Devices, systems and methods for predicting future consumption values of load(s) in power distribution systems

1. A method for predicting energy usage of a load associated with a meter in an electrical distribution system comprising:

building a prediction model based on historical channel data for at least one meter and historical weather data of a location of the at least one meter;

receiving a request for a load prediction for a load associated with the at least one meter;

extracting time series data from the at least one meter;

retrieving future weather conditions for a particular location of the load associated the at least one meter; and

providing the extracted data and the future weather conditions to the prediction model that outputs a predicted load usage;

building a hybrid monte carlo model by extracting fifteen minute energy periods of data from the historical channel data, aggregating and averaging one hour energy periods of data from the historical channel data, subtracting the fifteen minute energy periods of data from the aggregated hourly data and performing a hybrid monte carlo algorithm on the subtracted data;

generating a stochastic 15 minute sequence on the hybrid monte carlo model;

grouping the data output from the sequence and determining the maximum per hour;

adding the determined maximum per hour to the predicted load usage to predict a peak demand for the load; and

performing at least one action based on the predicted peak demand, wherein the action includes outputting at least one control signal to at least client or the at least one meter to shut off the load associated with the at least one meter when the peak demand is above a predetermined threshold.

2. The method of claim 1 , wherein the performing of the at least one action includes outputting at least one communication signal to at least one client or the at least one meter.

3. The method as in claim 1 , wherein the request includes metadata for the at least one meter, the metadata includes at least one of a serial number, meter name, meter type, channel for meter history, facility location and/or user information.

4. The method as in claim 1 , wherein if the load usage prediction exceeds a predetermined threshold, the at least one action includes shutting down the load associated with at least one meter.

5. The method as in claim 1 , further comprising merging the historical channel data and historical weather data based on time and date.

6. The method as in claim 5 , further comprising performing a regression analysis on the merged data for each day of the week.

7. The method as in claim 6 , further comprising:

gradient boosting results of the regression analysis;

outputting the results of the gradient boosting to a time-series forecasting model; and

storing an output of the time-series forecasting model as the prediction model.

8. A system for predicting energy usage of a load associated with a meter in an electrical distribution system comprising:

an interface that receives a request for a load prediction for at least one meter;

at least one processor that builds a time series forecasting model based on historical channel data for the at least one meter and historical weather data of a location of the at least one meter;

the at least one processor extracts data from the at least meter from a time series database and retrieves future weather conditions for a particular location of the load associated with the at east one meter; and

the time series forecasting model receives the extracted data and the future weather conditions and outputs a predicted load usage,

wherein the at least one processor builds a hybrid monte carlo model by extracting fifteen minute energy periods of data from the historical channel data, aggregating and averaging one hour energy periods of data from the historical channel data, subtracting the fifteen minute energy periods of data from the aggregated hourly data and performing a hybrid monte carlo algorithm on the subtracted data,

the at least one processor generates a stochastic 15 minute sequence on the hybrid monte carlo model, groups the data output from the sequence, determines the maximum per hour and adds the determined maximum per hour to the predicted load usage to predict a peak demand for the load, and

the at least one processor performs at least one action based on the predicted peak demand, wherein the action includes outputting at least one control signal to at least client or the at least one meter to shut off the load associated with the at least one meter when the peak demand is above a predetermined threshold.

9. The system as in claim 8 , wherein the at least one action includes outputting at least one communication signal to at east one client or the at east one meter.

10. The system as in claim 8 , wherein the request includes metadata for the at least one meter, the metadata includes at east one of a serial number, meter name, meter type, channel for meter history, facility location and/or user information.

11. The system as in claim 8 , wherein if the load usage prediction exceeds a predetermined threshold, the at least one action includes shutting down the load associated with the at least one meter.

12. The system as in claim 8 , wherein the at least one processor merges the historical channel data and historical weather data based on time and date.

13. The system as in claim 12 , wherein the at least on processor performs a regression analysis on the merged data for each day of the week.

14. The system as in claim 13 , wherein the at least one processor gradient boosts results of the regression analysis, outputs the results of the gradient boosting to the time-series forecasting model and stores an output of the time-series forecasting model as the prediction model.

15. The method as in claim 1 , wherein the building the prediction model employs elastic net regularization.

16. The method as in claim 6 , wherein the regression analysis includes at least an absolute shrinkage and selection operator (LASSO) algorithm.

17. The method as in claim 7 , wherein the gradient boosting employs an autoregressive integrated moving average (ARIMA) model.

18. The system as in claim 8 , wherein the at least one processor builds the forecasting model using elastic net regularization.

19. The system as in claim 13 , wherein the regression analysis includes at least an absolute shrinkage and selection operator (LASSO) algorithm.

20. The system as in claim 14 , wherein the at least one processor gradient boosts using an autoregressive integrated moving average (ARIMA) model.