US 11,921,257 B2
Forecasting method with machine learning
Daniel Rothenberg, Boston, MA (US)
Assigned to The Tomorrow Companies Inc., Boston, MA (US)
Filed by The Tomorrow Companies Inc., Boston, MA (US)
Filed on May 16, 2022, as Appl. No. 17/745,498.
Application 17/745,498 is a continuation of application No. 16/669,479, filed on Oct. 30, 2019, granted, now 11,402,542.
Application 16/669,479 is a continuation of application No. PCT/US2018/067203, filed on Dec. 21, 2018.
Application PCT/US2018/067203 is a continuation in part of application No. 16/181,137, filed on Nov. 5, 2018, granted, now 10,962,680, issued on Mar. 30, 2021.
Application 16/181,137 is a continuation in part of application No. 16/181,148, filed on Nov. 5, 2018, granted, now 11,294,096, issued on Apr. 5, 2022.
Claims priority of provisional application 62/752,781, filed on Oct. 30, 2018.
Prior Publication US 2022/0291420 A1, Sep. 15, 2022
Int. Cl. G01W 1/10 (2006.01); G06N 5/04 (2023.01); G06N 20/00 (2019.01)
CPC G01W 1/10 (2013.01) [G06N 5/04 (2013.01); G06N 20/00 (2019.01)] 33 Claims
OG exemplary drawing
 
1. A computer-automated method for generating a weather forecast from current weather data and an intensity value associated with an historical storm, comprising:
obtaining or deriving current weather data including weather data stored in a database associated with one or more weather parameter tiles, the current weather data derived at least in part from at least one sensor, and
performing, with at least one processor in response to instructions stored in at least one non-transitory memory, operations comprising:
determining when the obtained current weather data includes characteristics indicative of a storm, and detecting storm presence when the obtained current weather data is determined to include characteristics indicative of a storm,
retrieving one or more tracked storms from at least one database,
comparing characteristics of the detected storm presence with characteristics of the retrieved one or more tracked storms including retrieving a tracked storm intensity profile associated with the one or more tracked storms from the at least one database,
updating the tracked storm intensity profile based, at least in part, on matching of the characteristics of the detected storm presence with stored tracked storm characteristics, and saving the updated tracked storm intensity profile to the at least one database,
obtaining information corresponding to one or more historical storms from at least one historical storm database, wherein the obtained information includes characteristics of a historical storm and a historical storm intensity profile associated with the historical storm,
comparing reflectivity and/or storm path characteristics of the one or more tracked storms with reflectivity and/or storm path characteristics of the one or more historical storms,
selecting the historical storm in response to the comparison of the reflectivity and/or storm path characteristics of the one or more tracked storms with characteristics of the one or more historical storms,
predicting a future value of a weather parameter associated with the one or more tracked storms,
using the historical storm intensity profile associated with the selected historical storm to adjust the predicted future value of the weather parameter,
assigning the adjusted predicted future value of the weather parameter to one or more tile layers, and
storing the assigned predicted future value of the weather parameter in a data structure associated with the one or more tile layers.