Wind-power intra-day multi-step predictions using polynomial networks solutions of general PDEs based on Operational Calculus.

Result code in IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F19%3A10244102" target="_blank" >RIV/61989100:27240/19:10244102 - isvavai.cz</a>

Result on the web

<a href="https://inis.iaea.org/search/search.aspx?orig_q=RN:52034314" target="_blank" >https://inis.iaea.org/search/search.aspx?orig_q=RN:52034314</a>

DOI - Digital Object Identifier

—

Result language

angličtina

Original language name

Wind-power intra-day multi-step predictions using polynomial networks solutions of general PDEs based on Operational Calculus.

Original language description

Precise intra-day predictions of wind-power are challenging due to its intermittent nature and high correlation with large-scale atmospheric chaotic circulation processes. NWP systems solve sets of differential equations to predict a time-change of each 3D-grid cell in several atmospheric layers. Their surface forecasts of wind speed are not entirely adapted to specific local characteristics and anomalies, which largely influence its temporal-flow. AI methods us-ing historical observations can convert and refine the daily forecasts in consideration of wind farm siting, terrain asperity and ground level (hub height). Their independent wind-power predictions in horizon of several hours are also more precise then NWP model forecasts as these are usually produced every 6 hours. The designed method uses Polynomial neural networks to decompose and sub-stitute for the general linear Partial Differential Equation being able to describe n-variable functions of unknown complex dynamic systems. It solves specific 2-variable 2nd order PDEs, formed in PNN nodes, using a polynomial conver-sion based on Operational Calculus. The inverse Laplace transformation is ap-plied to the resulting rational terms to obtain the originals of node functions whose sum gives the complete PDE model. The composite PDE models are developed with data samples from the estimated optimal numbers of training days to represent spatial data relations in current weather, necessary for applicable predictions. They can predict wind power up to 12 hours ahead according to a trained data inputs-&gt;output time-shift. Intra-day multi-step predictions using the PDE models are more precise than those based on NWP model forecasts or statistical techniques allowing using local time-series of several variables only.

Czech name

—

Czech description

—

Type

D - Article in proceedings

CEP classification

—

OECD FORD branch

10200 - Computer and information sciences

Project

Result was created during the realization of more than one project. More information in the Projects tab.

Continuities

P - Projekt vyzkumu a vyvoje financovany z verejnych zdroju (s odkazem do CEP)

Publication year

2019

Confidentiality

S - Úplné a pravdivé údaje o projektu nepodléhají ochraně podle zvláštních právních předpisů

Article name in the collection

ITISE 2019 : International Conference on Time Series and Forecasting : proceedings of papers : 25-27 September 2019, Granada, Spain

ISBN

978-84-17970-78-9

ISSN

—

e-ISSN

—

Number of pages

10

Pages from-to

1-10

Publisher name

University of Granada

Place of publication

Granada

Event location

Granada

Event date

Sep 25, 2019

Type of event by nationality

WRD - Celosvětová akce

UT code for WoS article

—