1. technical name

Artificial intelligence-based accurate identification of pollution source-traceability analysis-prediction and prediction technology

2. scope of application

Identification, traceability and prediction of urban pollution conditions.

3. technical content

Integrate multiple data sources such as video data, pollutant concentrations, meteorological data, and emission sources in a unified manner, and use artificial intelligence (AI) technology to accurately identify pollution sources, analyze traceability contribution, and predict pollutant concentration. This technology realizes a series of functions from underlying data source access, data processing, data fusion to feature extraction, and model construction.

3.2 main innovations

(1) A technology for data fusion and exception processing based on the spatial distribution of data monitoring stations is proposed. The space is divided into regions in a certain way, and then the data in the same area is fused in a specified way. For areas where there are still missing data, the Gaussian regression process method is used to comprehensively consider time domain and spatial information and nonlinear changes in pollutant concentrations, and nonlinear interpolation is performed on the missing areas, thereby effectively improving the stability of the data and eliminating noise. At the same time, it can also play a role in reducing the dimension of the data.

(2) Based on the city's existing video data resources, use the FSSD detection model and the ResNet classification network model to intelligently identify whether muck trucks are covered and dust on roads and construction sites, so that relevant ecological and environmental illegal information can be grasped at the first time and improve law enforcement personnel. Case handling efficiency, effectively fix evidence, connect with the case handling system, and achieve refined management.

(3) Combined with Kalman filtering and fully connected neural network technology, the proportion of pollutants generated inside and outside the target area where the central station is located and the contribution of pollutants input from various directions can be quickly and quantitatively calculated in real time, so that relevant personnel can take relevant measures in a targeted manner. Carry out management and disposal.

(4) A regional air pollution prediction model based on AI technology has been developed to achieve hourly accurate prediction of long-term changes in the concentration of 6 parameters in the future.

4. Pollution control or environmental restoration effects

It can timely grasp ecological and environmental illegal information, effectively fix evidence, and realize closed-loop law enforcement such as rapid discovery, rapid response, and emergency command; realize dynamic scheduling and real-time management and control of pollution control, and complete closed-loop management from discovering problems to solving problems. After statistics, the accuracy of the model's prediction range is 78% higher on average than that of other similar technologies.

5. Technology demonstration

This technology has been applied in many cities such as Daxing District in Beijing, Hechuan District in Chongqing, Qingxu in Taiyuan, Liaocheng in Shandong, and Huaibei in Anhui. In the smart environmental protection project in Hechuan District, Chongqing City, environmental monitoring equipment is installed on social operating vehicles On (taxis, buses, etc.), mobile monitoring of the atmospheric environment is carried out. After discovering pollution incidents such as smoke burning, uncovered muck trucks, and black smoke trucks, it automatically alarms and pushes information synthesis data to the platform and the Internet of Things. The equipment carries out linkage display and accurate positioning, which facilitates grid workers to quickly handle problems. At the same time, it combines the four-level environmental protection grid management mechanism of "city, district, township, and community", It achieves traces and quantification of performance assessment throughout the entire process of handling related pollution incidents, effectively supports environmental management decisions, and also provides fast and reliable data support for environmental emergency management.

6. investment estimation

Based on estimates of the country's 117 key cities, the investment in each city is about 20 million yuan.

7. investment recovery period

The payback period is 2 to 3 years.

8. Prospects for transformation and promotion of technological achievements

This technology combines artificial intelligence technology with Internet of Things big data analysis to realize intelligent identification of complex urban pollution scenarios, traceability and prediction of air quality, provides efficient, rapid identification and accurate decision support for ecological and environmental management, and solves the problem of difficulty and slow evidence collection for violations. The problem of difficulty and slow evidence collection has changed the past situation of relying on human and sea tactical inspections with low efficiency and high investment, and has improved the efficiency of environmental governance work. Optimize the allocation of urban human resources and have broad future development prospects