Orchestra is the innovative product for the prediction of hourly energy production up to 36 hours in advance. It can currently be applied to both wind and photovoltaic power plants. Orchestra takes its name from the innovative architecture of its artificial intelligence model: it uses a multitude of sub-models (the 'musicians' of Orchestra) that provide preliminary estimates of the amount of energy produced to a neural network (the 'director' of Orchestra) that harmonizes them and produces the final estimate. Selecting the best combination of musicians (among autoregressive, meteorological and artificial intelligence models) and the conductor architecture took years of development, but finally led to results unmatched in the marketplace.
Input variables
Weather variables
Energy production (optional)
Musicians
Orchestral conductor
Physical models
Statistical models
Proprietary artificial intelligence
Output
Energy production estimates
metrics
Performance
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
PERFORMANCE
The power of Omnienergy's predictions
The higher quality of Orchestra's model predictions is evident from the graph below, which compares Orchestra's root mean squared error (RMSE) and normalized mean absolute error (NMAE) with those of one of its main competitors, for the same historical period, on a photovoltaic plant:
As the graph shows, Orchestra's 1- to 12-hour (t+1, t+12) forecasts result in errors that are at least two times smaller not only than the competitor's predictions, but also than its real-time (t+0) production estimates, which should be particularly accurate.
What we do
Orchestra's forecasts are 2 to 3 times more accurate than the real-time estimates of an established competitor
competitivo
Performance
competitive
Gli errori nelle stime predittive sono più che dimezzati rispetto ai principali competitor di mercato, valutati tramite NMAE (Normalized Mean Absolute Error) rispetto alla potenza nominale e RMSE (Root Mean Squared Error).
Omnienergy si differenzia dai competitor per l’attenzione alle previsioni energetiche nel Mercato Intraday (MI), offrendo orizzonti di previsione che vanno da 1 a 12 ore con un costante aggiornamento orario.
At each forecast horizon (12 horizons for the Intraday Market, 24 for the Day-Ahead Market), the Conductor evaluates the forecast quality of the underlying models and dynamically assigns the weight of their contributions. Each musician belongs to one of the following types:
weather models, based on the physical link between weather state and energy production
autoregressive models, which estimate future energy production from past energy production
artificial intelligence models, selected and developed by Omnienergy to extract energy production from a variety of variables
The importance of the Orchestra Conductor
Each musician in the Orchestra performs well in some scenarios (i.e., for certain facilities, times of day, and weather conditions) and less well in others. The Orchestra Conductor knows the strengths and weaknesses of each musician and can harmonize their predictions by drawing the best out of each. The chart below shows an example of this fact: individual musicians provide good estimates, but often some musicians overestimate energy production while others underestimate it. Orchestra manages to mediate these conflicting positions and produces the most accurate estimate.
What we do
Dynamic evaluation of the quality of musicians' predictions by the Orchestra Conductor
Example of effective energy production (kwh) and predicted by Orchestra (orchestra) and its musicians (M1 to M6). While individual musicians frequently tend to overestimate or underestimate energy production, Orchestra averages their opinions by producing a better estimate than those of individual musicians.
performance
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
KPI
The evaluation indicators of the model
To evaluate the performance of the Orchestra model, we chose to measure two key performance indicators (KPIs): NMAE (Normalized Mean Absolute Error) and RMSE (Root Mean Squared Error).The NMAE indicator allows the quantification of the average percentage error of Orchestra with respect to the maximum power output of the considered plant (rated power).The RMSE metric, on the other hand, allows quantification of the absolute difference between the output predicted by the model and the actual output: larger errors have a greater proportional impact on the RMSE indicator.
What we do
NMAE (Normalized Mean Absolute Error) to quantify the average percent error with respect to maximum deliverable output
RMSE (Root Mean Squared Error) to measure the absolute difference between output predicted by the model and actual output
Want to learn how to build websites like this?
Learn from award winning designers in live online classes. Each class is limited to 15 students. Save you're spot below!
