Controlling devices using a rich representation of their environment

A method, implemented by one or more computing devices, a physical environment, comprising:

• storing environment information in one or more data stores that describes the physical environment to be controlled, the physical environment including plural local energy devices that generate energy, and/or consume energy, and/or store previously-supplied energy,
• the environment information including:
• first information which describes entities associated with the physical environment, the entities including at least:
• one or more physical spaces;
• one or more devices including said plural local energy devices;
• one or more sensors;
• and one or more people who interact with the physical environment;
• and
• second information which describes relations among the entities in the physical environment,
• the environment information being expressed as a data structure having nodes connected by links, the nodes representing the entities, and the links representing the relations among the entities;
• receiving sensor data from one or more sensors provided by the physical environment, the sensor data describing states associated with at least some of the entities in the physical environment;
• generating forecast data, using at least one machine-trained model, that describes a future state of at least one aspect of the physical environment, said generating of the forecast data being based at least on:
• the environment information;
• and the sensor data;
• generating a control plan that provides a strategy for controlling operation of the plural local energy devices and/or other entities in the physical environment that satisfies a specified optimization objective, within a specified span of time, said generating of the control plan being based at least on:
• a constraint file that describes constraints associated with the physical environment, as expressed by properties of the nodes of the data structure;
• an objective function that describes the specified optimization objective;
• the sensor data;
• and the forecast data;
• submitting proposal information to a resource-providing system that describes the control plan;
• receiving reply information from the resource-providing system in response to the proposal information;
• and
• sending instructions to the physical environment when the reply information indicates that the resource-providing system has approved the proposal information, the instructions for use in executing the control plan in the physical environment,
• wherein the resource-providing system is a distribution system for distributing power to the physical environment, and wherein the specified optimization objective includes minimizing a cost incurred by the physical environment in consuming power from the distribution system in a flexibility market hosted by a distribution system operator, and
• wherein said generating of the control plan includes modeling said plural local energy devices in the physical environment as an aggregate flexible energy resource, the aggregate flexible energy resource being controlled to achieve the specified optimization objective.

A computer-readable storage medium for storing computer-readable instructions, the computer-readable instructions, when executed by one or more hardware processors, performing a method that comprises:

• storing environment information in one or more data stores that describes a physical environment to be controlled, the physical environment including plural local energy devices that generate energy, and/or consume energy, and/or store previously-supplied energy,
• the environment information including:
• first information which describes entities associated with the physical environment, the entities including at least:
• one or more physical spaces;
• one or more devices including said plural local energy devices;
• one or more sensors;
• and one or more people who interact with the physical environment;
• and
• second information which describes relations among the entities in the physical environment,
• the environment information being expressed as a data structure having nodes connected by links, the nodes representing the entities, and the links representing the relations among the entities;
• receiving sensor data from one or more sensors provided by the physical environment, the current sensor data describing states associated with at least some of the entities in the physical environment;
• generating forecast data, using at least one machine-trained model, that describes a future state of at least one aspect of the physical environment, said generating of the forecast data being based at least on:
• the environment information;
• and the sensor data;
• and
• generating a control plan that provides a strategy for controlling operation of the plural local energy devices and/or other entities in the physical environment that satisfies a specified optimization objective, within a specified span of time, said generating of the control plan being based at least on:
• a constraint file that describes constraints associated with the physical environment, as expressed by properties of the nodes of the data structure;
• an objective function that describes the specified optimization objective;
• the sensor data;
• and the forecast data,
• wherein the specified optimization objective includes minimizing a cost incurred by the physical environment in consuming power in a flexibility market hosted by a distribution system operator, and
• wherein said generating of the control plan includes modeling said plural local energy devices in the physical environment as a flexible aggregate energy resource, the aggregate flexible energy resource being controlled to achieve the specified optimization objective.