Registration Number
0223U005655,
(0121U110229)
, R & D reports
Title
Predictive analytics for meteorology using the Internet of Things
popup.stage_title
Предикативна аналітика для метеорології з використанням Інтернету речей.
Head
Korchak Yurii M., Кандидат фізико-математичних наук
Registration Date
26-12-2023
Organization
Ivan Franko National University of Lviv
popup.description1
Software implementation of algorithms for assessing and forecasting the development of the meteorological situation and supplementing data based on information collected using the Internet of Things.
popup.description2
The development of the basic weather station and the method of predicative forecasting of time series was carried out.
A station prototype based on an ESP32 microcontroller was assembled, an architectural solution based on the LORA protocol and data exchange with the TTN IoT network was chosen. The expansion of the weather complex to the required list of data was implemented and methods of data transmission using IoT were developed.
Modern methods of forecasting time series are considered. For this purpose, weather data collected by the weather station of the Lviv Airport (UKLL) for 2016-2021 were used, as well as weather data in the city of Lviv from the websites www.weather.com for 2 years (period 2019-2021) and www. accuweather.com for the period 2010 - 2023. These different sources of weather data served as a conditional network of weather complexes. A comparative analysis of the effectiveness of forecasting temperature changes based on a number of time series models with and without the use of exogenous parameters (humidity, atmospheric pressure) was carried out. Conventional ARIMA models are found to provide better short-term forecasting, while models that use seasonality can perform at longer time intervals. Taking into account exogenous parameters somewhat worsened the accuracy of forecasting, because then the error of their own forecast changes was taken into account. As a result, the use of a simple hybrid model for half-hourly weather forecasting based on ARIMA (for short-term perspective) and Prophet (for long-term perspective) is justified. Thus, on the basis of data collected at one point from different weather stations, it is possible to build a realistic forecast for the same location. An evaluation of the time limits of the applicability of the Prophet model for forecasting the local weather situation was carried out and it was found that the use of the Prophet library gives a good quality of prediction, at least in the case of a weekly perspective.
Product Description
popup.authors
Furhala Yurii M.
Blahitka Marta Ya.
popup.nrat_date
2023-12-26
