Autonomous electrical grid management

1. A system for controlling devices connected to an electrical grid, the system comprising:

a manager subsystem that is configured to perform operations comprising, at each time point in a sequence of multiple time points:

autonomously determining whether each control strategy update criterion in a set of control strategy update criteria is satisfied at the time point, wherein the set of control strategy update criteria comprises at least one of:

a first control strategy update criterion that is satisfied if a measured change in an electrical load on the grid exceeds a threshold amount in a time period prior to the time point; or

a second control strategy update criterion that depends on a magnitude of the electrical load on the grid at the time point;

and in response to determining that a control strategy update criterion in the set of control strategy update criteria is satisfied at a time point, updating controls applied to devices connected to the electrical grid, comprising:

identifying: (i) a plurality of devices connected to the electrical grid, and (ii) one or more objectives to be satisfied by the plurality of devices, wherein:

the one or more objectives specify a target load curve that defines a target for an aggregated load generated by the plurality of devices; and

the target load curve: (i) is greater than a baseline load curve over a first target time period, and (ii) is less than the baseline load curve over a second target time period; and

the baseline load curve defines an aggregate load generated by the plurality of devices if the plurality of devices are not controlled to satisfy the one or more objectives;

obtaining current monitoring data that characterizes a current operational state of the electrical grid and of the plurality of devices connected to the electrical grid as of the time point;

providing, to an optimization subsystem, the current monitoring data and data identifying: (i) the plurality of devices connected to the electrical grid, and (ii) the one or more objectives to be satisfied by the plurality of devices, wherein the optimization subsystem identifies a respective control to be applied to each of one or more of the devices to achieve a target operational state of the grid, and provides data identifying the respective control to be applied to each of one or more of the devices to the manager subsystem;

receiving, from the optimization subsystem, the data identifying the respective control to be applied to each of one or more of the devices to achieve the target operational state of the grid; and

transmitting, to a dispatcher subsystem, instructions to cause the dispatcher subsystem to instruct the respective control to be applied to each of one or more of the devices;

the optimization subsystem that is configured to perform operations comprising:

receiving the current monitoring data the data identifying: (i) the plurality of devices connected to the electrical grid, and (ii) the one or more objectives to be satisfied by the plurality of devices;

for each of one or more devices of the plurality of devices:

generating, using the current monitoring data and a model for the device and for a plurality of controls, a plurality of simulated load curves that simulate operation of the device based on application of a respective control of the plurality of controls;

identifying, by an optimization procedure and based on at least the plurality of simulated load curves for a respective device, a respective control to be applied to the respective device of the plurality of devices to satisfy the objectives; and

the dispatcher subsystem that is configured to perform operations comprising:

transmitting instructions to cause the respective control to be applied to each of one or more devices of the plurality of devices.

2. The system of claim 1 , wherein generating the plurality of simulated load curves that simulate operation of the device for each device of the plurality of devices based on application of the respective control of the plurality of controls comprises:

simulating operation of the plurality of devices based on application of the plurality of controls conditioned on the current monitoring data characterizing the current operational state of the grid and of the plurality of devices.

3. The system of claim 1 , wherein identifying the respective control to be applied to the respective device of the plurality of devices to satisfy the objectives comprises:

adjusting values of a plurality of weights during a plurality of optimization iterations using the optimization procedure to optimize a loss function, wherein each weight corresponds to a respective simulated load curve, wherein the loss function measures:

a sparsity of the values of the plurality of weights, and

an error between: (i) an aggregate load curve that is defined by combining the simulated load curves in accordance with the values of the plurality of weights, and (ii) the target load curve; and

identifying the respective control to be applied to each of the plurality of devices based on the values of the plurality of weights after a final optimization iteration.

4. The system of claim 1 , wherein receiving the current monitoring data and the data identifying: (i) the plurality of devices connected to the electrical grid, and (ii) the one or more objectives to be satisfied by the plurality of devices comprises:

receiving data identifying a set of multiple objectives to be satisfied by the plurality of devices, wherein the set of multiple objectives comprises a first objective that conflicts with a second objective.

5. The system of claim 4 , wherein the first objective defines a target operational state of the grid over a time interval, and the second objective defines a different target operational state of the grid over the time interval.

6. The system of claim 5 , wherein the optimization subsystem is further configured to perform operations comprising:

receiving, for each objective in the set of objectives, a prioritization score for the objective that defines a priority of the objective; and

determining a combined objective over the time interval based on the prioritization scores for the first objective and the second objective;

wherein identifying the respective control to be applied to each of one or more devices of the plurality of devices to satisfy the objectives comprises:

identifying a respective control to be applied to each of one or more devices of the plurality of devices to satisfy the combined objective over the time interval.

7. The system of claim 6 , wherein the first objective defines a first target load curve over the time interval, and the second objective defines a second, different target load curve over the time interval; and

wherein determining the combined objective over the time interval based on the prioritization scores for the first objective and the second objective comprises:

determining respective scaling factors for the first objective and the second objective based on their respective prioritization scores; and

determining the combined objective over the time interval as a combination of: (i) the first target load curve scaled by the scaling factor for the first objective, and (ii) the second target load curve scaled by the scaling factor for the second objective.

8. The system of claim 1 , wherein the plurality of devices connected to the grid comprises an energy storage device; and

wherein transmitting instructions to cause the respective control be applied to each of one or more devices of the plurality of devices comprises:

transmitting instructions to the energy storage device to inject energy into the grid.

9. The system of claim 8 , wherein the energy storage device comprises a battery.

10. The system of claim 1 , wherein the set of control strategy update criteria further comprises:

a third control strategy update criterion that is satisfied if a new objective is added to the one or more objectives to be satisfied by the plurality of devices; and

a fourth control strategy update criterion that is satisfied if an objective is removed from the one or more objectives to be satisfied by the plurality of devices.

11. The system of claim 1 , wherein the second target time period is before the first target time period.

12. A method performed by one or more data processing apparatus for controlling devices connected to an electrical grid, the method comprising:

autonomously determining, at each time point in a sequence of multiple time points, whether each control strategy update criterion in a set of control strategy update criteria is satisfied at the time point, wherein the set of control strategy update criteria comprises at least one of:

a first control strategy update criterion that is satisfied if a measured change in an electrical load on the grid exceeds a threshold amount in a time period prior to the time point; or

a second control strategy update criterion that depends on a magnitude of the electrical load on the grid at the time point;

and in response to determining that a control strategy update criterion in the set of control strategy update criteria is satisfied at a time point, updating controls applied to devices connected to the electrical grid, comprising:

identifying: (i) a plurality of devices connected to the electrical grid, and (ii) one or more objectives to be satisfied by the plurality of devices, wherein:

the one or more objectives specify a target load curve that defines a target for an aggregated load generated by the plurality of devices; and

the target load curve: (i) exceeds a baseline load curve over a first target time period, and (ii) does not exceed the baseline load curve over a second target time period; and

the baseline load curve defines an aggregate load generated by the plurality of devices if the plurality of devices are not controlled to satisfy the one or more objectives;

obtaining current monitoring data that characterizes a current operational state of the electrical grid and of the plurality of devices connected to the electrical grid as of the time point;

processing data identifying: (i) the plurality of devices connected to the electrical grid, and (ii) the one or more objectives to be satisfied by the plurality of devices, to identify a respective control to be applied to each of one or more of the devices to achieve a target operational state of the grid, comprising:

for each of one or more devices of the plurality of devices:

generating, using the current monitoring data and a model for the device and for a plurality of controls, a plurality of simulated load curves that simulate operation of the device based on application of a respective control of the plurality of controls;

identifying, by an optimization procedure and based on at least the plurality of simulated load curves for a respective device, a respective control to be applied to the respective device of the plurality of devices to satisfy the objectives;

transmitting instructions to cause the respective control to be applied to each of one or more of the devices of the plurality of devices.

13. The method of claim 12 , wherein generating the plurality of simulated load curves that simulate operation of the device for each device of the plurality of devices based on application of the respective control of the plurality of controls comprises:

simulating operation of the plurality of devices based on application of the plurality of controls conditioned on the current monitoring data characterizing the current operational state of the grid and of the plurality of devices.

14. The method of claim 12 , wherein identifying the respective control to be applied to the respective device of the plurality of devices to satisfy the objectives comprises:

adjusting values of a plurality of weights during a plurality of optimization iterations using the optimization procedure to optimize a loss function, wherein each weight corresponds to a respective simulated load curve, wherein the loss function measures:

a sparsity of the values of the plurality of weights, and

an error between: (i) an aggregate load curve that is defined by combining the simulated load curves in accordance with the values of the plurality of weights, and (ii) the target load curve; and

identifying the respective control to be applied to each of the plurality of devices based on the values of the plurality of weights after a final optimization iteration.

15. A method performed by one or more data processing apparatus for controlling devices connected to an electrical grid, the method comprising:

autonomously determining, at each time point in a sequence of multiple time points, whether each control strategy update criterion in a set of control strategy update criteria is satisfied at the time point, wherein the set of control strategy update criteria comprises at least one of:

a first control strategy update criterion that is satisfied if a measured change in an electrical load on the grid exceeds a threshold amount in a time period prior to the time point; or

a second control strategy update criterion that depends on a magnitude of the electrical load on the grid at the time point;

and in response to determining that a control strategy update criterion in the set of control strategy update criteria is satisfied at a time point, updating controls applied to devices connected to the electrical grid, comprising:

identifying: (i) a plurality of devices connected to the electrical grid, and (ii) a set of multiple objectives to be satisfied by the plurality of devices;

obtaining current monitoring data that characterizes a current operational state of the electrical grid and of the plurality of devices connected to the electrical grid as of the time point;

processing data identifying: (i) the plurality of devices connected to the electrical grid, and (ii) the set of multiple objectives to be satisfied by the plurality of devices, to identify a respective control to be applied to each of one or more of the devices to achieve the set of multiple objectives, comprising:

for each of one or more devices of the plurality of devices, generating, using the current monitoring data and a model for the device and for a plurality of controls, a plurality of simulated load curves that simulate operation of the device based on application of a respective control of the plurality of controls;

identifying, by an optimization procedure and based on at least the plurality of simulated load curves for a respective device, a respective control to be applied to the respective device of the plurality of devices to satisfy the objectives, comprising:

determining that the set of multiple objectives comprises a first objective that conflicts with a second objective, wherein the first objective and the second objective define conflicting target operational states of the grid over a time interval;

obtaining, for each of the first objective and the second objective, a respective prioritization score for the objective that defines a priority of the objective;

determining a combined objective over the time interval based on the prioritization scores for the first objective and the second objective; and

identifying, by the optimization procedure, a respective control to be applied to each of one or more devices of the plurality of devices to satisfy the combined objective over the time interval; and

transmitting instructions to cause the respective control to be applied to each of one or more of the devices of the plurality of devices.

16. The method of claim 15 , wherein the first objective defines a first target load curve over the time interval, and the second objective defines a second, different target load curve over the time interval; and

wherein determining the combined objective over the time interval based on the prioritization scores for the first objective and the second objective comprises:

determining respective scaling factors for the first objective and the second objective based on their respective prioritization scores; and

determining the combined objective over the time interval as a combination of: (i) the first target load curve scaled by the scaling factor for the first objective, and (ii) the second target load curve scaled by the scaling factor for the second objective.

17. The method of claim 15 , wherein the set of control strategy update criteria comprises a third control strategy update criterion that is satisfied if a new objective is added to the one or more objectives to be satisfied by the plurality of devices.

18. The method of claim 15 , wherein the set of control strategy update criteria comprises a fourth control strategy update criterion that is satisfied if an objective is removed from the one or more objectives to be satisfied by the plurality of devices.

19. The method of claim 15 , wherein generating the plurality of simulated load curves that simulate operation of the device for each device of the plurality of devices based on application of the respective control of the plurality of controls comprises:

simulating operation of the plurality of devices based on application of the plurality of controls conditioned on the current monitoring data characterizing the current operational state of the grid and of the plurality of devices.

20. The method of claim 15 , wherein identifying the respective control to be applied to the respective device of the plurality of devices to satisfy the objectives comprises:

adjusting values of a plurality of weights during a plurality of optimization iterations using the optimization procedure to optimize a loss function, wherein each weight corresponds to a respective simulated load curve, wherein the loss function measures:

a sparsity of the values of the plurality of weights, and

an error between: (i) an aggregate load curve that is defined by combining the simulated load curves in accordance with the values of the plurality of weights, and (ii) a target load curve; and

identifying the respective control to be applied to each of the plurality of devices based on the values of the plurality of weights after a final optimization iteration.