METHOD FOR INTELLIGENT PREDICTION OF PROBABILITY OF OCCURRENCE OF NATURAL FIRES Russian patent published in 2024 - IPC G06N3/02 A62C3/02 

FIELD: information technology.

SUBSTANCE: invention relates to means of intelligent prediction of probability of occurrence of natural fires. A set of geospatial data is obtained, containing meteorological measurements for 7 days before the probability of occurrence of natural fires, Earth cover data (type of vegetation cover, vegetation indices of vegetation cover, fraction of absorbed photosynthetic active radiation), spatial data (height above sea level, direction of slope, incline, population density and distance from roads). Obtained data set is processed with the help of convolutional neural network (CNN), trained on archive data sets, containing all of the above groups of features, as well as reference marking with coordinates of the actually recorded occurrence of natural fires. Based on this CNN processing, a regular spatial grid is constructed for a given area of prediction of the probability of occurrence of new natural fires. Then the probability of occurrence of natural fires is predicted for each of the next five days on the given regular grid, as a result of which 5 maps are obtained.

EFFECT: high efficiency of forecasting probability of occurrence of natural fires for five days of forecasting.

9 cl, 3 dwg