Power Quality 24-h Prediction Based on L-Transform Derivative Modular and Deep Lear-ning Statistics Using Environmental Data in Detached Smart Buildings

Kód výsledku v IS VaVaI

<a href="https://www.isvavai.cz/riv?ss=detail&h=RIV%2F61989100%3A27240%2F25%3A10258614" target="_blank" >RIV/61989100:27240/25:10258614 - isvavai.cz</a>

Výsledek na webu

<a href="https://link.springer.com/chapter/10.1007/978-3-032-02725-2_1" target="_blank" >https://link.springer.com/chapter/10.1007/978-3-032-02725-2_1</a>

DOI - Digital Object Identifier

<a href="http://dx.doi.org/10.1007/978-3-032-02725-2" target="_blank" >10.1007/978-3-032-02725-2</a>

Jazyk výsledku

angličtina

Název v původním jazyce

Power Quality 24-h Prediction Based on L-Transform Derivative Modular and Deep Lear-ning Statistics Using Environmental Data in Detached Smart Buildings

Popis výsledku v původním jazyce

Verification of power quality (PQ) in autonomous grid systems is essential to prepare credible daily plans for the use of renewable energy (RE) and consequent load scheduling considering the source potential and user needs. Various combinations of connected household equipment and unexpected parameter changes require the evaluation of specific operational modes and incidents to control undesirable variations in PQ and avoid eventual system malfunction. The complexity and variability of system states, determined by backup charges and load switching, require the application of artificial intelligence (AI) strategies to model load instabilities and RE power ramp events. The system specifics and problem uncertainty do not allow for the unique representation by numerical models in relation to possible undefined states with great changeability. Differential Learning (DfL) is a novel unconventional neuro-computing strategy that composes step by step a modular-based derivative model that enables computing of next states for high alterative indefinable physical and electrical systems. Training data are first searched for optimal day intervals using AI testing evaluation. The proper assessment of initialisation times allows one to evolve AI statistics models for computable next-day PQ-series by processing the latest 24-h data in the final sequenced procedure. The 2-step PQ-assessment management verifies the preliminary composed load scheduling plans according to the availability of the RE sources and backup charge. The effectiveness of the system is guaranteed without failures in the operation of the system according to the first 24 h forecast of photovoltaic (PV) and wind power supply. Secondary PQ check-up handles the user-adapted algorithmically generated load utilisation schemes by evaluating uncertain state charges and RE production variational capacity. Algorithmic scheduling of the switching load in realistic PQ scenarios and reevaluating changes in available PV power resources are necessary in operational planning and control. Optimisation of system parameters is determined by regular early morning PQ verification and adequate RE use. Parametric C++ software with historical PQ &amp; meteo data archives is freely available for additional comparative experiments with the results presented and the deep and statistical learning model.

Název v anglickém jazyce

Power Quality 24-h Prediction Based on L-Transform Derivative Modular and Deep Lear-ning Statistics Using Environmental Data in Detached Smart Buildings

Popis výsledku anglicky

Verification of power quality (PQ) in autonomous grid systems is essential to prepare credible daily plans for the use of renewable energy (RE) and consequent load scheduling considering the source potential and user needs. Various combinations of connected household equipment and unexpected parameter changes require the evaluation of specific operational modes and incidents to control undesirable variations in PQ and avoid eventual system malfunction. The complexity and variability of system states, determined by backup charges and load switching, require the application of artificial intelligence (AI) strategies to model load instabilities and RE power ramp events. The system specifics and problem uncertainty do not allow for the unique representation by numerical models in relation to possible undefined states with great changeability. Differential Learning (DfL) is a novel unconventional neuro-computing strategy that composes step by step a modular-based derivative model that enables computing of next states for high alterative indefinable physical and electrical systems. Training data are first searched for optimal day intervals using AI testing evaluation. The proper assessment of initialisation times allows one to evolve AI statistics models for computable next-day PQ-series by processing the latest 24-h data in the final sequenced procedure. The 2-step PQ-assessment management verifies the preliminary composed load scheduling plans according to the availability of the RE sources and backup charge. The effectiveness of the system is guaranteed without failures in the operation of the system according to the first 24 h forecast of photovoltaic (PV) and wind power supply. Secondary PQ check-up handles the user-adapted algorithmically generated load utilisation schemes by evaluating uncertain state charges and RE production variational capacity. Algorithmic scheduling of the switching load in realistic PQ scenarios and reevaluating changes in available PV power resources are necessary in operational planning and control. Optimisation of system parameters is determined by regular early morning PQ verification and adequate RE use. Parametric C++ software with historical PQ &amp; meteo data archives is freely available for additional comparative experiments with the results presented and the deep and statistical learning model.

Druh

D - Stať ve sborníku

CEP obor

—

OECD FORD obor

10201 - Computer sciences, information science, bioinformathics (hardware development to be 2.2, social aspect to be 5.8)

Projekt

—

Návaznosti

S - Specificky vyzkum na vysokych skolach

Rok uplatnění

2025

Kód důvěrnosti údajů

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

Název statě ve sborníku

Lecture Notes in Computer Science. Volume 16008

ISBN

978-3-032-02724-5

ISSN

0302-9743

e-ISSN

1611-3349

Počet stran výsledku

10

Strana od-do

3-13

Název nakladatele

Springer

Místo vydání

Cham

Místo konání akce

A Coruña

Datum konání akce

16. 6. 2025

Typ akce podle státní příslušnosti

WRD - Celosvětová akce

Kód UT WoS článku

—