Chiller and air handler control using customizable artificial intelligence system
Methods and systems are disclosed for determining a plan to optimize key performance indicators (KPIs) of an industrial process using a trained artificial intelligence agent and a custom objective function determined based on process information from the industrial system.
1 . A method, comprising:
determining a set of key performance indicators (KPIs) associated with an industrial process comprising a set of cooling equipment, wherein the set of cooling equipment comprises a chiller;
receiving process information for the industrial process that are associated with the set of KPIs from a user, wherein the process information comprises a set of semantic human-readable variable names and a mapping of the variable names to a set of industrial process data streams;
generating a custom objective function based on the process information;
using a reinforcement learning artificial intelligence agent comprising a set of hidden neural network layers, predicting a set of setpoints that optimize the set of KPIs based on the custom objective function and sensor data from the set of cooling equipment, wherein the sensor data comprise chiller data from the chiller; and
controlling the set of cooling equipment based on the set of setpoints, comprising selectively activating or deactivating the chiller.
2 . The method of claim 1 , wherein determining the set of setpoints comprises: predicting a future state of the industrial process based on the sensor data; and determining the set of setpoints based on the future state of the industrial process.
3 . The method of claim 1 , wherein the reinforcement learning artificial intelligence agent is trained using the custom objective function.
4 . The method of claim 1 , wherein the sensor data comprises historical sensor data and real-time sensor data from the set of cooling equipment components.
5 . The method of claim 1 , wherein the process information comprises a set of process variables associated with the custom objective function, wherein the set of setpoints are further determined based on a relationship between the set of process variables and the set of setpoints.
6 . The method of claim 1 , wherein the industrial process comprises at least one of a chemical manufacturing process, a heating and cooling process, a pharmaceutical manufacturing process, an oil and gas manufacturing process, a paper and pulp manufacturing process, a metal manufacturing process, a cement manufacturing process, a glass manufacturing process, or a power generation process.
7 . The method of claim 1 , wherein the reinforcement learning artificial intelligence agent (RL AI agent) further comprises an exploration module configured to explore setpoint values outside of a training data set for the RL AI agent.
8 . The method of claim 1 , wherein the process information is received from a user using interactive hierarchical querying.
9 . A system, comprising:
a processing system configured to:
determine a set of key performance indicators (KPIs);
determine process information associated with the set of KPIs from a user using interactive querying, wherein the process information comprises a set of semantic human-readable variable names and a mapping of the variable names to a set of industrial process data streams; generate a custom objective function for the set of KPIs based on the process information;
using a trained artificial intelligence agent, determine a plan for optimizing the set of KPIs based on the custom objective function and sensor data from a set of equipment, the set of equipment comprising an air handler and wherein the plan comprises a set of process variable setpoints for the air handler; and
control the set of equipment based on the plan, comprising controlling the air handler based on the set of process variable setpoints.
10 . The system of claim 9 , wherein the artificial intelligence agent is trained using the custom objective function.
11 . The system of claim 9 , wherein the process information comprises at least one of: a set of process variables associated with the custom objective function, a set of constraints associated with the set of process variables, or a dependency between process variables in the set of process variables.
12 . The system of claim 9 , wherein the process information comprises a mapping between a set of elements of the objective function and the sensor data.
13 . The method of claim 9 , wherein the process information comprises a set of responses to a set of queries.
14 . The system of claim 9 , wherein the artificial intelligence agent comprises a reinforcement learning agent.
15 . The system of claim 9 , wherein the equipment further comprises at least one of: a chiller, a pump, a fan, or a cooling tower.
16 . The system of claim 9 , wherein the set of KPIs comprises at least one of: energy usage, product quality, product yield, or process stability.
17 . A method, comprising:
generating an objective function associated with a set of key performance indicators (KPIs) based on a set of semantic human-readable variable names received from a user;
receiving a set of equipment data streams from a set of equipment of an industrial process, wherein the set of equipment comprises a chiller;
determining a mapping between the set of variable names and the set of equipment data streams from the user;
using a reinforcement learning artificial intelligence agent, determining a plan for optimizing the set of KPIs based on the objective function, the set of equipment data streams, and the mapping, wherein the plan comprises a set of chiller setpoints; and
controlling the set of equipment based on the plan, comprising controlling the chiller based on the set of chiller setpoints.
18 . The method of claim 17 , wherein a set of elements of the objective function comprise the semantic human-readable variable names.
19 . The method of claim 17 , wherein determining the plan comprises predicting a future state of the industrial process based on the set of equipment data streams, and determining the plan based on the future state of the industrial process.
20 . The method of claim 17 , further comprising iteratively training the reinforcement learning artificial intelligence agent using the objective function.
21 . The method of claim 17 , wherein the industrial process comprises at least one of a chemical manufacturing process, a heating and cooling process, a pharmaceutical manufacturing process, an oil and gas manufacturing process, a paper and pulp manufacturing process, a metal manufacturing process, a cement manufacturing process, a glass manufacturing process, or a power generation process.