Multi-model Structure Research Articles

Koopman operator-based multi-model for predictive control

AbstractThis work describes a new model structure developed for prediction in Model Predictive Control (MPC). The model has a multi-model structure in which independent sub-models are employed for the consecutive sampling instants. The model lifts process states into a high-dimensional space in which a linear process description is applied. Depending on the influence of the manipulated variables on lifted states, three general model versions are described and model identification algorithms are derived. As a result of the multi-model structure, model parameters are found analytically from computationally uncomplicated least squares problems using the Extended Dynamic Mode Decomposition algorithm, but the evolution of states over the horizon used in MPC is taken into account. Next, the MPC algorithm for the described model is derived. It requires solving online simple quadratic optimisation tasks. The effectiveness of three considered model configurations and three versions of the lifting functions is examined for a nonlinear DC motor benchmark. Their impact on model accuracy, complexity, possible control accuracy and MPC calculation time is thoroughly discussed. Finally, a more complex polymerisation reactor process is considered to showcase the practical applicability of the presented approach to modelling and MPC.

Impacts of spatiotemporal resolutions of precipitation on flood event simulation based on multimodel structures – a case study over the Xiang River basin in China

Abstract. Accurate flood event simulation and prediction, enabled by effective models and reliable data, are critical for mitigating the potential risk of flood disaster. This study aims to investigate the impacts of spatiotemporal resolutions of precipitation on flood event simulation in a large-scale catchment of China. We use high-spatiotemporal-resolution Integrated Multi-satellite Retrievals for Global Precipitation Measurement (IMERG) products and a gauge-based product as precipitation forcing for hydrologic simulation. Three hydrological models (HBV, SWAT and DHSVM) and a data-driven model (long short-term memory (LSTM) network) are utilized for flood event simulation. Two calibration strategies are carried out, one of which targets matching of the flood events, with peak discharge exceeding 8600 m3 s−1 between January 2015 and December 2017, and the other one is the conventional strategy for matching the entire streamflow time series. The results indicate that the event-based calibration strategy improves the performance of flood event simulation compared with a conventional calibration strategy, except for DHSVM. Both hydrological models and LSTM yield better flood event simulation at a finer temporal resolution, especially in flood peak simulation. Furthermore, SWAT and DHSVM are less sensitive to the spatial resolutions of IMERG, while the performance of LSTM obtains improvement when degrading the spatial resolution of IMERG-L. Generally, LSTM outperforms the hydrological models in most flood events, which implies the usefulness of the deep learning algorithms for flood event simulation.

Data-Driven Model-Predictive Control for Nonlinear Systems With Stochastic Sampling Interval

Stochastic sampling is ubiquitous for the complex nonlinear systems, which may destroy the stability of the closed-loop system. Concerned with the sampled-data stabilization problem for the nonlinear systems with complex dynamics and stochastic sampling, a data-driven model-predictive control (DDMPC) strategy is proposed to control the system at each sampling instant. The main merits of DDMPC are threefold. First, a data-driven method based on the fuzzy neural network (FNN) is used to establish the mathematical model for the complex nonlinear system. Especially, considering the stochastic sampling characteristics of the system, a multimodel structure is constructed to predict the future system outputs at possible sampling instants. Second, the objective function is designed to deal with the stochastic sampling control problem. The reasonable control laws are computed to stabilize the nonlinear system with stochastic sampling. Finally, the stability of DDMPC is proved in theory, and the numerical simulation results and industrial application for the wastewater treatment process (WWTP) reveal that the proposed DDMPC can achieve a satisfactory control performance for the stochastic sampling nonlinear system (SSNS).

Fast Training of Neural Affine Models for Model Predictive Control: An Explicit Solution

This work describes a nonlinear affine model developed especially for prediction in Model Predictive Control (MPC). A multi-model structure is used in which independent sub-models compute predictions for the consecutive sampling instants. All sub-models are affine with respect to the future values of the manipulated variable that are calculated in MPC. Model coefficients are time-varying; they are determined online by a neural network of the Radial Basis Function type. It is proved that the described model configuration makes it possible to formulate the training task as a least squares problem that has an explicit solution (the global optimum). A chemical reactor benchmark is considered to show advantages of the discussed modeling approach and the resulting MPC scheme. As a result of model affinity, MPC requires solving online simple quadratic optimization tasks.

Self-Organizing Fuzzy Belief Inference System for Classification

Evolving fuzzy systems (EFSs) are widely known as a powerful tool for streaming data prediction. In this article, a novel zero-order EFS with a unique belief structure is proposed for data stream classification. Thanks to this new belief structure, the proposed model can handle the interclass overlaps in a natural way and better capture the underlying multimodel structure of data streams in the form of prototypes. Utilizing data-driven soft thresholds, the proposed model self-organizes a set of prototype-based <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">if</small> – <sc xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">then</small> fuzzy belief rules from data streams for classification, and its learning outcomes are practically meaningful. With no requirement of prior knowledge in the problem domain, the proposed model is capable of self-determining the appropriate level of granularity for rule-based construction, while enabling users to specify their preferences on the degree of fineness of its knowledge base. Numerical examples demonstrate the superior performance of the proposed model on a wide range of stationary and nonstationary classification benchmark problems.

Input convex neural networks in nonlinear predictive control: A multi-model approach

The presented input convex neural multi-modelling approach to Model Predictive Control (MPC) has two essential advantages. Firstly, the MPC algorithm solves only convex optimisation tasks; complex not convex and multi-modal problems are not considered. Secondly, the model is trained in a simple one-step-ahead configuration and it is not used recurrently for prediction. Two neural multi-model structures are considered: fully input convex and partially input convex ones. The fully convex model structure turns out to be too restrictive for the considered polymerisation reactor benchmark process. The partially convex model leads to very good prediction accuracy and excellent control quality. Development of fully and partially input convex multi-models is thoroughly described, particularly selection of the number of the hidden nodes. Additionally, low effectiveness of a linear multi-model is discussed for comparison.

A multi-scale multi-model deep neural network via ensemble strategy on high-throughput microscopy image for protein subcellular localization

Protein subcellular locations are closely related to the function of proteins. By detecting the abnormalities in subcellular locations, we can infer the occurrence of some diseases and mine new drug targets. A large number of single-cell high-throughput microscopy images provide us with relevant resources for studying protein distribution patterns. However, in the existing image-based protein subcellular localization methods, the traditional techniques are lack of efficiency and accuracy, and the potential of deep learning methods has not been fully tapped. So in this study, we propose a multi-scale multi-model deep neural network via ensemble strategy for protein subcellular localization on single-cell high-throughput images. First of all, we employ a deep convolutional neural network as multi-scale feature extractor and use global average pooling to map extracted features at different stages into feature vectors, then concatenate these multi-scale features to form a multi-model structure for image classification. In addition, we add Squeeze-and-Excitation Blocks to the network to emphasize more informative features. What is more, we use an ensemble method to fuse the classification results from the multi-model structure to obtain the final subcellular location of each single-cell image. Experiments show the validity and effectiveness of our method on yeast cell images, it can significantly improve the accuracy of high-throughput microscopy image-based protein subcellular localization, and we achieve the classification accuracy of 0.9098 on the high-throughput microscopy images of yeast cells. In the work of protein subcellular localization, our method provides a framework for processing and classifying microscope images, and further lays the foundation for the study of protein and gene functions.

Fuzzy decoupled-states multi-model identification of gas turbine operating variables through the use of their operating data

Practically the rotating machines degradation, such as gas turbines, is due to the quality of construction and online operation of their dynamic state models, of different physical phenomena affecting these machines which cause their total malfunction. To maintain their stable operation, it is essential to correctly describe these real dynamic behaviors by reliable and robust representations, by models that can be used in monitoring and diagnostics. To achieve the performance objectives in terms of security, reliability, availability, and operating safety, this work proposes the development of a fuzzy multi-model identification approach with states decoupled from the operating variables, uploaded for monitoring a TITAN 130 turbine. This fuzzy multi-model structure with decoupled states is of interest for the monitoring of industrial systems because it adapts to the different changes in dynamic behavior of the system, makes it possible to represent the nonlinear behavior of the real system in a linear multi-model form without loss of information. In this work, through the different implementations and obtained results, this approach clearly shows how the gas turbine dynamics were reproduced when using the proposed fuzzy multi-models, thus allowing better performance when exploiting it for the synthesis of the faults diagnosis strategy for this rotating machine.

Which Features Are More Correlated to Illuminant Estimation: A Composite Substitute

Computational color constancy (CCC) is to endow computers or cameras with the capability to remove the color bias effect caused by different scene illuminations. The first procedure of CCC is illuminant estimation, i.e., to calculate the illuminant color for a given image scene. Recently, some methods directly mapping image features to illuminant estimation provide an effective and robust solution for this issue. Nevertheless, due to diverse image features, it is uncertain to select which features to model illuminant color. In this research, a series of artificial features weaved into a mapping-based illuminant estimation framework is extensively investigated. This framework employs a multi-model structure and integrates the functions of kernel-based fuzzy c-means (KFCM) clustering, non-negative least square regression (NLSR), and fuzzy weighting. By comparing the resulting performance of different features, the features more correlated to illuminant estimation are found in the candidate feature set. Furthermore, the composite features are designed to achieve the outstanding performances of illuminant estimation. Extensive experiments are performed on typical benchmark datasets and the effectiveness of the proposed method has been validated. The proposed method makes illuminant estimation an explicit transformation of suitable image features with regressed and fuzzy weights, which has significant potential for both competing performances and fast implementation against state-of-the-art methods.

A Stacked Ensemble-Based Classifier for Breast Invasive Ductal Carcinoma Detection on Histopathology Images

Breast cancer is one of the main causes of death in women. When body tissues start behaves abnormally and the ratio of tissues growth becomes asymmetrical then this stage is called cancer. Invasive ductal carcinoma (IDC) is the early stage of breast cancer. The early detection and diagnosis of invasive ductal carcinoma is a significant step for the cure of IDC breast cancer. This paper presents a convolutional neural network (CNN) approach to detect and visualize the IDC tissues in breast on histological images dataset. The dataset consists of 90 thousand histopathological images containing two categories: Invasive Ductal Carcinoma positive (IDC+) and Invasive Ductal Carcinoma negative (IDC-). For the detection of this disease, we have proposed an ensemble learning-based deep learning model that consists of the multi-model structure having different internal configuration settings of CNN. As a result, the proposed method has significant progress as per present techniques in terms of accuracy as 92.7%. Therefore, the ensemble approach of multi-model CNN can provide the evolution of IDC at different stages, and this would help the pathologist in the cure of cancer. The experimented results show that the proposed ensemble-based approach of multi-model CNN generated the preeminent results with an accuracy of 92.7% as compared to the already existing techniques.

Receding interval prediction of district heat load via finite difference multi-operating-domain dynamic modelling

To realize low carbon emission, district heat load (DHL) prediction is of great significance to the efficient operation of urban district heating system (DHS). Due to spatial topology of central heating network, weather conditions and heat users show great spatiotemporal dispersion, whose influence to DHL is very complex and its prediction is challengeable. To solve this problem, a multi-operating-domain dynamic modelling structure is proposed in this paper. Firstly, through detailed analysis from first-principle, external factors such as weather and user factors are quantitatively included to form candidate inputs set where the NCA (Neighbourhood Component Analysis) algorithm is used to extract feature inputs. Considering time-delays of feature inputs to DHL, AIC (Akaike Information Criterion) is applied to determine delay orders. Then, feature inputs, DHL output and their delay-orders are integrated to form a finite difference operating space, representing by a regression vector. After its high-dimensional clustering and hyperplane estimation, the whole space is convexly partitioned into multiple polyhedral domains. In each domain, using the finite difference input vector and output, multi-step receding prediction of heat load is realized using the Bi-LSTM (Bidirectional Long-short Temporal Memory) neural network. To further quantify prediction uncertainty, the nonparametric CKDE (Conditional Kernel Density Estimation) method are adopted to evaluate prediction error, where confidence boundaries can be obtained as interval prediction models. Finally, a DHS in Northwest China is selected for verification. The results show that the proposed method can accurately predict short-term DHL under the proposed multi-model structure. It is meaningful to achieve the purposes of on-demand heating, energy saving and environmental protection.

A multi-model approach to assessing the impacts of catchment characteristics on spatial water quality in the Great Barrier Reef catchments

Water quality monitoring programs often collect large amounts of data with limited attention given to the assessment of the dominant drivers of spatial and temporal water quality variations at the catchment scale. This study uses a multi-model approach: a) to identify the influential catchment characteristics affecting spatial variability in water quality; and b) to predict spatial variability in water quality more reliably and robustly. Tropical catchments in the Great Barrier Reef (GBR) area, Australia, were used as a case study. We developed statistical models using 58 catchment characteristics to predict the spatial variability in water quality in 32 GBR catchments. An exhaustive search method coupled with multi-model inference approaches were used to identify important catchment characteristics and predict the spatial variation in water quality across catchments. Bootstrapping and cross-validation approaches were used to assess the uncertainty in identified important factors and robustness of multi-model structure, respectively. The results indicate that water quality variables were generally most influenced by the natural characteristics of catchments (e.g., soil type and annual rainfall), while anthropogenic characteristics (i.e., land use) also showed significant influence on dissolved nutrient species (e.g., NOX, NH4 and FRP). The multi-model structures developed in this work were able to predict average event-mean concentration well, with Nash-Sutcliffe coefficient ranging from 0.68 to 0.96. This work provides data-driven evidence for catchment managers, which can help them develop effective water quality management strategies.

Identification of nonlinear time‐varying systems using wavelet neural networks

AbstractThe dynamic model of an aircraft changes significantly by altering the flight speed and the vehicle altitude. Thus, a conventional aircraft has a nonlinear time‐varying dynamic model in different regions of the flight envelope, and a dynamic model developed for a specific operating point is not valid in the entire flight envelope. This paper presents a novel identification approach that can deal with nonlinear and time‐varying characteristics of complex dynamic systems, especially an aerial vehicle in the entire flight envelope. In this regard, a set of local submodels are first developed at different operating points of the system, and subsequently, a multimodel structure is introduced to aggregate the outputs of the local models as a single model. Wavelet Neural Networks (WNNs), which combine both the universal approximation property of neural networks and the wavelet decomposition capability, are used as the local models of the proposed scheme. Also, three different approaches for determining the validity functions of the local models are introduced to allow for identifying the time‐varying dynamics of the system. The simulation results obtained for the Generic Transport Model aircraft suggest that the proposed WNN‐based multimodel structure can be used satisfactorily as the prediction model of model‐based flight control systems for long prediction horizons.

Intelligent vehicle lateral tracking control based on multiple model prediction

A new multi-model predictive control (MMPC) algorithm was proposed and applied in an intelligent vehicle lateral tracking control system in this paper, which is better to adapt the intelligent vehicle lateral tracking control under complex multi-conditions. First, the Gustafson–Kessel algorithm was used for the cluster analysis based on the vehicle test data to obtain the clustering center and train the sample data of each typical steering condition. Then, a multi-model structure was constructed by least squares support vector machines, and the sub-models of each category were taken as the prediction model for the application of MPC. Hence, the objective function of multi-objective optimization can be established and the multi-objective optimization problem was solved by the non-dominated sorted genetic algorithm-II algorithm to obtain the optimal control quantity. Finally, the MMPC-based intelligent vehicle lateral tracking control system was used to control the vehicle lateral tracking under three steering conditions, including straight line, normal right turn, and U-turn, through a simulation study in the MATLAB/Simulink environment. By comparing the vehicle trajectory, steering wheel angle, lateral deviation, and lateral angle, the performance of the proposed control method was verified. The experimental analysis results show that the proposed method can track the steering wheel angle of the vehicle reference trajectory under various working conditions. The vehicle lateral deviation value can be controlled in the range of (−1.0 m, 0.5 m). The high-precision lateral tracking control ensures that the yaw rate of the vehicle can track the yaw velocity under the reference driving track and guarantees the driving stability of intelligent vehicles.

Learning salient features to prevent model drift for correlation tracking

Correlation Filter (CF) based algorithms play an important role in the field of Visual Object Tracking (VOT) due to their high accuracy and low computational complexity. While existing CF tracking algorithms suffer performance degradation due to inaccurate object modeling. In this paper, we improve the object modeling accuracy in both CF training stage and target detection procedure to preventing the drift problem. Specifically, we propose a multi-model structure for CF trackers to capture the target appearance changes, where different appearance models are trained with specific samples to catch the salient features of the target and reduce the computational cost. Furthermore, a space filter for detection features is designed to suppress the boundary effect under Gaussian motion prior, which contributes to improving the accuracy of position estimation. We deploy our method to three hand-crafted features based CF trackers to perform real-time visual tracking on popular benchmarks. The experimental results demonstrate the efficacy of our proposed scheme and the efficiency of our trackers. In addition, we provide a comprehensive analysis of the proposed method to facilitate application.

On considering the output in space-filling test signal designs for the identification of dynamic Takagi-Sugeno models

The model-based design of test signals for the identification of dynamical Takagi-Sugeno (TS) fuzzy models is addressed. The multi-model structure is exploited to reduce computational cost. Space-filling designs usually only address the input but the nonlinear behavior of dynamic systems depends on the lagged output in general. This is considered as an additional constraint regarding the test signal design. This contribution investigates whether a control input can be calculated exploiting the local structure and whether considering the output in space-filling designs yields identified models of higher quality.

Observer-based output feedback integral terminal sliding mode control for nonlinear systems: multi-model approach

In this paper we propose an observer-based output feedback integral terminal sliding mode control for nonlinear systems. The state estimation problem is treated by using the Utkin observer under a multi-model structure. The design of the control law is based on the sliding mode theory using multi-model approach. In fact, we propose two types of multi-model integral terminal sliding mode which are sign integral and fraction integral. The objective of such observer-based control algorithms is to estimate the states of the system and to stabilise the output tracking error to zero. The stability proof is demonstrated by Lyapunov approach. A numerical example of denitrification process is presented in order to validate the performances of the proposed control laws.

Observer-based output feedback integral terminal sliding mode control for nonlinear systems: multi-model approach

In this paper we propose an observer-based output feedback integral terminal sliding mode control for nonlinear systems. The state estimation problem is treated by using the Utkin observer under a multi-model structure. The design of the control law is based on the sliding mode theory using multi-model approach. In fact, we propose two types of multi-model integral terminal sliding mode which are sign integral and fraction integral. The objective of such observer-based control algorithms is to estimate the states of the system and to stabilise the output tracking error to zero. The stability proof is demonstrated by Lyapunov approach. A numerical example of denitrification process is presented in order to validate the performances of the proposed control laws.

Predicting fish kills and toxic blooms in an intensive mariculture site in the Philippines using a machine learning model

Harmful algal blooms (HABs) that produce toxins and those that lead to fish kills are global problems that appear to be increasing in frequency and expanding in area. One way to help mitigate their impacts on people's health and livelihoods is to develop early-warning systems. Models to predict and manage HABs typically make use of complex multi-model structures incorporating satellite imagery and frequent monitoring data with different levels of detail into hydrodynamic models. These relatively more sophisticated methods are not necessarily applicable in countries like the Philippines. Empirical statistical models can be simpler alternatives that have also been successful for HAB forecasting of toxic blooms. Here, we present the use of the random forest, a machine learning algorithm, to develop an early-warning system for the prediction of two different types of HABs: fish kill and toxic bloom occurrences in Bolinao-Anda, Philippines, using data that can be obtained from in situ sensors. This site features intensive and extensive mariculture activities, as well as a long history of HABs. Data on temperature, salinity, dissolved oxygen, pH and chlorophyll from 2015 to 2017 were analyzed together with shellfish ban and fish kill occurrences. The random forest algorithm performed well: the fish kill and toxic bloom models were 96.1% and 97.8% accurate in predicting fish kill and shellfish ban occurrences, respectively. For both models, the most important predictive variable was a decrease in dissolved oxygen. Fish kills were more likely during higher salinity and temperature levels, whereas the toxic blooms occurred more at relatively lower salinity and higher chlorophyll conditions. This demonstrates a step towards integrating information from data that can be obtained through real-time sensors into a an early-warning system for two different types of HABs. Further testing of these models through times and different areas are recommended.

Benchmarking the predictive capability of hydrological models for river flow and flood peak predictions across over 1000 catchments in Great Britain

Abstract. Benchmarking model performance across large samples of catchments is useful to guide model selection and future model development. Given uncertainties in the observational data we use to drive and evaluate hydrological models, and uncertainties in the structure and parameterisation of models we use to produce hydrological simulations and predictions, it is essential that model evaluation is undertaken within an uncertainty analysis framework. Here, we benchmark the capability of several lumped hydrological models across Great Britain by focusing on daily flow and peak flow simulation. Four hydrological model structures from the Framework for Understanding Structural Errors (FUSE) were applied to over 1000 catchments in England, Wales and Scotland. Model performance was then evaluated using standard performance metrics for daily flows and novel performance metrics for peak flows considering parameter uncertainty. Our results show that lumped hydrological models were able to produce adequate simulations across most of Great Britain, with each model producing simulations exceeding a 0.5 Nash–Sutcliffe efficiency for at least 80 % of catchments. All four models showed a similar spatial pattern of performance, producing better simulations in the wetter catchments to the west and poor model performance in central Scotland and south-eastern England. Poor model performance was often linked to the catchment water balance, with models unable to capture the catchment hydrology where the water balance did not close. Overall, performance was similar between model structures, but different models performed better for different catchment characteristics and metrics, as well as for assessing daily or peak flows, leading to the ensemble of model structures outperforming any single structure, thus demonstrating the value of using multi-model structures across a large sample of different catchment behaviours. This research evaluates what conceptual lumped models can achieve as a performance benchmark and provides interesting insights into where and why these simple models may fail. The large number of river catchments included in this study makes it an appropriate benchmark for any future developments of a national model of Great Britain.