Research On The Adaptive Selection Method Of Air Quality Prediction Model Based On Scenario Characteristics

Air quality prediction models have been established as an important solution for air pollution forecasting and control.With the development of air quality prediction model,its type and quantity are more and more abundant,and the application requirements of these models are becoming more and more extensive.However,different models have significantly distinguished characteristics in terms of mechanism,applicable area and applicable scale,and thus actual modeling applications often require tedious consideration and treatment of the applicable scenarios of various models.This increases not only the difficulty of model application for users,but also the uncertainty of the comprehensive simulation results involving integrated applications of multi-disciplinary and multi-domain models.Contemporarily,the air quality prediction models' evaluation strategies and indicators are independent,it is difficult to support the model selection which is suitable for diverse geographic problems,leading to the lack of systematic research on the applicable scenarios of models.Therefore,beginning with the analysis of applicable scenarios of the air quality prediction model,this paper takes the establishment of the scenario feature system as the starting point and summarizes the model features and the user features to form the scenario feature system.Based on this scenario feature system,a model feature database and a user scenario database were constructed separately.The two databased were associated by index to meet the application requirement of the air quality forecast models.Led by matching and recommendation,this paper studies the adaptive selection method of models to achieve both the model matching based on situational features and model-driven data resource recommendation.The main observations and conclusions of this research include:(1)Facing the systemic research needs of model applicable scenarios,a scenario feature system is established based on both the model features and user scenario features in order to form a model feature sub-system and user scenario feature subsystem.Guided by the model feature sub-system,the model feature database is established from the perspective of modeling mechanism and applicable area.Meanwhile,under the guidance of the user's use scenario feature sub-system,the user scenario database is built under unlimited and limited conditions.Through the establishment of scenario feature system and feature database,this paper provides accordance and support the model matching in the adaptive selection method.(2)This research develops an adaptive selection method for air quality prediction models.Towards the application requirement of air quality prediction models,an adaptive selection method is proposed,including a model matching method based on scenario characteristics,and a data resource recommendation method driven by model execution.For the model matching method,the similarity calculation method is used as the core to realize model matching based on scenario features.In terms of the data resource recommendation method,a data resource database and a data processing tool resource database are built and expressed structurally.With the resource association method as the core,the model execution-driven data resource recommendation is realized through the binding association between scenario feature library and structured document.(3)This study proposes a model adaptive selection prototype system.A prototype system for model adaptive selection was designed and established based on the research on the scenario feature system and the model adaptive selection method.The userdefined scenario supported by front-end page has been developed to meet the user's application requirements of air quality prediction models in different scenarios.With the open access of model resources,data resources and data processing tool resources,users are capable to use the networked resources for online matching and running of models in an open network environment.On the basis of this prototype system,the proposed model adaptive selection method is tested and verified for two types of user scenarios.In conclusion,along with the establishment of the scenario feature system and feature database and the model adaptive selection method,this paper systematically associates the air quality forecast model with the actual application requirements,which breaks the limitation to a specific research field with a greater applicability.In addition,the adaptive selection method of the air quality forecast model supports the application of air quality prediction models in different scenarios,which helps to reduce the difficulty of model application and provide a good reference for further promotion and application of the air quality forecast model.