Development progress of power prediction robot and platform: Its world level very long term prototyping example

Authors : Burak Omer SARACOGLU

Pages : 253-289

Doi:10.30521/jes.1021838

View : 12 | Download : 11

Publication Date : 2022-06-30

Article Type : Research Paper

Abstract :Global Power Prediction Systems prototype version 2021 is presented with its system decomposition, scope, geographical/administrative/power grid decompositions, and similar. “Welcome”, “sign-up”, “log-in”, and “non-registered user main” web-interfaces are designed as draft on Quant UX. Map canvas is given as world political map with/without world power grid layers on QGIS 3.16.7-Hannover. Data input file is prepared based on several sources insert ignore into journalissuearticles values(1971-2018);. It includes minimum and maximum values due to source value differences. 70/30 principle is applied for train/test splitting insert ignore into journalissuearticles values(training/testing sets: 1971-2003/2004-2018);. 10 models are prepared on R version 4.1.1 with RStudio 2021.09.0+351. These are R::baseinsert ignore into journalissuearticles values(lm);, R::baseinsert ignore into journalissuearticles values(glm);, R::tidymodels::parsnipinsert ignore into journalissuearticles values(engineinsert ignore into journalissuearticles values(`lm`););, R::tidymodels::parsnipinsert ignore into journalissuearticles values(engineinsert ignore into journalissuearticles values(`glmnet`);); with lasso regularization, R::tidymodels::parsnipinsert ignore into journalissuearticles values(engineinsert ignore into journalissuearticles values(`glmnet`);); with ridge regularization, R::forecastinsert ignore into journalissuearticles values(auto.arima); auto autoregressive integrated moving average insert ignore into journalissuearticles values(ARIMA);, R::forecastinsert ignore into journalissuearticles values(arima); ARIMAinsert ignore into journalissuearticles values(1,1,2);, and ARIMAinsert ignore into journalissuearticles values(1,1,8);. Electricity demand in kilowatt-hours at the World level zone for up to 500-years insert ignore into journalissuearticles values(2019-2519); prediction period with only 1-year interval is forecasted. The best model is the auto ARIMA insert ignore into journalissuearticles values(mean absolute percentage error MAPE and symmetric mean absolute percentage error SMAPE for minimum and maximum electricity consumption respectively 1,1652; 6,6471; 1,1622; 6,9043);. Ex-post and ex-ante plots with 80%-95% confidence intervals are prepared in R::tidyverse::ggplot2. There are 3 alternative scripts insert ignore into journalissuearticles values(long, short, RStudio Cloud);. Their respective runtimes are 41,45; 25,44; and 43,33 seconds. Ex-ante 500-year period insert ignore into journalissuearticles values(2019-2519); is indicative and informative.
Keywords : Global power prediction system, Platform, Power, Prototyping, Robot