Machine Learning Approach for Analysis and Prediction of Cloud Particle Size and Shape Distribution

摘要

Techniques for analysis and prediction of cloud particle distribution and solar radiation are provided. In one aspect, a method for analyzing cloud particle characteristics includes the steps of: (a) collecting meteorological data; (b) calculating solar radiation values using a radiative transfer model based on the meteorological data and blended guess functions of a cloud particle distribution (c) optimizing the cloud particle distribution by optimizing the weight coefficients used for the blended guess functions of the cloud particle distribution based on the solar radiation values calculated in step (b) and measured solar radiation values; (d) training a machine-learning process using the meteorological data collected in step (a) and the cloud particle distribution optimized in step (c) as training samples; and (e) predicting future solar radiation values using forecasted meteorological data and the machine-learning process trained in step (d).

主权项

1. A system for analyzing cloud particle characteristics, comprising:
(a) a databus module configured to collect meteorological data; (b) a guess function module configured to provide blended guess functions of a cloud particle distribution; (c) a radiative transfer model module configured to calculate solar radiation values using a radiative transfer model based on the meteorological data from the databus module and the blended guess functions of cloud particle distribution from the guess function module such that the solar radiation values are generated for each of the blended guess functions; (d) a radiation measurement data module configured to collect measured solar radiation values; and (e) a machine learning module configured to (i) optimize the cloud particle distribution by optimizing the weight coefficients used for the blended guess functions of the cloud distribution based on the solar radiation values from the radiative transfer model module and the measured solar radiation values from the radiation measurement data module, (ii) train a machine-learning process using the meteorological data collected by the databus module and the cloud particle distribution optimized in step (i) as training samples, and (iii) predict future solar radiation values using forecasted meteorological data and the machine-learning process trained in step (ii).