Abstract: Disclosed are methods and systems that help identify critical variables for more efficient and robust plan validation process. An example embodiment is a computer implemented method of industrial process control. The example method includes receiving in computer memory a dataset including initial process parameters representing operational data of a subject industrial process, and, using filtering operations and grouping operations on the dataset, identifying a subset of the process parameters indicative of control data for controlling the subject industrial process. The example method further includes automatically applying the identified subset of process parameters controlling the subject industrial process.

Abstract: A controller has improved closed-loop step testing of a dynamic process of an industrial processing plant. The controller performs economic optimization relaxation on process variables, such that operating range of the variables (MVs and CVs) during the testing are not skewed by variations in optimization cost factors. The controller employs computer-implemented methods and systems that receive a user-defined giveaway tolerance representing an allowable range between a current process variable value and a target process variable value. In response to the variables not meeting the giveaway tolerance, embodiments adjust the MPC controller configuration to drive the variables inside the tolerance, while relaxing optimization of the variables already meeting the giveaway tolerance. Using the adjusted configuration, embodiments calculate a new set of targets and generate a dynamic move plan from the new target.

Abstract: A method and system for estimating wax or hydrate deposits is desirable for the oil industry and important for assuring flow conditions and production, avoiding downtime, and reducing or preventing costly interventions. The method and system disclosed herein use artificial intelligence and machine learning techniques combined with oil well historical operational sensor data and historical operational event records (such as diesel hot flush, slick line, coil tubing, etc.) to build an oil well model. The method and system enable oil well practitioners to test and validate the built model and deploy the model online to estimate and/or detect wax or hydrate deposition status. By using one or more such models in operating an oil well, users can monitor and/or detect the status of wax of hydrate deposits in an oil well and can optimize production, maintenance, and planning for oil wells.

Abstract: Computer-based process modeling and simulation methods and systems combine first principles models and machine learning models to benefit where either model is lacking. In one example, input values (measurements) are adjusted by first principles techniques. A machine learning model of the chemical process of interest is trained on the adjusted values. In another example, a machine learning model represents the residual (delta) between a first principles model prediction and empirical data. Residual machine learning models correct physical phenomena predictions in a first principles model of the chemical process. In another example, a first principles simulation model uses the process input data and predictions of the machine learning model to generate simulated results of the chemical process.

Abstract: Deep Learning is a candidate for advanced process control, but requires a significant amount of process data not normally available from regular plant operation data. Embodiments disclosed herein are directed to solving this issue. One example embodiment is a method for creating a Deep Learning based model predictive controller for an industrial process. The example method includes creating a linear dynamic model of the industrial process, and based on the linear dynamic model, creating a linear model predictive controller to control and perturb the industrial process. The linear model predictive controller is employed in the industrial process and data is collected during execution of the industrial process. The example method further includes training a Deep Learning model of the industrial process based on the data collected using the linear model predictive controller, and based on the Deep Learning model, creating a Deep Learning model predictive controller to control the industrial process.

Abstract: A computer-implemented method and system for process schedule reconciliation receives a scheduling model and an initial schedule for reconciliation, where the initial schedule includes projected plant data. Current plant data is imported into the system. The current plant data and projected plant data is processed using mathematical modeling techniques to identify event boundaries, stream flowrates associated with tanks and process units. The system builds an optimization model applying identified event boundaries, stream flowrates and pre-determined constraints along a period of time that includes priority slots to reconcile the projected plant data of the initial schedule with the current plant data, and then solves the optimization model to develop a reconciled schedule.

Abstract: Systems and methods for building predictive and prescriptive analytics of wind turbines generate a historical operational dataset by loading historical operational SCADA data of one or more wind turbines. Each sensor measurement is associated with an engineering tag and at least one component of a wind turbine. The system creates one or more performance indicators corresponding to one or more sensor measurements, and applies at least one data clustering algorithm onto the dataset to identify and label normal operation data clusters. The system builds a normal operation model using normal operational data clusters with Efficiency of Wind-To-Power (EWTP) and defines a statistical confidence range around the normal operation model as criterion for monitoring wind turbine performance. As real-time SCADA data is received by the system, the system can detect an anomalous event, and issue an alert notification and prescriptive early-action recommendations to a user, such as a turbine operator, technician or manager.

Abstract: Systems and methods provide a new paradigm of Advanced Process Control that includes building and deploying APC seed models. Embodiments provide automated data cleansing and selection in model identification and adaption in multivariable process control (MPC) techniques. Rather than plant pre-testing onsite for building APC seed models, the embodiments help APC engineers to build APC seed models from existing plant historical data with self-learning automation and pattern recognition, AI techniques. Embodiments further provide “growing” and “calibrating” the APC seed models online with non-invasive closed loop step testing techniques. PID loops and associated SP, PV, and OPs are searched and identified. Only “informative moves” data is screened, identified, and selected among a long history of process variables for seed model development and MPC application.

Abstract: Deep learning models and other complex models provide accurate representations of complex industrial processes. However, these models often fail to satisfy properties needed for their use in closed loop systems such as Advanced Process Control. In particular, models need to satisfy gain-constraints. Methods and systems embodying the present invention create complex closed-loop compatible models. In one embodiment, a method creates a controller for an industrial process. The method includes accessing a model of an industrial process and receiving indication of at least one constraint. The method further includes constructing and solving an objective function based on at least one constraint and the model of the industrial process. The solution of the objective function defines a modified model of the industrial process that satisfies the received constraint and can be used to create a closed-loop controller to control the industrial process.

Abstract: Computer tool determines target feedstock for a refinery, process complex, or plant. The tool receives a dataset of market conditions and preprocesses the data based on properties of the plant. Using the preprocessed data and machine learning, the tool trains predictive models. Each predictive model calculates a breakeven value of a candidate feedstock for the given plant under an individual market condition. Different predictive models optimize for different market conditions. A trained predictive model is selected based on a current market condition. The tool applies the selected predictive model and determines whether a candidate feedstock is a target feedstock for the refinery under the current market condition.

Abstract: Computer implemented methods and systems generate an improved predicted model of an industrial process or process engineering system. The model is a function of measurable features of the subject process and selected first principle features. First principle features are selected that capture linearities in a residual of a linear model constructed using a received dataset of the subject process. The model can further be a function of a scaled spline. The scaled spline is generated by computing a spine for a measurable feature of the subject process, fitting the computer spline to the residual of the constructed linear model, and scaling the fitting spline with a scaling factor. The model results in improved predictions of behavior of the subject process by relying primarily on the data of the measurable features of the subject process.

Abstract: Embodiments include a computer-implemented method (and system) for performing automated batch data alignment for modeling, monitoring, and control of an industrial batch process. The method (and system) loads, scales, and screens plant historian batch data for an industrial batch process. The method (and system) selects a reference batch as basis of the batch alignment, defines and adds or modifies one or more batch phases, and selects one or more batch variables based on one or more profiles and corresponding curvatures of the batch data. The method (and system) estimates one or more weightings, adjust one or more tuning parameters and uses a sliding time window combined with DTW, DTI and GSS algorithms, performs the batch alignment in offline mode or online mode.

Abstract: Computer-based methods and systems provide automated batch data alignment for a batch production industrial process. An example embodiment selects a reference batch from batch data for a subject industrial process and configures batch alignment settings. In turn, a seed model configured to predict alignment quality given settings for one or more alignment hyperparameters is constructed. Collectively the selected reference batch, the configured batch alignment settings, the constructed seed model, and a set of representative batches, representative of the batch data for the industrial process, are used to perform at least one of: (i) automated active learning, (ii) interactive active learning, and (iii) guided learning to determine settings for the one or more alignment hyperparameters. Then, a batch alignment is performed using the determined settings for the one or more alignment hyperparameters and the configured batch alignment settings.

Abstract: Computer implemented methods and systems incorporate physics-based and/or chemistry-based constraints into a model of a chemical, physical, or industrial process. The model is derived from a representative dataset of the subject process. The constrained model provides predictions of process behavior that are guaranteed to be consistent with incorporated constraints such as mass balances, atom balances, and/or energy balances while being less computationally intensive than equivalent first principle models. The constrained model can be constructed by matrix multiplication, namely multiplying the solution of an unconstrained linear model by a matrix that enforces the constraints. Improved process control models result, as well as improved process modeling and simulation models result.