Monitoring electrical substation networks

1. An apparatus, comprising a device to forecast performance for a noninstrumented substation, comprising:

a storage device comprising a historic database comprising historic data for instrumented substations;

the storage device comprising a metadata database comprising metadata for the instrumented substations and for noninstrumented substations;

an unsupervised machine learning manager to generate a cluster of substations by metadata, wherein at least a portion of the substations in the cluster are noninstrumented substations; and

a supervised machine learning manager to generate a forecast for a noninstrumented substation in the cluster from historic data generated from the instrumented substations,

wherein generating the forecast comprises:

creating a vector of features, wherein the vector of features comprises a binary vector of clustering features and a binary vector of forecasting features:

simulating a predictive performance across all substations; and

determining if termination criteria have been met to terminate the simulation.

2. The apparatus of claim 1 , wherein the historic data comprises active power (P), reactive power (Q), voltage, apparent power (S), or voltage-ampere reactive (VAR), or any combinations thereof.

3. The apparatus of claim 1 , wherein the device comprises an auxiliary data source manager to retrieve data from an external database in a cloud, an internet-of-things (IoT) device, or a sensor, or any combinations thereof.

4. The apparatus of claim 3 , wherein the external database comprises weather data, metadata for substations, or scheduling data, or any combinations thereof.

5. The apparatus of claim 3 , wherein the IoT device comprises a traffic monitoring device, a weather station, or a temperature sensor, or any combinations thereof.

6. The apparatus of claim 1 , wherein the device comprises a performance simulator to simulate forecasts of deployment strategies for instrumenting substations.

7. The apparatus of claim 6 , wherein the performance simulator creates a graph of deployment cost versus forecasting performance.

8. A method for forecasting data for a noninstrumented substation, comprising:

accessing metadata for a plurality of substations;

grouping the plurality of substations into a plurality of dusters, based, at least in part, on the metadata; and

assigning the cluster id to each of the plurality of dusters;

determining a cluster id for the noninstrumented substation;

creating a model from historic data for instrumented substations having the duster id; and

forecasting the data for the noninstrumentcd substation from the model;

creating a vector of features, wherein the vector of features comprises a binary vector of clustering features and a binary vector of forecasting features;

simulating a predictive performance across all substations; and determining if termination criteria have been met.

9. The method of claim 8 , comprising measuring historic data for the instrumented substations, wherein the historic data comprises active power (P), reactive power (Q), voltage, apparent power (S), or voltage-ampere reactive (VAR), or any combinations thereof.

10. The method of claim 8 , comprising forecasting an equipment failure in a noninstrumentcd substation, based, at least in part, on values measured at the instrumented substations for reactive power (Q), voltage, apparent power (S), or voltage-ampere reactive (VAR), or any combinations thereof.

11. The method of claim 8 , wherein grouping the plurality of substations into the plurality of clusters comprises:

selecting metadata features related to power consumption; and

performing a clustering algorithm to create the clusters.

12. The method of claim 11 , wherein the metadata features comprise average power, average inter-day correlation, area load classifications, work schedules, regional energy tariffs, weather data, seasonal data, or traffic patterns, or any combinations thereof.

13. The method of claim 8 , comprising:

determine if the predictive performance has improved; and, if so,

copy a current binary vector of clustering features into a best feature store;

select a random element in the current binary vector of clustering features and Inver a feature bit; and

rerun the simulation.

14. The method of claim 8 , wherein creating the model comprises:

aggregating the data for all of the instrumented substations having the cluster id; and

creating the model from the aggregated data.

15. The method of claim 8 , wherein creating the model comprises:

creating a model for each instrumented substation having the cluster id; and

performing an individual forecast for the noninstrumented substation using each model.

16. The method of claim 15 , Wherein forecasting the data comprises aggregating the individual forecast for each noninstrumented substation to form an aggregated forecast.

17. The method of claim 8 , comprising generating a forecast for an instrumented substation.

18. The method of claim 17 , comprising:

creating a model for the instrumented substation from the historic data; and

generating the forecast.

19. The method of claim 8 , comprising determining a deployment strategy for deploying instrumentation to a portion of the plurality of substations.

20. The method of claim 19 comprising:

generating initial state features;

iterating state features;

evaluating average substation performance;

determining if termination criteria have been met; and

generating implementation plan.

21. The method of claim 20 , wherein the implementation plan comprises a graph of average substation performance versus deployment cost.

22. A non-transitory, machine readable medium, comprising instructions, which when executed, direct a processor to:

obtain metadata for a plurality of substations comprising both instrumented and noninstrumented substations;

create a cluster of substations based on the metadata;

obtain historic data for the instrumented substations;

generate a forecast for a noninstrumented substation;

create a vector of features, wherein the vector of features comprises a binary vector of clustering features and a binary vector of forecasting features;

simulate a predictive performance across the plurality of substations; and

determining if termination criteria have been met.

23. The non-transitory, machine readable medium of claim 22 , comprising instructions, which when executed, direct the processor to:

simulate performance for different distributions of instrumentation across the plurality of substations;

determine if termination criteria are met; and

create a graph of performance versus deployment cost for the different, distributions of instrumentation.