Primary Separation Vessel Research Articles

Autonomous Control of Primary Separation Vessel using Reinforcement Learning

Reinforcement Learning (RL) adapts to a dynamic environment through exploration and continual learning. However, the risks associated with exploration have limited RLC applications in industrial settings. This research designed a safe RLC framework for the bitumen extraction process from oil sands. We explore transfer learning techniques through Behavioral Cloning (BC), Generative Adversarial Imitation Learning (GAIL), and Simulation-to-Reality (Sim2Real) pretraining, which allow the RL agent to acquire fundamental knowledge before real-world exploration. This transferred knowledge steers exploration towards near-optimal and safer regions. GAIL and Sim2Real pretrained agents deliver satisfactory control performance right after pretraining, cutting process trips during online training by factors of 8 and 27, respectively. Further online training reduces Integral Squared Error (ISE) for interface tracking by 71% with GAIL and 22% with Sim2Real. ISE for tailings density tracking also decreases by 73% and 33%, respectively. This significant reduction demonstrates RL's continual learning ability suitable for autonomous control. This work highlights the potential for safely deploying RL in process automation and reveals that RLC can handle complex control problems in multivariate, multimodal, partially observable, and uncertain environments. RLC achieves Model Predictive Control (MPC)-level performance but with less controller effort, and computation time faster by a factor of 10.

Kalman Filter-Based Convolutional Neural Network for Robust Tracking of Froth-Middling Interface in a Primary Separation Vessel in Presence of Occlusions

Bitumen in the oil sands industry is separated from sand using a water-based gravity separation process in a primary separation vessel (PSV). The interface between the froth and the middlings layer is an important parameter to control for optimal operation of the PSV unit. In this paper, a method using computer vision based on convolutional neural networks (CNNs) and Kalman filter (KF) is designed for the detection of the interface in PSV, with the CNN estimating both the interface level and the image quality. The proposed method consists of two parts: an offline and an online stage, wherein the parameters of CNN and KF are trained in the offline stage using the available data. The algorithm is made robust to any new type of occlusions, not present in the training dataset, in the online stage. Experimental results demonstrate that the proposed method is accurate and robust to different abnormalities in the process.

Sensor Fusion and Computer Vision Integrated System for Primary Separation Vessel Interface Level Estimation

Abstract In oil sands industry, primary separation vessel (PSV) is a critical component to recover bitumen from oil sands slurry. Accurate interface level estimation between froth and middlings layers ensures economical and environmental benefits of bitumen recovery. Nuclear density profiler, differential pressure (DP) cell, and image processing based computer vision system are usually used to estimate the interface level. The computer vision system, which uses a camera to capture sight glass vision frames, is considered to be the most accurate. Although the accuracy of computer vision system is high in normal operational conditions, its qualities are influenced by abnormalities, such as sight glass vision blocking, stains, and level switching between sight glasses. A sensor fusion approach, which recursively updates fusion parameters according to accurate computer vision results whenever they are reliable, is proposed. The fused results can then be used to provide reliable interface level estimation under abnormal scenarios. The sensor fusion algorithm is further integrated with computer vision system to improve froth-middlings interface level estimation accuracy and robustness. Industrial environment simulations and factory accepted test (FAT) demonstrate the advantages and effectiveness of the sensor fusion and computer vision integrated system, which is applied in the industry.

A hierarchical constrained reinforcement learning for optimization of bitumen recovery rate in a primary separation vessel

This work proposes a two-level hierarchical constrained control structure for reinforcement learning (RL) with application in a Primary Separation Vessel (PSV). The lower level is concerned with servo tracking and regulation of the interface level against variances in ore quality by manipulating middlings flow rate. At the higher level, with the objective to optimize bitumen recovery rate, a supervisory interface level setpoint control is implemented. To prevent sanding, tailings density regulation using tailings withdrawal flow rate is proposed. For each case, an asynchronous advantage actor-critic (A3C) based agent is chosen to interact with a high-fidelity PSV model to learn the near optimal control strategy through episodic interactions. Each of the three control loops is sequentially learnt. In the interface level control loop, a behavioral cloning based two-phase learning scheme to promote stable state space exploration is proposed. The proposed hierarchical structure successfully demonstrates improved bitumen recovery rate by manipulating the interface level while preventing sanding.

Economic MPC with terminal cost and application to an oilsand primary separation vessel

In this work, a novel approach to construct the terminal cost for economic model predictive control (EMPC) is developed and two EMPC algorithms are designed based on the terminal cost for both infinite-time and finite-time operations. In the proposed approach, an auxiliary nonlinear controller which renders the desired optimal steady state asymptotically stable is taken advantage of. The idea is to take the auxiliary nonlinear controller as a benchmark control system and to design EMPC algorithms that gives practically better economic performance than the auxiliary nonlinear controller. The terminal cost ensures that the two EMPC algorithms give provable (practically) improved economic performance than the auxiliary controller with any prediction horizon. This means that the proposed EMPC algorithms could be very computationally efficient. The proposed EMPC algorithms are applied to an oilsand primary separation vessel. The simulation results demonstrate the effectiveness of the proposed approaches.

Adaptive Soft Sensing and On-line Estimation of the Critical Minimum Velocity with Application to an Oil Sand Primary Separation Vessel

Most of the previous studies on the required critical minimum velocity to move the solid bed inside the slurry pipelines are related to the design step where the dynamics of the process are not thoroughly considered. In this paper, a general framework for on-line estimation of the critical minimum velocity is proposed and applied to the underflow stream of an important Primary Separation Vessel (PSV) in an oil sand industry. Appropriate statistical and probabilistic models are used to improve the on-line measurements and estimations. The proposed method demonstrates a satisfactory performance in detection of different conditions of the PSV operations.

Economic MPC with Terminal Cost and Application to Oilsand Separation

In this work, a novel approach to construct the terminal cost for economic model predictive control (EMPC) is developed. In the proposed approach, an auxiliary nonlinear controller which renders the desired optimal steady state asymptotically stable is taken advantage of. The proposed terminal cost is constructed such that it reflects the economic performance of the auxiliary controller over sufficiently large but finite steps. The proposed EMPC design provides guaranteed closed-loop economic performance which does not depend on the length of the prediction horizon. This means that the proposed EMPC algorithm could be very computationally efficient. The proposed EMPC algorithm is applied to an oilsand primary separation vessel. The simulation results demonstrate the effectiveness of the proposed approach.

Operating condition diagnosis based onHMMwith adaptive transition probabilities in presence of missing observations

A new approach for modeling and monitoring of the multivariate processes in presence of faulty and missing observations is introduced. It is assumed that operating modes of the process can transit to each other following a Markov chain model. Transition probabilities of the Markov chain are time varying as a function of the scheduling variable. Therefore, the transition probabilities will be able to vary adaptively according to different operating modes. In order to handle the problem of missing observations and unknown operating regimes, the expectation maximization algorithm is used to estimate the parameters. The proposed method is tested on two simulations and one industrial case studies. The industrial case study is the abnormal operating condition diagnosis in the primary separation vessel of oil‐sand processes. In comparison to the conventional methods, the proposed method shows superior performance in detection of different operating conditions of the process. © 2014 American Institute of Chemical EngineersAIChE J, 61: 477–493, 2015

Adaptive monitoring of the process operation based on symbolic episode representation and hidden Markov models with application toward an oil sand primary separation

This paper presents a novel procedure for classification of normal and abnormal operating conditions of a process when multiple noisy observation sequences are available. Continuous time signals are converted to discrete observations using the method of triangular representation. Since there is a large difference in the means and variances of the durations and magnitudes of the triangles at different operating modes, adaptive fuzzy membership functions are applied for discretization. The expectation maximization (EM) algorithm is used to obtain parameters of the different modes for the durations and magnitudes assuming that states transit to each other according to a Markov chain model. Applying Hamilton's filter, probability of each state given new duration and magnitude is calculated to weight the membership functions of each mode previously obtained from a fuzzy C-means clustering. After adaptive discretization step, having discrete observations available, the combinatorial method for training hidden Markov models (HMMs) with multiple observations is used for overall classification of the process. Application of the method is studied on both simulation and industrial case studies. The industrial case study is the detection of normal and abnormal process conditions in the primary separation vessel (PSV) of an oil sand industry. The method shows an overall good performance in detecting normal and risky operating conditions.

The use of physical modeling in the optimisation of a primary separation vessel feedwell

AbstractA physical modeling program was undertaken to assess and eliminate a significant vessel wear problem observed in a primary separation vessel used in the recovery of bitumen from oil sands. A 1:20 scale model was fabricated, and process rates and materials were selected on the basis of dimensional analysis. Experiments showed that the existing feedwell produced poor circumferential distribution and formed a localised jet that entered the vessel over a narrow sector. Tests were then conducted with several different modifications to the feedwell to improve the flow distribution. The final design produced uniform circumferential distribution and a modest improvement in model separation efficiency. This design has been in commercial service for several years and has eliminated the local wear problems.

X-ray Diffraction (XRD)-Derived Processability Markers for Oil Sands Based on Clay Mineralogy and Crystallite Thickness Distributions

An X-ray diffraction (XRD) methodology has been developed for characterizing clays in unextracted oil sands. Application of the new technique to five estuarine and five marine ores directly identified three clay mineral properties that may impact bitumen recovery: (1) The specific surface area of illite was significantly greater for four oil sand ores identified as problematic in batch extraction unit tests. (2) The correlation of illite/kaolinite XRD peak area ratios with bitumen recovery produced a processability classification similar to that proposed in earlier work. (3) Significant amounts of chlorite, as measured by XRD, were observed only in marine oil sands; this may provide a means to distinguish marine from estuarine ores. A combination of XRD analysis on separated clays and laser diffraction determination of clay contents provided a quantitative estimate for the illite and kaolinite contents of the oil sands. Also, the contribution from ultrathin illite and kaolinite for each oil sand (i.e., the mass fractions of illite and kaolinite with crystallite thicknesses of 1-3 composite layers) was determined. This methodology thus provides a direct method for the determination of the ultrafines content in unextracted oil sands and obviates the necessity for the time-consuming wet chemistry technique for separation of this component. For the 10 oil sands analyzed here, ultrathin crystallites occurred almost entirely in the illite clay fraction. The amount of ultrathin illite was critical and closely matched the ultrafines concentration required to cause sludging (gelation) in the primary separation vessel, with concomitant loss of bitumen recovery during extraction.

05/02421 Recovery of bitumen from oilsands: gelation of ultra-fine clay in the primary separation vessel

05/02421 Recovery of bitumen from oilsands: gelation of ultra-fine clay in the primary separation vessel

Effects of Chemical Application on Antifouling in Steam-Assisted Gravity Drainage Operations

The fouling behavior of water streams in steam-assisted gravity drainage (SAGD) operations can be affected by chemical selection, including a reverse emulsion breaker, demulsifier, flocculant, and dispersant. Reverse emulsion breakers and demulsifiers were chosen to give the cleanest water quality and best dehydration during laboratory testing of reverse emulsions produced at SAGD facilities. A group of polyquaternary amine reverse emulsion breakers were synthesized to have a range of molecular weights with equivalent charge densities. The effect of their molecular weights on water clarification was investigated by standard bottle testing. It was found that the higher the molecular weight, the more effective the chemical was at removing suspected fouling agents. A flocculant has to be used with care to avoid overtreatment of the system. If fouling occurs during normal operation due to produced water from the primary separation vessel containing residual amounts of oil or other impurities, remediation can be used to address the fouling. A series of dispersants were tested to check if the oil could be kinetically stabilized in the produced water phase. It was shown that decreased viscosity of the oil increased the tendency for the oil to be dispersed. Certain surfactants helped the oil to be dispersed into the water and then to be released upon settling for 30 min. This type of dispersant could be applied in any SAGD operation with heat exchanger fouling problems.

Recovery of bitumen from oilsands: gelation of ultra-fine clay in the primary separation vessel

The bulk of bitumen recovery from oilsands takes place via a flotation/settling process in the primary separation vessel (PSV). Under certain conditions, some oilsands slurries become non-segregating, i.e. both fines and coarse solids remain in suspension and little or no bitumen froth is generated. An ultra-fine (<0.3 μm) component of the oilsands fines fraction (<44 μm) is identified as having the potential to be the major contributor to the thickening (gelation), or sludging, phenomenon observed in some ores. In this work we determined the amounts of ultra-fines in oilsands from different depositional environments. Although a generally linear correlation exists between ultra-fines and fines or clay contents, some samples were found to have a disproportionately high ultra-fines contribution relative to other oilsands. Waste units, in particular, were found to be extremely rich in the ultra-fines and clay-sized fractions. We used a 2H NMR method to measure both the rate and degree of gelation of ultra-fines suspensions separated from oilsands. Sludging conditions are reached when the ultra-fines concentration is between about 1.5 and 2.0 w/w%. Measurement of ultra-fines gelation times showed them to be of the same order as slurry residence time in the PSV. Laboratory jar tests have demonstrated that poor segregation of oilsands components can be correlated with the gelation conditions determined by NMR measurements.

Bitumen-air aggregates flow to froth layer: II. Effect of ore grade and operating conditions on aggregate composition and bitumen recovery

Size, rise velocity, shape, and composition of bitumen-air aggregates floating to the froth layer during the processing of 11.1% bitumen Estuarine ore, and 7.2% bitumen Marine ore under various operating conditions (caustic dosage and feed density) were determined. The flux of the bitumen-air aggregates inside a Primary Separation Vessel (PSV) processing 2 tonnes/hour of the ore was in-situ monitored and video recorded. The mass of bitumen contained in the aggregates floating to the froth layer under different operating conditions was calculated on the basis of the measured rise velocities and dimensions of the aggregates. It was found that the type of ore processed was the parameter having the largest effect on size and composition of the bitumen-air aggregates floating to the froth layer inside the PSV during a 50°C warm slurry extraction process. In the case of the Estuarine ore the average mass of bitumen contained in an average aggregate was ca. 9 ∗10 −4 g, and the average aggregate size was ca. 1.0 mm. During processing of the low grade (Marine) ore the aggregates were smaller (within 0.33–0.69 mm) and they carried much less (from 0.17–1.37 ∗10 −4 g) bitumen, depending on other operating conditions (caustic dosage and feed density). A good correlation between size of the bitumen-air aggregates and bitumen recovery in the primary separation vessel was found. Thus, size of the aggregates can be used as an indication of better flotation recovery.

The Physical Chemistry Of The Hot Water Process

Abstract The roles in the hot water process of sodium hydroxide, surfactants, water soluble electrolytes, fine solids and electric properties of the bitumen/water and solids/water interfaces are examined. Recovery of bitumen from low- to medium-grade oil sands can be increased by ensuring the fine solids remain dispersed during digestion by adjustment of digestion water salinity and pH. Recovery of bitumen from high-grade oil sands appears to be controlled by surfactants which affect bitumen water interfacial tension and the size of bitumen droplets formed during the initial separation of the bitumen from sand grains. Processing behaviour of mixtures of oil sands is explained within the framework of these mechanisms. Introduction In the past few years, there has been a renewed effort to understand the hot water process which is used by Suncor Inc. and Syncrude Canada Ltd. to commercially extract bitumen from mined Athabasca oil sand. The process can be considered in terms of two distinct stages; separation of the bitumen from the sand followed by flotation which encompasses both primary flotation and scavenging. Separation is induced by mixing the oil sand in a conditioning drum with hot water and often a process aid, usually sodium hydroxide. The conditioned slurry is pumped to a primary separation vessel (PSV) where the bitumen froth is skimmed from the surface. The sand tailings which settle in the PSV are removed. The middlings (water) are pumped to a secondary recovery vessel where air is introduced through the bottom of the vessel, attaches to some of the suspended bitumen and causes it to float. The amount of sodium hydroxide added to the conditioning stage of the process is the major variable used to maximize recovery of bitumen. The addition rate historically has been based on empirical but sometimes imprecise correlations of oil sand grade with processability. Recently Schramm and co-workers proposed that the process could be optimized around the concentration of free surfactants formed due to reactions between sodium hydroxide and organic acids in the bitumen, and found in the tailings water(l). Even if the surfactant concentration in the tailings stream of a hot water processing plant could be accurately monitored there would be a delay between the time recovery fell below its optimum level and an adjustment to the digestion stage of the process could be made. A model that is used to predict the best operating conditions based upon properties of oil sand that are easily measured either during a coring program or on-line, would be beneficial compared to one which is used in a reactive mode. This paper describes a model for the hot water process based on the Derjaguin-Landau-Verwey-Overbeek (DLVO) and Ionizable Surface Group theories that can be used to predict from concentrations of inorganic salts in the digestion stage of the process, the conditions for maximizing recovery of bitumen from low grades of oil sand. The limitations of this approach especially for oil sands containing low levels of fine solids, arc addressed through consideration of the role of surfactants in the process.