Search NASA⌕ Search

SEARCH · Search NASA

Results for “Control”

Search indexed NASA NTRS and DOE OSTI research on propulsion, heat transfer, battery materials and energy systems. Follow report and document links to the original sources.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 1,639 records · Page 91
All sourcesNASA NTRSDOE OSTI
NASA NTRS2010

Simulation Training Versus Real Time Console Training for New Flight Controllers

For new flight controllers, the two main learning tools are simulations and real time console performance training. These benefit the new flight controllers in different ways and could possibly be improved. Simulations: a) Allow for mistakes without serious consequences. b) Lets new flight controllers learn the working style of other new flight controllers. c) Lets new flight controllers eventually begin to feel like they have mastered the sim world, so therefore they must be competent in the real time world too. Real time: a) Shows new flight controllers some of the unique problems that develop and have to be accounted for when dealing with certain payloads or systems. b) Lets new flight controllers experience handovers - gathering information from the previous shift on what the room needs to be aware of and what still needs to be done. c) Gives new flight controllers confidence that they can succeed in the position they are training for when they can solve real anomalies. How Sims could be improved and more like real-time ops for the ISS Operations Controller position: a) Operations Change Requests to review. b) Fewer anomalies (but still more than real time for practice). c) Payload Planning Manager Handover sheet for the E-1 and E-3 reviews. d) Flight note in system with at least one comment to verify for the E-1 and E-3 reviews How the real time console performance training could be improved for the ISS Operations Controller position: a) Schedule the new flight controller to be on console for four days but with a different certified person each day. This will force them to be the source of knowledge about every OCR in progress, everything that has happened in those few days, and every activity on the timeline. Constellation program flight controllers will have to learn entirely from simulations, thereby losing some of the elements that they will need to have experience with for real time ops. It may help them to practice real time console performance training in the International Space Station or Space Shuttle to gather some general anomaly resolution and day-to-day task management skills.

Heaton, Amanda↗
NASA NTRS2010

Data-Based Predictive Control with Multirate Prediction Step

Data-based predictive control is an emerging control method that stems from Model Predictive Control (MPC). MPC computes current control action based on a prediction of the system output a number of time steps into the future and is generally derived from a known model of the system. Data-based predictive control has the advantage of deriving predictive models and controller gains from input-output data. Thus, a controller can be designed from the outputs of complex simulation code or a physical system where no explicit model exists. If the output data happens to be corrupted by periodic disturbances, the designed controller will also have the built-in ability to reject these disturbances without the need to know them. When data-based predictive control is implemented online, it becomes a version of adaptive control. One challenge of MPC is computational requirements increasing with prediction horizon length. This paper develops a closed-loop dynamic output feedback controller that minimizes a multi-step-ahead receding-horizon cost function with multirate prediction step. One result is a reduced influence of prediction horizon and the number of system outputs on the computational requirements of the controller. Another result is an emphasis on portions of the prediction window that are sampled more frequently. A third result is the ability to include more outputs in the feedback path than in the cost function.

Barlow, Jonathan S.↗
NASA NTRS2005

Controlling Force and Depth in Friction Stir Welding

Feedback control of the penetration force applied to a pin tool in friction stir welding has been found to be a robust and reliable means for controlling the depth of penetration of the tool. This discovery has made it possible to simplify depth control and to weld with greater repeatability, even on workpieces with long weld joints. Prior to this discovery, depths of penetration in friction stir welding were controlled by hard-tooled roller assemblies or by depth actuators controlled by feedback from such external sensors as linear variable-differential transformers or laser-based devices. These means of control are limited: A hard-tooled roller assembly confines a pin tool to a preset depth that cannot be changed easily during the welding process. A measurement by an external sensor is only an indirect indicative of the depth of penetration, and computations to correlate such a measurement with a depth of penetration are vulnerable to error. The present force-feedback approach exploits the proportionality between the depth and the force of penetration Unlike a depth measurement taken by an external sensor, a force measurement can be direct because it can be taken by a sensor coupled directly to the pin tool. The reading can be processed through a modern electronic servo control system to control an actuator to keep the applied penetration force at the desired level. In comparison with the older depth-control methods described above, this method offers greater sensitivity to plasticizing of the workpiece metal and is less sensitive to process noise, resulting in a more consistent process. In an experiment, a tapered panel was friction stir welded while controlling the force of penetration according to this method. The figure is a plot of measurements taken during the experiment, showing that force was controlled with a variation of 200 lb (890 N), resulting in control of the depth of penetration with a variation of 0.004 in. (0.1 mm).

Adams, Glynn↗
NASA NTRS2012

Intelligent Systems Approach for Automated Identification of Individual Control Behavior of a Human Operator

Results have been obtained using conventional techniques to model the generic human operator?s control behavior, however little research has been done to identify an individual based on control behavior. The hypothesis investigated is that different operators exhibit different control behavior when performing a given control task. Two enhancements to existing human operator models, which allow personalization of the modeled control behavior, are presented. One enhancement accounts for the testing control signals, which are introduced by an operator for more accurate control of the system and/or to adjust the control strategy. This uses the Artificial Neural Network which can be fine-tuned to model the testing control. Another enhancement takes the form of an equiripple filter which conditions the control system power spectrum. A novel automated parameter identification technique was developed to facilitate the identification process of the parameters of the selected models. This utilizes a Genetic Algorithm based optimization engine called the Bit-Climbing Algorithm. Enhancements were validated using experimental data obtained from three different sources: the Manual Control Laboratory software experiments, Unmanned Aerial Vehicle simulation, and NASA Langley Research Center Visual Motion Simulator studies. This manuscript also addresses applying human operator models to evaluate the effectiveness of motion feedback when simulating actual pilot control behavior in a flight simulator.

Zaychik, Kirill B.↗
NASA NTRS2012

Interactive robot control system and method of use

A robotic system includes a robot having joints, actuators, and sensors, and a distributed controller. The controller includes command-level controller, embedded joint-level controllers each controlling a respective joint, and a joint coordination-level controller coordinating motion of the joints. A central data library (CDL) centralizes all control and feedback data, and a user interface displays a status of each joint, actuator, and sensor using the CDL. A parameterized action sequence has a hierarchy of linked events, and allows the control data to be modified in real time. A method of controlling the robot includes transmitting control data through the various levels of the controller, routing all control and feedback data to the CDL, and displaying status and operation of the robot using the CDL. The parameterized action sequences are generated for execution by the robot, and a hierarchy of linked events is created within the sequence.

Sanders, Adam M.↗
NASA NTRS2010

Control of Risks Through the Use of Procedures: A Method for Evaluating the Change in Risk

This paper considers how procedures can be used to control risks faced by an organization and proposes a means of recognizing if a particular procedure reduces risk or contributes to the organization's exposure. The proposed method was developed out of the review of work documents and the governing procedures performed in the wake of the Columbia accident by NASA and the Space Shuttle prime contractor, United Space Alliance, LLC. A technique was needed to understand the rules, or procedural controls, in place at the time in the context of how important the role of each rule was. The proposed method assesses procedural risks, the residual risk associated with a hazard after a procedure's influence is accounted for, by considering each clause of a procedure as a unique procedural control that may be beneficial or harmful. For procedural risks with consequences severe enough to threaten the survival of the organization, the method measures the characteristics of each risk on a scale that is an alternative to the traditional consequence/likelihood couple. The dual benefits of the substitute scales are that they eliminate both the need to quantify a relationship between different consequence types and the need for the extensive history a probabilistic risk assessment would require. Control Value is used as an analog for the consequence, where the value of a rule is based on how well the control reduces the severity of the consequence when operating successfully. This value is composed of two parts: the inevitability of the consequence in the absence of the control, and the opportunity to intervene before the consequence is realized. High value controls will be ones where there is minimal need for intervention but maximum opportunity to actively prevent the outcome. Failure Likelihood is used as the substitute for the conventional likelihood of the outcome. For procedural controls, a failure is considered to be any non-malicious violation of the rule, whether intended or not. The model used for describing the Failure Likelihood considers how well a task was established by evaluating that task on five components. The components selected to define a well established task are: that it be defined, assigned to someone capable, that they be trained appropriately, that the actions be organized to enable proper completion and that some form of independent monitoring be performed. Validation of the method was based on the information provided by a group of experts in Space Shuttle ground processing when they were presented with 5 scenarios that identified a clause from a procedure. For each scenario, they recorded their perception of how important the associated rule was and how likely it was to fail. They then rated the components of Control Value and Failure Likelihood for all the scenarios. The order in which each reviewer ranked the scenarios Control Value and Failure Likelihood was compared to the order in which they ranked the scenarios for each of the associated components; inevitability and opportunity for Control Value and definition, assignment, training, organization and monitoring for Failure Likelihood. This order comparison showed how the components contributed to a relative relationship to the substitute risk element. With the relationship established for Space Shuttle ground processing, this method can be used to gauge if the introduction or removal of a particular rule will increase or decrease the .risk associated with the hazard it is intended to control.

Praino, Gregory T.↗
NASA NTRS2015

SimSup's Loop: A Control Theory Approach to Spacecraft Operator Training

Immersive simulation is a staple of training for many complex system operators, including astronauts and ground operators of spacecraft. However, while much has been written about simulators, simulation facilities, and operator certification programs, the topic of how one develops simulation scenarios to train a spacecraft operator is relatively understated in the literature. In this paper, an approach is presented for using control theory as the basis for developing the immersive simulation scenarios for a spacecraft operator training program. The operator is effectively modeled as a high level controller of lower level hardware and software control loops that affect a select set of system state variables. Simulation scenarios are derived from a STAMP-based hazard analysis of the operator's high and low level control loops. The immersive simulation aspect of the overall training program is characterized by selecting a set of scenarios that expose the operator to the various inadequate control actions that stem from control flaws and inadequate control executions in the different sections of the typical control loop. Results from the application of this approach to the Lunar Atmosphere and Dust Environment Explorer (LADEE) mission are provided through an analysis of the simulation scenarios used for operator training and the actual anomalies that occurred during the mission. The simulation scenarios and inflight anomalies are mapped to specific control flaws and inadequate control executions in the different sections of the typical control loop to illustrate the characteristics of anomalies arising from the different sections of the typical control loop (and why it is important for operators to have exposure to these characteristics). Additionally, similarities between the simulation scenarios and inflight anomalies are highlighted to make the case that the simulation scenarios prepared the operators for the mission.

LADEE↗
NASA NTRS2016

Component Control System for a Vehicle

A vehicle includes a chassis, a modular component, and a central operating system. The modular component is supported by the chassis. The central operating system includes a component control system, a primary master controller, and a secondary master controller. The component control system is configured for controlling the modular component. The primary and secondary master controllers are in operative communication with the component control system. The primary and secondary master controllers are configured to simultaneously transmit commands to the component control system. The component control system is configured to accept commands from the secondary master controller only when a fault occurs in the primary master controller.

Fraser-Chanpong, Nathan↗
NASA NTRS2017

Dynamic Control System Mode Performance of the Space Technology-7 Disturbance Reduction System

The Space Technology-7 (ST-7) Disturbance Reduction System (DRS) is an experiment package aboard the European Space Agency (ESA) LISA Pathfinder spacecraft, launched on December 3, 2015. DRS consists of three primary components: Colloidal MicroNewton Thrusters (CMNTs), an Integrated Avionics Unit (IAU), and flight-software implementing the Command and Data Handling (C&DH) and Dynamic Control System (DCS) algorithms. The CMNTs were designed to provide thrust from 5 to 30 micro Newton, with thrust controllability and resolution of 0.1 micro Newton and thrust noise of 0.1 micro Newton/(square root of (Hz)) in the measurement band from 1-30 mHz. The IAU hosts the C&DH and DCS flight software, as well as interfaces with both the CMNT electronics and the LISA Pathfinder spacecraft. When in control, the DCS uses star tracker attitude data and capacitive or optically-measured position and attitude information from LISA Pathfinder and the LISA Technology Package (LTP) to control the attitude and position of the spacecraft and the two test masses inside the LTP. After completion of the nominal ESA LISA Pathfinder mission, the DRS experiment was commissioned followed by its nominal mission. DRS operations extended over the next five months, interspersed with station keeping, anomaly resolution, and periods where control was handed back to LISA Pathfinder for them to conduct further experiments. The primary DRS mission ended on December 6, 2016, with the experiment meeting all of its Level 1 requirements. The DCS, developed at the NASA Goddard Space Flight Center, consists of five spacecraft control modes and six test mass control modes, combined into six 'DRS Mission Modes'. Attitude Control and Zero-G were primarily used to control the spacecraft during initial handover and during many of the CMNT characterization experiments. The other Mission Modes, Drag Free Low Force, 18-DOF Transitional, and 18-DOF, were used to provide drag-free control of the spacecraft about the test masses. This paper will discuss the performance of these DCS spacecraft and test mass control modes. Flight data will be shown from each mode throughout the mission, both from nominal operations and during various flight experiments. The DCS team also made some changes to controller, filter, and limit parameters during operations; the motivation and results of these changes will be shown and discussed.

O'Donnell, James R., Jr.↗
NASA NTRS2020

Pterodactyl: Development and Comparison of Control Architectures for a Mechanically Deployed Entry Vehicle

The Pterodactyl project, seeks to advance the current state-of-the-art for entry vehicles by developing novel guidance and control technologies for Deployable Entry Vehicles (DEVs) that can be applied to various entry vehicle configurations. This paper details the efforts on the NASA-funded Pterodactyl project to investigate and implement multiple control techniques for an asymmetric mechanical DEV. We design multiple control architectures for a Pterodactyl Baseline Vehicle (PBV) and evaluate their performance in achieving varying guidance commands during entry. The control architectures studied are (i) propulsive control systems such as reaction control systems and (ii) non-propulsive control systems such as aerodynamic control surfaces and internal moving masses. For each system, state-feedback integral controllers based on linear quadratic regulator (LQR) optimal control methods are designed to track guidance commands of either (i) bank angle or (ii) angle of attack and sideslip angle as determined by the desired guidance trajectory. All control systems are compared for a lunar return reference mission and by providing a comparative analysis of these systems, configurations, and performance, the efforts detailed in this paper and the Pterodactyl project as a whole will help entry vehicle system designers determine suitable control architectures for integration with DEVs and other entry vehicle types.

Okolo, Wendy A.↗
NASA NTRS2004

Control systems of the large millimeter telescope

The paper presents the analysis results (in terms of settling time, bandwidth, and servo error in wind disturbances) of four control systems designed for the Large Millimeter Telescope (LMT). The first system, called PI’, consists of the proportional and integral (PI) controllers in the rate and position loops, and is widely used in the antenna and radiotelescope industry. The analysis shows that the PP control system performance is remarkably good when compared to similar control systems applied to typical antennas. This performance is achieved because the LMT structure is exceptionally rigid, however, it does not meet the stringent LMT pointing requirements. The second system, called PL, consists of the PI controller in the rate loop, and the Linear-Quadratic-Gaussian (LQG) controller in the position loop. This type of controller is implemented in the NASA Deep Space Network antennas, where pointing accuracy is twice that of PP control system. The third system, called LP, consists of the LQG controller in the rate loop, and the propositional-integral-derivative (PID) controller in the position loop. This type of loop has not been yet implemented at known antennas or radiotelescopes, but the analysis shows that its pointing accuracy is the ten times better than PP control system. The fourth system, called LL, consists of the LQG controller in both the rate loop, and the position loop. It is the best of the four, with accuracy 250 better than the PP system, thus is worth further investigations, to identify implementation challenges for the telescopes of high pointing requirements.

Souccar, Kainal↗
DOE OSTI2020

Flexible Reinforcement Learning Framework for Building Control using EnergyPlus-Modelica Energy Models

In recent years, reinforcement learning (RL) methods have been greatly enhanced by leveraging deep learning approaches. RL methods applied to building control have shown potential in many applications due to their ability to complement or replace conventional methods such as model-based or rule-based controls. However, RL-based building control software is likely tailored either to one target building system or to a specific RL method so that significant additional effort would be required to customize the RL-based controller for use in other building systems or with other RL approaches. Also, RL-based building controls usually depend on building energy simulations to train controllers, so emulating building dynamics (i.e., thermal dynamics and control dynamics) and capturing sub-hourly dynamic profiles are crucial to further the development of effective RL-based building control methods. To address these challenges, we present an open source RL-based control software employing a high-fidelity hybrid EnergyPlus-Modelica building energy model which emulates building dynamics at 1-minute resolution. This software consists of decoupled components (environment, building emulator, control agent, and RL algorithm), which allows for quick prototyping and benchmarking of standard RL algorithms in different systems; for example, a single component can be replaced without revising all of the software. To demonstrate this software framework, we conducted a benchmark study using an EnergyPlus-Modelica building energy model for a Chicago office building with an RL-based controller to dynamically control the chilled water temperature setpoint and the air handling unit supply air temperature setpoint on selected floors.

Lee, Joon-Yong↗
DOE OSTI2023

Model predictive current profile control in tokamaks by exploiting spatially moving electron cyclotron current drives

The plasma control systems in next-generation tokamaks like ITER will balance competing control objectives to achieve the desired level of performance in advanced scenarios while preventing magnetohydrodynamic instabilities and disruptions. During normal tokamak operation, the points of incidence of the electromagnetic waves generated by the electron cyclotron heating and current drives (EC H&CDs) are usually fixed in space. However, the points of incidence can be modified in real-time by changing the angles of the mirrors that reflect the EC H&CD waves. Altering the points of incidence, in turn, varies the ability of the plasma control system to regulate a plasma property. For instance, changing the EC H&CD wave incidence location may place the power demands necessary to achieve a particular plasma target within saturation limits. Therefore, using the EC H&CD deposition location, which is related to the EC H&CD mirror angle, as a supplementary controllable variable may facilitate access to a given target scenario. However, active scenario-control algorithms have not been designed so far to fully exploit this capability in real time. In this work, a model predictive controller that can handle actuation locations as control inputs is developed. In particular, the controller is designed to regulate both the auxiliary powers and the EC H&CD deposition locations in a pre-defined optimal sense to achieve the control objective of attaining and sustaining a target current profile. Further, the proposed controller is tested for a DIII-D tokamak scenario in nonlinear simulations using the Control Oriented Transport SIMulator (COTSIM).

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

Model predictive control of KSTAR equilibrium parameters enabled by TRANSP

Due to the complex behavior of tokamak plasmas and the importance of optimizing performance while avoiding instabilities and machine limits, plasma control algorithms are becoming increasingly dependent on sophisticated model-based control approaches. It is anticipated that the use of integrated modeling codes in the model-based control design process will reduce the amount of experimental time needed to implement new control algorithms by facilitating development of control-oriented models and enabling higher-fidelity closed-loop simulations. In this work, a reduced model is developed from a series of TRANSP simulations and is used to develop a model predictive control (MPC) algorithm for controlling important equilibrium parameters in KSTAR [1] discharges. The control algorithm uses the KSTAR neutral beam injection system and the target plasma current and plasma boundary as actuators, and optimizes the plasma stored energy, loop voltage, and internal inductance while avoiding constraints that could lead to disruptions. Higher fidelity testing of the control algorithm is performed using a flexible framework for enabling external processes to actively control plasma parameters in TRANSP simulations. Furthermore, closed-loop simulations demonstrate the ability of the control algorithm to respond to disturbances in density and confinement, handle actuator failures, and move the discharge to high non-inductive fraction conditions.

70 PLASMA PHYSICS AND FUSION TECHNOLOGY↗
DOE OSTI2020

MPC-Based Local Voltage Control Strategy of DGs in Active Distribution Networks

The increasing penetration of distributed generators (DGs) exacerbates the voltage violations in active distribution networks (ADNs). The commonly used centralized voltage control may aggravate the burden of computation and communication. Local voltage control can realize fast response to frequent voltage fluctuations. However, effective local control for DGs still needs to be further investigated. In this paper, a model predictive control (MPC)-based local voltage control strategy of DGs is proposed to cope with voltage violation problems and enhance the adaptation to the DG volatility. DGs are regulated with local Q-V control curves to respond to voltage fluctuations rapidly. Based on the dynamic network partition, a unified control curve for DGs is set in the same area to realize efficient operation. To further improve the performance of voltage control, the parameter tuning of control curves is regularly conducted based on decentralized inter-area coordination with the rolling of the control horizon. Finally, case study is verified on the modified IEEE 33-node distribution system. The proposed local control strategy of DGs can effectively improve the voltage profile and reduce power losses of the whole network without high computation and communication burden.

active distribution network (ADN)↗
DOE OSTI2022

Model predictive control for robust quantum state preparation

A critical engineering challenge in quantum technology is the accurate control of quantum dynamics. Model-based methods for optimal control have been shown to be highly effective when theory and experiment closely match. Consequently, realizing high-fidelity quantum processes with model-based control requires careful device characterization. In quantum processors based on cold atoms, the Hamiltonian can be well-characterized. For superconducting qubits operating at milli-Kelvin temperatures, the Hamiltonian is not as well-characterized. Unaccounted for physics (i.e., mode discrepancy), coherent disturbances, and increased noise compromise traditional model-based control. This work introduces model predictive control (MPC) for quantum control applications. MPC is a closed-loop optimization framework that (i) inherits a natural degree of disturbance rejection by incorporating measurement feedback, (ii) utilizes finite-horizon model-based optimizations to control complex multi-input, multi-output dynamical systems under state and input constraints, and (iii) is flexible enough to develop synergistically alongside other modern control strategies. We show how MPC can be used to generate practical optimized control sequences in representative examples of quantum state preparation. Specifically, we demonstrate for a qubit, a weakly-anharmonic qubit, and a system undergoing crosstalk, that MPC can realize successful model-based control even when the model is inadequate. These examples showcase why MPC is an important addition to the quantum engineering control suite.

71 CLASSICAL AND QUANTUM MECHANICS, GENERAL PHYSIC↗
DOE OSTI2021

Development of a new control rod drive mechanism design for the ISU AGN-201M reactor

The Aerojet General Nucleonics (AGN) model 201-Modified, known as the AGN-201M reactor, plays an essential role in the educational and research activities at Idaho State University (ISU). The ISU AGN-201M's original Control Rod Drive Mechanism (CRDM) has been in operation for more than fifty years with no large-scale redesigns. The CRDM is required to eject the fuel rods within one second during a SCRAM event (also known as a 'reactor trip') and adjust the control rods' insertion speed, and keeps the rod insertion sequence correct. The existing control rod drive mechanisms meet these criteria but experience a few concerns due to aging. Concerns include complex maintenance and costly repairs for old electromechanical components, rod position feedback errors, and the impediment of the plate during a SCRAM due to binding of the lead screws of the existing mechanism. During a binding event, the drive mechanism becomes locked, preventing the control rod's magnetic plate from moving in or out under the reactor's normal and emergency operating conditions. Although, the binding has no effect on the ability of the rod to exit the core during the SCRAM. To counteract the concerns and issues with the current CRDM, a new design has been proposed using newer components and a simplified design. The new design utilizes more advanced electric and mechanical components that are commercially available. The new CRDM system is divided into four main aspects: (1) control rod movement design (motor, lead screw, guide rods), (2) control rod ejection (springs, electromagnet), (3) control rod position and feedback (position transducer, microswitches), and (4) material selection and structural analysis. The new design aims to reduce the overall complexity and probability of failure to improve the reactor's overall reliability. With proper material selection and improved structural design, the new drives are lighter with little to no change in structural integrity. The new control rod drive mechanism eliminates binding scenarios by using a single lead screw and implementing additional guide rods. An advanced linear position sensor and microswitches replace the existing and aging synchro system for accurate rod position feedback resulting in better reactivity control. The new design meets the reactor's operational limits by having an average reactivity insertion of 0.065% Δk/k per second, which corresponds to a total control rod insertion time of 19.23 s, while the control rod's ejection time remains less than one second during a SCRAM event. The new design ensures the reactor's long-term viability for educational and research activities by increasing the reliability and safety of operation for years to come.

21 SPECIFIC NUCLEAR REACTORS AND ASSOCIATED PLANTS↗
DOE OSTI2021

Controlling distributed energy resources via deep reinforcement learning for load flexibility and energy efficiency

Behind-the-meter distributed energy resources (DERs), including building solar photovoltaic (PV) technology and electric battery storage, are increasingly being considered as solutions to support carbon reduction goals and increase grid reliability and resiliency. However, dynamic control of these resources in concert with traditional building loads, to effect efficiency and demand flexibility, is not yet commonplace in commercial control products. Traditional rule-based control algorithms do not offer integrated closed-loop control to optimize across systems, and most often, PV and battery systems are operated for energy arbitrage and demand charge management, and not for the provision of grid services. More advanced control approaches, such as MPC control have not been widely adopted in industry because they require significant expertise to develop and deploy. Recent advances in deep reinforcement learning (DRL) offer a promising option to optimize the operation of DER systems and building loads with reduced setup effort. However, there are limited studies that evaluate the efficacy of these methods to control multiple building subsystems simultaneously. Additionally, most of the research has been conducted in simulated environments as opposed to real buildings. This paper proposes a DRL approach that uses a deep deterministic policy gradient algorithm for integrated control of HVAC and electric battery storage systems in the presence of on-site PV generation. The DRL algorithm, trained on synthetic data, was deployed in a physical test building and evaluated against a baseline that uses the current best-in-class rule-based control strategies. Performance in delivering energy efficiency, load shift, and load shed was tested using price-based signals. The results showed that the DRL-based controller can produce cost savings of up to 39.6% as compared to the baseline controller, while maintaining similar thermal comfort in the building. The project team has also integrated the simulation components developed during this work as an OpenAIGym environment and made it publicly available so that prospective DRL researchers can leverage this environment to evaluate alternate DRL algorithms.

24 POWER TRANSMISSION AND DISTRIBUTION↗