1,720,983 research outputs found
Learning markov jump affine systems via regression trees for MPC
Full text linkModel Predictive Control is a well consolidated technique to design optimal control strategies, leveraging the capability of a mathematical model to predict a system's behavior over a time horizon. However, building physics-based models for complex large-scale systems can be cost and time prohibitive. To overcome this problem we propose a methodology to exploit Regression Trees technique in order to build a Markov Jump System model of a large-scale system using historical data, and apply Model Predictive Control. A comparison with an optimal benchmark and related techniques is provided on an energy management system to validate the performance of the proposed methodology
Data Predictive Control using Regression Trees and Ensemble Learning
No full textDecisions on how to best operate large complex plants such as natural gas processing, oil refineries, and energy efficient buildings are becoming ever so complex that model-based predictive control (MPC) algorithms must play an important role. However, a key factor prohibiting the widespread adoption of MPC, is the cost, time, and effort associated with learning first-principles dynamical models of the underlying physical system. An alternative approach is to employ learning algorithms to build black-box models which rely only on real-time data from the sensors. Machine learning is widely used for regression and classification, but thus far data-driven models have not been used for closed-loop control. We present novel Data Predictive Control (DPC) algorithms that use Regression Trees and Random Forests for receding horizon control. We demonstrate the strength of our approach with a case study on a bilinear building model identified using real weather data and sensor measurements. We further apply DPC to a large scale multi-story EnergyPlus building model to curtail total power consumption in a Demand Response setting. In such cases, when the model-based controllers fail due to modeling cost, complexity and scalability, our results show that DPC curtails the desired power usage with high confidence
DR-Advisor: A data-driven demand response recommender system
No full textDemand response (DR) is becoming increasingly important as the volatility on the grid continues to increase. Current DR approaches are predominantly completely manual and rule-based or involve deriving first principles based models which are extremely cost and time prohibitive to build. We consider the problem of data-driven end-user DR for large buildings which involves predicting the demand response baseline, evaluating fixed rule based DR strategies and synthesizing DR control actions. The challenge is in evaluating and taking control decisions at fast time scales in order to curtail the power consumption of the building, in return for a financial reward. We provide a model based control with regression trees algorithm (mbCRT), which allows us to perform closed-loop control for DR strategy synthesis for large commercial buildings. Our data-driven control synthesis algorithm outperforms rule-based DR by 17% for a large DoE commercial reference building and leads to a curtailment of up to 380 kW and over $45,000 in savings. Our methods have been integrated into an open source tool called DR-Advisor, which acts as a recommender system for the building's facilities manager and provides suitable control actions to meet the desired load curtailment while maintaining operations and maximizing the economic reward. DR-Advisor achieves 92.8-98.9% prediction accuracy for 8 buildings on Penn's campus. We compare DR-Advisor with other data driven methods and rank 2nd on ASHRAE's benchmarking data-set for energy prediction
Learning affine predictors for MPC of nonlinear systems via artificial neural networks
No full textNonlinear model predictive control (MPC) problems can be well approximated by linear time-varying (LTV) MPC formulations in which, at each sampling step, a quadratic programming (QP) problem based on linear predictions is constructed and solved at runtime. To reduce the associated computation burden, in this paper we explore and compare two methodologies for learning the entire output prediction over the MPC horizon as a nonlinear function of the current state but affine with respect to the sequence of future control moves to be optimized. Such a learning process is based on input/output data collected from the process to be controlled. The approach is assessed in a simulation example and compared to other similar techniques proposed in the literature, showing that it provides accurate predictions of the future evolution of the process and good closed-loop performance of the resulting MPC controller. Guidelines for tuning the proposed method to achieve a desired memory occupancy / quality of fit tradeoff are also given
On the stability of switched arx models, with an application to learning via regression trees
Full text linkThis work studies the stability properties of Switched AutoRegressive eXogenous (SARX) models subject to arbitrary switching sequences. We provide necessary and sufficient conditions for the arbitrary switching stability of multiple-input, single-output SARX models under nonnegativity constraints, and sufficient-only conditions removing sign constraints. The conditions are equivalentlv formulated on state-space representations of SARX models, due to their influential use in designing control strategies. As an application of the aforementioned results, we propose a novel algorithm for the identification of switched models with stability guarantees via Regression Trees, a powerful machine learning technique
Stability analysis of switched ARX models and application to learning with guarantees
No full textThe main subject of this work is the stability analysis of Switched Auto-Regressive models with eXogenous inputs (SARX), which constitute a reference class for switched and hybrid system identification. The work introduces novel conditions for the arbitrary switching stability of multiple-input multiple-output SARX models which exploit the peculiar structure of their state-space realization. The analysis relies on the properties of block companion matrices, and partly leverages results from the theory of non-negative matrices, without nevertheless asking for an input–output positive behavior of the model. The novel stability conditions have a simple formulation in terms of linear co-positive common Lyapunov functions, and come at a remarkably low computational cost, being solvable by Linear Programming. The low computational burden is particularly attractive in an identification context, as it allows to efficiently constrain learning procedures in order to obtain SARX models with stability guarantees. The latter is itself a contribution of the work, as it fills a gap in the literature on the estimation of SARX models. The results are validated on a particular learning technique based on Regression Trees – a well known machine learning algorithm – which has shown remarkable accuracy in experimental environments
Data-driven model predictive control with regression trees-an application to building energy management
No full textModel Predictive Control (MPC) plays an important role in optimizing operations of complex cyber-physical systems because of its ability to forecast system's behavior and act under system level constraints. However, MPC requires reasonably accurate underlying models of the system. In many applications, such as building control for energy management, Demand Response, or peak power reduction, obtaining a high-fdelity physics-based model is cost and time prohibitive, thus limiting the widespread adoption of MPC. To this end, we propose a data-driven control algorithm for MPC that relies only on the historical data. We use multioutput regression trees to represent the system's dynamics over multiple future time steps and formulate a fnite receding horizon control problem that can be solved in real-time in closed-loop with the physical plant. We apply this algorithm to peak power reduction in buildings to optimally trade-off peak power reduction against thermal comfort without having to learn white/grey box models of the systems dynamics
Optimal co-design of control, scheduling and routing in multi-hop control networks
No full textA Multi-hop Control Network consists of a plant where the communication between sensors, actuators and computational units is supported by a (wireless) multi-hop communication network, and data flow is performed using scheduling and routing of sensing and actuation data. Given a SISO LTI plant, we will address the problem of co-designing a digital controller and the network parameters (scheduling and routing) in order to guarantee stability and maximize a performance metric on the transient response to a step input, with constraints on the control effort, on the output overshoot and on the bandwidth of the communication channel. We show that the above optimization problem is a polynomial optimization problem, which is generally NP-hard. We provide sufficient conditions on the network topology, scheduling and routing such that it is computationally feasible, namely such that it reduces to a convex optimization problem
- …
