- Статьи
- Science and technology
- Weather conditions in the house: heating in a separate courtyard will save utilities from accidents
Weather conditions in the house: heating in a separate courtyard will save utilities from accidents
Russian scientists have created an artificial intelligence-based model that is able to predict the weather in a particular yard. She is able to "land" large-scale weather data on a local terrain map. Unlike analogues for average forecasts, the system is aimed at calculating abnormal phenomena such as hurricane-force winds or heavy rains. This allows her to warn public utilities and special services about upcoming emergencies. According to experts, the introduction of technology opens up the possibility for municipal institutions to avoid accidents and move faster to a "smart housing and communal services" based on saving resources.
AI-weather forecast for the yard
Specialists from the MIPT Institute of Artificial Intelligence have developed a climate analysis platform capable of predicting the weather with accuracy to an individual yard or even a house. Unlike analogs that can also work with such high resolution, the new AI-based system is designed to forecast abnormal events such as hurricanes, winds, or heavy downpours. Such events often cause emergencies, so the development will be useful for utilities and special services, which will be able to prepare in advance for possible danger.
— Even high-resolution weather forecasting services calculate averages for daily predictions of events that occur with high probability. For example, they can easily predict wind speeds of 10 m/s. But a wind with a speed of 30 m/s is very rare, but if it happens, it leads to a serious emergency. Our system is designed to prevent such extreme events. We set a different task for artificial intelligence than in standard forecasting models, and this required a complete redesign of our model," said Ivan Novikov, a junior researcher at MIPT.
Most weather forecasts describe the situation on a scale of tens of kilometers. Although conditions inside the city vary from street to street. In the center of a megalopolis, the air is often 10 degrees or more warmer than on the outskirts, and the geometry of streets and dense buildings often forms "heat traps" and "street canyons" of wind, which noticeably change the local microclimate. All this increases the peak load on the power grid and can overload city services, from storm sewers to public transport.
The new development takes all these factors into account, as it is based on downscaling technology: algorithms analyze the general forecast for the city and "land" it to the streets, filling it with real data from city cameras, weather stations, sensors and metering devices. If standard models predict the weather for sections with a step of 25-30 km, then the MIPT invention narrows it to 30-50 m — within one house.
The AI platform does not just transmit data, but constantly analyzes how and why atmospheric processes occurred at a specific point with known parameters (high-rise buildings, asphalt, green areas). She uses the accumulated relationships to further increase the accuracy of forecasts.
— For example, our system will be able to predict that the wind in a narrow alley between skyscrapers in this direction will increase threefold, and in one of the squares, due to the "heat island" effect, the temperature will be 10 degrees higher than in the park around the corner, - said Denis Lobas, head of the industrial products department at the Institute of Artificial Intelligence at MIPT.
Transition to "lean housing and communal services"
The platform visualizes threats on an interactive map and provides services with a convenient tool for real-time monitoring. This way, developers can assess in advance how the architecture of new buildings will affect the climate of a particular area.
Alexey Karnaukhov, a climatologist and a leading researcher at the Pushchino Scientific Center for Biological Research of the Russian Academy of Sciences, told Izvestia that the methodology presented by the developers for linking weather forecasts to a specific area may be useful.
— For example, heavy rains are expected. Then the house on the hill remains out of the risk of flooding, and the house next to the reservoir, which may overflow its banks, may be in an emergency situation. However, to apply the system, a very large amount of data will need to be entered into the model. But we already know examples where developers have succeeded," he said.
According to Marina Kholod, a leading researcher at the Laboratory of Artificial Intelligence, Neurotechnology and Business Analytics at Plekhanov Russian University of Economics, the technology of downscaling forecasting allows us to move from responding to and eliminating the consequences of weather emergencies to anticipating them, which means that it is possible to mobilize resources in advance. With its help, you can calculate the most vulnerable places in the infrastructure and target equipment and repair teams there.
— By predicting "heat islands" in buildings and abnormal heat in specific areas, energy supply services can more accurately predict the load on the network and prevent possible accidents, that is, it becomes possible to reduce peak loads on the infrastructure. The system will allow you to send personal warnings to residents with health risks. For example, people with cardiovascular diseases will know in advance about the onset of abnormal heat in their particular yard, and asthmatics will know about unfavorable conditions with a high concentration of pollutants in the air," the expert said.
It is also possible to predict the formation of ice on specific overpasses or road sections, strong wind gusts on bridges, which will allow roads to be treated with reagents in advance and warn drivers about dangerous areas, she added.
For the solution to work, it needs access to local weather observation data, but weather stations and other sources of information are extremely unevenly distributed throughout the city, and it will not be possible to achieve accurate results based on fragmentary data, said Alexander Bukhanovsky, director of the ITMO University's Megafacult of Translational Information Technologies and an expert at the ITMO-based National Center for Cognitive Research. However, the proposed approach may be very much in demand in the tasks of "lean housing and communal services", for example, for flexible control of building heating modes, he concluded.
Переведено сервисом «Яндекс Переводчик»
