposed by the present invention, advantage
It is：Ozone concentration prediction is carried out using statistical model, calculating speed is fast, and accuracy rate is high, is convenient for rail vehicle roller test-rig, compares numerical value
Pattern, calculating speed in the same order of magnitude, do not there is significant increase.The method of the present invention uses incremental analysis, by metastable shifting
The influence reduction in dynamic source and natural source, pay close attention to larger point source emission influences caused by ozone concentration.This method simultaneously
It is trained using truthful data, avoids influence of the inaccurate pollution sources listings data input to prediction result.The method of the present invention
Ozone generating machine reason and pollutant diffusion mechanism are introduced into statistical model feature extraction process, more accurately portray ozone concentration
Reason, the statistical model before comparing are increased, accuracy has significant increase.The method of the present invention has very strong versatility, Ke Yishi
For different environmental monitoring websites；This method can also capture the otherness of varying environment monitoring station simultaneously, to different websites
Carry out personalized prediction model training.
Finally, method of the invention is only preferable embodiment, is not intended to limit the scope of the present invention.It is all
Within the spirit and principles in the present invention, any modification, equivalent replacement, improvement and so on should be included in the protection of the present invention
Within the scope of.
Claims (10)
1. a kind of ozone concentration prediction technique, which is characterized in that including：
S1, each environmental monitoring website each historical juncture corresponding fixed reference feature of acquisition and each environmental monitoring website are every
One historical juncture corresponding reference result, for any historical juncture of any environment monitoring station, any historical juncture
Corresponding fixed reference feature includes corresponding ozone concentration of any historical juncture, any historical juncture corresponding point source row
Put feature and the meteorological condition feature at historical forecast moment, any historical juncture, corresponding reference result was that the history is pre-
The difference of moment corresponding ozone concentration and corresponding ozone concentration of any historical juncture is surveyed, the historical forecast moment is in institute
After stating any historical juncture；
S2, training verification collection and test set are obtained, when each by each environmental monitoring website history of the training verification collection
It is default according to first to carve corresponding fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture
Proportioning is constituted, and the test set is by each environmental monitoring website each historical juncture corresponding fixed reference feature and described each
Environmental monitoring website each historical juncture corresponding reference result is constituted according to the second default proportioning；
S3, using statistical learning algorithm, on the training verification collection by ten folding cross validations carry out model parameter selection and
Model training obtains prediction model, carries out model measurement to the prediction model by the test set, utilizes model measurement
As a result the prediction model is compared and is evaluated；
S4, according to environmental monitoring website current time to be measured corresponding target signature and the prediction model, obtain described to be measured
Environmental monitoring website current time corresponding objective result, according to environmental monitoring website current time to be measured corresponding target
As a result the corresponding ozone concentration with environmental monitoring website current time to be measured, it is current to obtain the environmental monitoring website to be measured
The ozone concentration of prediction time, the target signature include the current time corresponding ozone concentration, the current time pair
The meteorological condition feature of the point source discharge characteristics and the environmental monitoring website current predictive moment to be measured answered, the objective result
For the difference of the current predictive moment corresponding ozone concentration and the current time corresponding ozone concentration, the current predictive
Moment indicates 24 hours after the current time.
2. method according to claim 1, which is characterized in that any history of any environment monitoring station described in step S1
Moment, corresponding point source discharge characteristics was obtained by following steps：
S11, sensitizing range is obtained, the sensitizing range includes the institute bordered on the city where any environment monitoring station
There is city；
S12, the sensitizing range is divided into several identical rectangles, obtains described corresponding source emission of each rectangle
Feature obtains the point source discharge characteristics according to the point source emission subcharacter of each rectangle.
3. method according to claim 2, which is characterized in that corresponding source emission subcharacter of any rectangle includes several
The corresponding mechanistic features of historical period, mechanistic features corresponding for any historical period, if point source number in any rectangle
Mesh is 0, the corresponding mechanistic features of any historical period is set as 0, the mechanistic features include Pollutant Dispersion Law
Feature and dispersion of pollutants temporal characteristics, a historical period are a period of time before the historical juncture.
4. method according to claim 3, which is characterized in that if point source number is more than 0, and described in any rectangle
In one rectangle the wind direction of the first future time period and the environmental monitoring website to be measured relative direction not within a preset range, will
The corresponding mechanistic features of any historical period are set as 0, after first future time period is any historical period
A period of time.
5. method according to claim 3, which is characterized in that if point source number is more than 0, and described in any rectangle
In one rectangle in the wind direction of the first future time period and the relative direction of the environmental monitoring website to be measured in the preset range,
Calculate the corresponding mechanistic features of any historical period.
6. method according to claim 5, which is characterized in that according to quick in first future time period in any rectangle
Feel all point sources nitrogen oxide emission in any historical period in the mean wind speed and any rectangle of wind
Summation, obtains the Pollutant Dispersion Law feature, and the wind direction of the sensitivity wind is opposite with the environmental monitoring website to be measured
Direction is in the preset range.
7. method according to claim 5, which is characterized in that according to any rectangle and the environmental monitoring website to be measured
Distance, in any rectangle sensitive wind described in first future time period mean wind speed, any historical period
With the time difference between any historical juncture, the dispersion of pollutants temporal characteristics are obtained.
8. method according to claim 7, which is characterized in that any rectangle and the environmental monitoring website to be measured away from
It is obtained from by following steps：
Obtain the longitude and latitude of any rectangular centre point longitude and latitude and the environmental monitoring website to be measured；
By Gauss Kru&4&ger projection's formula, the longitude and latitude of any rectangular centre point is converted in rectangular coordinate system
First rectangular co-ordinate is converted to the longitude and latitude of the environmental monitoring website to be measured described by Gauss Kru&4&ger projection's formula
The second rectangular co-ordinate in rectangular coordinate system；
According to first rectangular co-ordinate and second rectangular co-ordinate, any rectangle and the environmental monitoring to be measured are obtained
The distance of website.
9. method according to claim 1, which is characterized in that when any historical forecast of environmental monitoring website described in step S1
The meteorological condition feature at quarter specifically includes：The humidity average value of three periods of the environmental monitoring website, humidity maximum value, humidity
Minimum value, wind speed average value, wind speed maximum value and wind speed minimum value, three periods are 4 before the historical forecast moment
Hour is pre- to the historical forecast moment, the history to 6 hours before the historical forecast moment, the historical forecast moment
24 hours to the historical forecast moment before the survey moment.
10. a kind of ozone concentration forecasting system, which is characterized in that including：
Historical data module, for obtaining corresponding fixed reference feature of each environmental monitoring website each historical juncture and described each
Environmental monitoring website each historical juncture corresponding reference result, for any historical juncture of any environment monitoring station, institute
When to state corresponding fixed reference feature of any historical juncture include corresponding ozone concentration of any historical juncture, any history
The meteorological condition feature at corresponding point source discharge characteristics and historical forecast moment is carved, any historical juncture is corresponding with reference to knot
Fruit is the difference of the historical forecast moment corresponding ozone concentration and corresponding ozone concentration of any historical juncture, described to go through
History prediction time is after any historical juncture；
Module is matched, for obtaining training verification collection and test set, the training verification collection is by each environmental monitoring website
Each historical juncture corresponding fixed reference feature and corresponding reference result of each environmental monitoring website each historical juncture are pressed
It is constituted according to the first default proportioning, the test set is by each environmental monitoring website each historical juncture corresponding fixed reference feature
Reference result corresponding with each environmental monitoring website each historical juncture is constituted according to the second default proportioning；
Training test module is carried out on the training verification collection by ten folding cross validations for utilizing statistical learning algorithm
Model parameter selects and model training, obtains prediction model, and model measurement is carried out to the prediction model by the test set,
The prediction model is compared and evaluated using the result of model measurement；
Prediction module, for according to environmental monitoring website current time to be measured corresponding target signature and the prediction model, obtaining
The environmental monitoring website current time to be measured corresponding objective result is obtained, according to the environmental monitoring website current time to be measured
Corresponding objective result and the environmental monitoring website current time to be measured corresponding ozone concentration obtain the environment prison to be measured
The ozone concentration at survey station point current predictive moment, the target signature includes the current time corresponding ozone concentration, described
The meteorological condition feature of current time corresponding point source discharge characteristics and the environmental monitoring website current predictive moment to be measured, institute
State the difference that objective result is the current predictive moment corresponding ozone concentration and the current time corresponding ozone concentration, institute
Stating the current predictive moment indicates 24 hours after the current time.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711168594.7A CN107943928B (en) | 2017-11-21 | 2017-11-21 | A kind of ozone concentration prediction technique and system based on space-time data statistical learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711168594.7A CN107943928B (en) | 2017-11-21 | 2017-11-21 | A kind of ozone concentration prediction technique and system based on space-time data statistical learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107943928A CN107943928A (en) | 2018-04-20 |
| CN107943928B true CN107943928B (en) | 2018-10-30 |
Family
ID=61930560
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711168594.7A Active CN107943928B (en) | 2017-11-21 | 2017-11-21 | A kind of ozone concentration prediction technique and system based on space-time data statistical learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107943928B (en) |
Families Citing this family (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109085296B (en) * | 2018-08-14 | 2019-07-05 | 成都市环境保护科学研究院 | Judge that ozone generates sensibility and precursor control method, device, storage medium, terminal |
| CN109685246B (en) * | 2018-11-13 | 2024-04-23 | 平安科技（深圳）有限公司 | Environment data prediction method and device, storage medium and server |
| CN109541730A (en) * | 2018-11-23 | 2019-03-29 | 长三角环境气象预报预警中心（上海市环境气象中心） | A kind of method and apparatus of pollutant prediction |
| CN110333556A (en) * | 2019-06-03 | 2019-10-15 | 深圳中兴网信科技有限公司 | Air Quality Forecast method, apparatus, computer equipment and readable storage medium storing program for executing |
| CN110378520A (en) * | 2019-06-26 | 2019-10-25 | 浙江传媒学院 | A kind of PM2.5 concentration prediction and method for early warning |
| CN110346517B (en) * | 2019-07-25 | 2021-06-08 | 中南大学 | Smart city industrial atmosphere pollution visual early warning method and system |
| CN113125635B (en) * | 2019-12-31 | 2023-04-28 | 北京蛙鸣华清环保科技有限公司 | Atmospheric pollution early warning method and device and readable storage medium |
| CN111157682A (en) * | 2020-01-06 | 2020-05-15 | 上海应用技术大学 | Air quality monitoring and predicting system and method |
| CN111289040A (en) * | 2020-04-03 | 2020-06-16 | 中科三清科技有限公司 | Navigation monitoring system |
| CN113722288A (en) * | 2021-07-27 | 2021-11-30 | 张博 | Modeling method for time-space data statistics |
| CN114518611B (en) * | 2021-12-24 | 2023-06-06 | 山东省青岛生态环境监测中心(中国环境监测总站黄海近岸海域环境监测分站) | Ozone forecasting method based on similar case discriminant analysis |
| CN117236528B (en) * | 2023-11-15 | 2024-01-23 | 成都信息工程大学 | Ozone concentration forecasting method and system based on combined model and factor screening |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105243444A (en) * | 2015-10-09 | 2016-01-13 | 杭州尚青科技有限公司 | City monitoring station air quality prediction method based on online multi-core regression |
| CN106019409A (en) * | 2016-05-11 | 2016-10-12 | 北京市环境保护监测中心 | Partition prediction method and system for ozone concentration |
| CN106651036A (en) * | 2016-12-26 | 2017-05-10 | 东莞理工学院 | Air quality forecasting system |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| SE0003054D0 (en) * | 2000-08-30 | 2000-08-30 | Sc Trade & Consult Ab | Method and device |
| US20040050798A1 (en) * | 2002-09-14 | 2004-03-18 | Schulz Christopher R. | Method and apparatus for ozone disinfection of liquid-carrying conduits |
| US20040053415A1 (en) * | 2002-09-14 | 2004-03-18 | Schulz Christopher R. | Ozone-in-water decay rate analyzer |
| CN105868559A (en) * | 2016-03-29 | 2016-08-17 | 北京师范大学 | Atmospheric particulate mass concentration fitting method |
-
2017
- 2017-11-21 CN CN201711168594.7A patent/CN107943928B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105243444A (en) * | 2015-10-09 | 2016-01-13 | 杭州尚青科技有限公司 | City monitoring station air quality prediction method based on online multi-core regression |
| CN106019409A (en) * | 2016-05-11 | 2016-10-12 | 北京市环境保护监测中心 | Partition prediction method and system for ozone concentration |
| CN106651036A (en) * | 2016-12-26 | 2017-05-10 | 东莞理工学院 | Air quality forecasting system |
Non-Patent Citations (1)
| Title |
|---|
| Assessment and prediction of tropospheric ozone concentration levels using artificial neural networks;S.A. Abdul-Wahab等;《Environmental Modelling & Software》;20021231;第219-228页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107943928A (en) | 2018-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107943928B (en) | A kind of ozone concentration prediction technique and system based on space-time data statistical learning | |
| Kumar et al. | Time series data prediction using IoT and machine learning technique | |
| Guo et al. | Air pollution forecasting using artificial and wavelet neural networks with meteorological conditions | |
| CN110567510A (en) | Atmospheric pollution monitoring method, system, computer equipment and storage medium | |
| CN112766549A (en) | Air pollutant concentration forecasting method and device and storage medium | |
| Nguyen et al. | Development of the typical driving cycle for buses in Hanoi, Vietnam | |
| Bravo et al. | Spatial heterogeneity of PM10 and O3 in São Paulo, Brazil, and implications for human health studies | |
| Kingsy et al. | Air pollution analysis using enhanced K-Means clustering algorithm for real time sensor data | |
| Mubea et al. | Assessing application of Markov chain analysis in predicting land cover change: a case study of Nakuru municipality | |
| Tella et al. | GIS-based air quality modelling: Spatial prediction of PM10 for Selangor State, Malaysia using machine learning algorithms | |
| CN103197299A (en) | Extraction and quantitative analysis system of weather radar radial wind information | |
| Zhu et al. | Granger-causality-based air quality estimation with spatio-temporal (ST) heterogeneous big data | |
| Huang et al. | Surveillance efficiency evaluation of air quality monitoring networks for air pollution episodes in industrial parks: Pollution detection and source identification | |
| Maantay et al. | Loose-coupling an air dispersion model and a geographic information system (GIS) for studying air pollution and asthma in the Bronx, New York City | |
| CN110533239A (en) | A kind of smart city air quality high-precision measuring method | |
| Leung et al. | Integration of air pollution data collected by mobile sensors and ground-based stations to derive a spatiotemporal air pollution profile of a city | |
| CN113836808A (en) | PM2.5 deep learning prediction method based on heavy pollution feature constraint | |
| CN117809439A (en) | River discharge abnormality early warning system based on multiple environmental factors | |
| Kim et al. | Long-term behavior and stability of calibration models for NO and NO 2 low-cost sensors | |
| CN117332906B (en) | Machine learning-based three-dimensional space-time grid air quality prediction method and system | |
| Hassani et al. | Citizen-operated mobile low-cost sensors for urban PM2. 5 monitoring: field calibration, uncertainty estimation, and application | |
| Jia et al. | Comparison of the Gaussian plume and puff atmospheric dispersion models for methane modeling on oil and gas sites | |
| Parkavi et al. | Deep learning model for air quality prediction based on big data | |
| CN115526330A (en) | Organic matter navigation data calibration method and device, computer equipment and storage medium | |
| Martínez et al. | TAQE: a data modeling framework for traffic and air quality applications in smart cities |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |