Autoregressive Exogenous Input Research Articles

Using nonlinear auto-regressive with exogenous input neural network (NNARX) in blood glucose prediction

BackgroundPredicting of future blood glucose (BG) concentration is important for diabetes control. Many automatic BG monitoring or controlling systems use BG predictors. The accuracy of the prediction for long prediction time is a major factor affecting the performance of the control system. The predicted BG can be used for glycemia management in the form of early hypoglycemic/hyperglycemic alarms or adjusting insulin injections. Recent developments in continuous glucose monitoring (CGM) devices open new opportunities for glycemia management of diabetic patients. Many of those systems need prediction for long prediction horizons to avoid going through hypo or hyperglycemia.MethodsIn this article a nonlinear autoregressive exogenous input neural network (NNARX) is proposed to predict the glucose concentration for longer prediction horizons (PHs) than that was obtained previously with an established recurrent neural network (RNN). The proposed NNARX is a modified version from our previously published RNN with different initialization and building technique but has the same architecture. The modification is based on starting with building nonlinear autoregressive exogenous input model using MATLAB and train it, then close the loop to get NNARX network.ResultsThe results of using the proposed NNARX indicate that the proposed NNARX is better in prediction and stability than unmodified RNN as PH becomes higher than 45 minutes.ConclusionsModification in RNN building extends the ability of the prediction till 100 minutes. It performs statistically significant improvements in the FIT and RMSE values for 100 minutes prediction. It also decreases root mean squared error (RMSE) for both 45 and 60 minutes of prediction.

Efficient model predictive control of boiler coal combustion based on NARX neutral network

During coal-fired power generation, uniform combustion temperature in the boiler is desired which will benefit both economical efficiency and pollution reduction. To this end, a model predictive control (MPC) algorithm based on the Nonlinear Auto-Regressive Exogenous Inputs (NARX) neural network and KS-function is proposed, and the uniform combustion in the boiler is realized by controlling the opening travel of secondary windgates. In the modeling process, a multi-input and multi-output(MIMO) NARX neural network is developed using the historical data of the real system The NARX neural network is then used to predict the state variables, and the optimal control input is achieved by applying sequential quadratic programming (SQP), comparing with linear MPC the mean temperature difference is reduced by 64.2%. In addition, this paper proposes a new method to reduce the computational time of the online optimization process based on KS-function, which greatly accelerates the searching speed of SQP by 67.3%. The proposed MPC algorithm is applied to a 660 MW power generating unit. The results show that by applying the proposed algorithm, the temperature difference in the boiler is kept within 100 °C, the average coal consumption of the power plant is reduced by 5.71 g/kWh, and the NOx emission is reduced to 23.84 mg/m3. It can be concluded that the proposed algorithm greatly improves the economical efficiency of the power plant and reduces the emission of pollutants.

Careful feedback active noise and vibration control algorithm robust to large secondary path changes

Feedback active noise and vibration control (ANVC) reduces narrowband noise and vibration when there is no reference signal. However, most ANVC algorithms require a secondary path (plant) model and can become unstable when this model is inaccurate. This work proposes an algorithm less sensitive to plant modeling errors and can cope with large and fast plant changes. This is achieved by modeling the plant and the noise using an autoregressive exogenous input (ARX) model. This model allows obtaining the expected values of the future residual noise and determining the optimal control signal. Since the model coefficient uncertainty (variance) is considered when calculating the expected value, the resulting controller is a careful controller. This allows solving problems due to significant model estimation errors when information is insufficient to estimate the full ARX model.

Hourly Water Level Forecasting in an Hydroelectric Basin Using Spatial Interpolation and Artificial Intelligence.

In this work, a new hydroelectric basin modelling approach is described and applied to the Pontecosi basin, Italy. Several types of data sources were used to learn the model: a number of weather stations, satellite observations, the reanalysis dataset, and basin data. With the goal of predicting the water level of the basin, the model was composed by three cascade modules. Firstly, different spatial interpolation methods, such as Kriging, Radial Basis Function, and Natural Neighbours, were compared and applied to interpolate the weather stations data nearby the basin area to infer the main environmental variables (air temperature, air humidity, precipitation, and wind speed) in the basin area. Then, using these variables as inputs, a neural network was trained to predict the mean soil moisture concentration over the area, also to improve the low availability due to satellite orbits. Finally, a non-linear auto regressive exogenous input (NARX) model was trained to simulate the basin level with different prediction horizons, using the data from the previous modules and past basin data (water level, discharge flow rate, and turbine flow rate). Accurate predictions of the basin water level were achieved within 1 to 6 h ahead, with mean absolute errors (MAE) between 2 cm and 10 cm, respectively.

Real-Time Power Quality Enhancement in a Hybrid Micro-Grid Using Nonlinear Autoregressive Neural Network

The extensive use of renewable energy sources (RESs) in energy sectors plays a vital role in meeting the present energy demand. The widespread utilization of allocated resources leads to multiple usages of converters for synchronization with the power grid, introducing poor power quality. The integration of distributed energy resources produces uncertainties which are reflected in the distribution system. The major power quality problems such as voltage sag/swell, voltage unbalancing, poor power factor, harmonics distortion (THD), and power transients appear during the transition of micro-grids (MGs). In this research, a single micro-grid is designed with PVs, wind generators, and fuel cells as distributed energy resources (DERs). A nonlinear auto regressive exogenous input neural network (NARX-NN) controller has been investigated in this micro-grid in order to maintain the above power quality issues within the specific standard range (IEEE/IEC standards). The performance of the NARX-NN controller is compared with PID and fuzzy-PID controllers. The single micro-grid is extended to design a three-phase large-scale realistic micro-grid structure to test the feasibility of the proposed controller. The realistic micro-grid is verified through addition of line-impedance, communication delay, demand response, and off-nominal situations. The proposed controller is also validated by simulating different test scenarios using MATLAB/Simulink and TMS320-based processor-in-loop (PIL) for real-time implementation.

Quantized‐output‐based least squares of ARX systems

SummaryThis paper considers the estimates of parameters of stochastic autoregressive exogenous input (ARX) systems with quantized data. The quantized error caused by quantized data brings difficulties to the parameter estimation of ARX systems. Therefore, I start with the boundedness of quantized error and design the size of quantized error thanks to good properties of the least squares. Then the influences of the noise of the system itself and the bounded noise caused by quantized error on the parameter estimation error are analyzed, respectively. The inputs of the system are designed to make the system satisfy the persistent excitation condition, so as to ensure the boundedness of the parameter estimation error. A numerical example is given to demonstrate the theorem.

Online unsupervised detection of structural changes using train–induced dynamic responses

This paper exploits unsupervised data-driven structural health monitoring (SHM) in order to propose a continuous online procedure for damage detection based on train-induced dynamic bridge responses, taking advantage of the large-magnitude loading for enhancing sensitivity to small-scale structural changes. While such large responses induced by trains might create more damage-sensitive information in the measured response, it also amplifies the effects on those measurements from the environment. Thus, one of the biggest contributions herein is a methodology that exploits the large bridge responses induced by train passage while rejecting the confounding influences of the environment in such a way that false positive detections are mitigated. Furthermore, this research work introduces an adaptable confidence decision threshold that further improves damage detection over time. To ensure an online continuous assessment, a hybrid combination of autoregressive exogenous input (ARX) models, principal components analysis (PCA), and clustering algorithms was sequentially applied to the monitoring data, in a moving window process. A comparison between the performance obtained from autoregressive (AR) and ARX models as feature extractors was conducted, and it was concluded that ARX models lead to increased sensitivity to damage due to their ability to capture cross information between the sensors. The PCA proved its importance and effectiveness in removing observable changes induced by variations in train speed or temperature without the need to measure them, and the clustering methods allowed for an automatic classification of the damage-sensitive features. Since it was not possible to introduce damage to the bridge, several structural conditions were simulated with a highly reliable digital twin of the Sado Bridge, tuned with experimental data acquired from a SHM system installed on site, in order to test and validate the efficiency of the proposed procedure. The strategy proved to be robust when detecting a comprehensive set of damage scenarios with a false detection incidence of 2%. Moreover, it showed sensitivity to smaller damage levels (earlier in life), even when it consists of small stiffness reductions that do not impair structural safety and are imperceptible in the original signals.

Time-varying parameter estimation and adaptive Kalman filter in computer aided control application

Estimating the unknown parameter vector of the system model is the most important problems in system identification. Especially in cases where the system's parameters are time-variable, it is observed that estimations obtained using estimator have deviated from the actual values, and therefore that the estimator must be corrected to some extent. In this paper, some methods for the parameter estimation in cases where a system is modelled with ARX Autoregressive Exogenous Input) are considered. After reviewing the problems, a simulation study has been made on comparing different estimation methods. Corrected (Adaptive) Kalman Filter (CKF) gives results more accurately than Normal Kalman Filter (NKF) for time varying parameter estimation. Moreover, after an introduction to the method of minimum variance feed-back control, using this method and CKF, a heating control is done in computer aided experimental study. CKF ensures that the system is kept under control by correctly estimating the parameter that changes over time.

Decoupling P-NARX models using filtered CPD

Abstract Nonlinear Auto-Regressive eXogenous input (NARX) models are a popular class of nonlinear dynamical models. Often a polynomial basis expansion is used to describe the internal multivariate nonlinear mapping (P-NARX). Resorting to fixed basis functions is convenient since it results in a closed form solution of the estimation problem. The drawback, however, is that the predefined basis does not necessarily lead to a sparse representation of the relationship, typically resulting in very large numbers of parameters. So-called decoupling techniques were specifically designed to reduce large multivariate functions. It was found that, often, a more efficient parameterisation can be retrieved by rotating towards a new basis. Characteristic to the decoupled structure is that, expressed in the new basis, the relationship is structured such that only single-input single-output nonlinear functions are required. Classical decoupling techniques are unfit to deal with the case of single-output NARX models. In this work, this limitation is overcome by adopting the filtered CPD decoupling method of Decuyper et al. (2021b). The approach is illustrated on data from the Sliverbox benchmark: measurement data from an electronic circuit implementation of a forced Duffing oscillator.

PRONÓSTICO HORARIO DE CAUDALES MEDIANTE FILTRO DE KALMAN DISCRETO EN EL RÍO HUAYNAMOTA, NAYARIT, MÉXICO

Debido a los eventos de precipitación extrema provocada por el cambio climático y a la alteración acelerada de las cuencas por el crecimiento poblacional, es importante pronosticar los caudales que generan las cuencas por los eventos de precipitación. El objetivo de este estudio fue predecir caudales horarios en la cuenca del río Huaynamota usando el Filtro de Kalman Discreto (DKF) junto con un modelo autorregresivo con entrada exógena (ARX). Al inicio los parámetros del filtro de Kalman se definen y después se recalculan por periodos definidos, es decir los valores de los parámetros del modelo se actualizan constantemente. El pronóstico de caudales se realizó en seis pasos hacia adelante (L=1, 2, 3, 4, 5 y 6 horas). La cuenca de estudio es parte del río Huaynamota, delimitada por la estación hidrométrica Chapalagana, aguas arriba de la presa Aguamilpa, en Nayarit, México. La cuenca del río Huaynamota es un tributario del río Santiago. Series de datos horarias se emplearon para precipitación y caudal, de agosto a septiembre del 2017. El modelo de pronóstico DKF-ARX mostró índices de eficiencia de Nash-Sutcliffe entre 0.99 y 0.85 con L=1 y L=6, respectivamente. Se concluye que es factible obtener un buen pronóstico de caudales horarios con filtro de Kalman discreto.

Predictive Analysis of Machine Learning Schemes in Forecasting Offshore Wind Speed

High variability of wind in the farm areas causes a drastic instability in the energy markets. Therefore, precise forecast of wind speed plays a key role in the optimal prediction of offshore wind power. In this study, we apply two deep learning models, i.e. Long Short-Term Memory (LSTM) and Nonlinear Autoregressive EXogenous input (NARX), for predicting wind speed over long-range of dependencies. We use a four-month-long wind speed/direction, air temperature, and atmospheric pressure time series (all recorded at 10 m height) from a meteorological mast (Vigra station) in the close vicinity of the Havsul-I offshore area near Ålesund, Norway. While both predictive methods could efficiently predict the wind speed, the LSTM with update generally outperforms the NARX. The NARX suffers from vanishing gradient issue and its performance declines by abrupt variability inherited in the input data during training phase. It is observed that this sensitivity will significantly decrease by integrating, for example, the wind direction at low frequencies in the learning process. Generally, the results showed that the predictive models are robust and accurate in short-term and somewhat long-term forecasting of wind.

Data Analytics for Predicting COVID-19 Cases in Top Affected Countries: Observations and Recommendations.

The outbreak of the 2019 novel coronavirus disease (COVID-19) has adversely affected many countries in the world. The unexpected large number of COVID-19 cases has disrupted the healthcare system in many countries and resulted in a shortage of bed spaces in the hospitals. Consequently, predicting the number of COVID-19 cases is imperative for governments to take appropriate actions. The number of COVID-19 cases can be accurately predicted by considering historical data of reported cases alongside some external factors that affect the spread of the virus. In the literature, most of the existing prediction methods focus only on the historical data and overlook most of the external factors. Hence, the number of COVID-19 cases is inaccurately predicted. Therefore, the main objective of this study is to simultaneously consider historical data and the external factors. This can be accomplished by adopting data analytics, which include developing a nonlinear autoregressive exogenous input (NARX) neural network-based algorithm. The viability and superiority of the developed algorithm are demonstrated by conducting experiments using data collected for top five affected countries in each continent. The results show an improved accuracy when compared with existing methods. Moreover, the experiments are extended to make future prediction for the number of patients afflicted with COVID-19 during the period from August 2020 until September 2020. By using such predictions, both the government and people in the affected countries can take appropriate measures to resume pre-epidemic activities.

A Promising Wavelet Decomposition –NNARX Model to Predict Flood

Flood is a major disaster that happens around the world. It has caused many casualties and massive destruction of property. Estimating the chance of a flood occurring depends on several factors, such as rainfall, the structure and the flow rate of the river. This research used the neural network autoregressive exogenous input (NNARX) model to predict floods. One of the research challenges was to develop accurate models and improve the forecasting model. This research aimed to improve the performance of the neural network model for flood prediction. A new technique was proposed for modelling nonlinear data of flood forecasting using the wavelet decomposition-NNARX approach. This paper discusses the process of identifying the parameters involved to make a forecast as the rainfall value requires the flow rate of the river and its water level. The original data were processed by wavelet decomposition and filtered to generate a new set of data for the NNARX prediction model where the process can be compared. This research compared the performance of the wavelet and the non-wavelet NNARX model. Experimental results showed that the proposed approach had better performance testing results in relation to its counterpart in terms of hourly forecast, with the mean square error (MSE) of 2.0491e-4 m2 compared to 6.1642e-4 m2, respectively. The proposed approach was also studied for long-term forecast up to 5 years, where the obtained MSE was higher, i.e., 0.0016 m2.

Tourism, Accommodations, Food Services, and Regional GDP

The development of the airport city makes it the center of national-regional economic growth. The continuous development provides several alternative revenues to the surrounding areas and affects the price of the urban houses. The Bank of Indonesia has analyzed the development of commercial property since 2016. It covers five types of commercial property, which include offices (rent/sale), retail (rent/sale), apartments (rent/sale), hotels, industrial land, convention halls, and warehouse complexes, in Jabodetabek, Banten, Bandung, Makassar, Medan, Semarang, Surabaya, Balikpapan, Denpasar, and Palembang. The research aims to find variables that affect the contribution of accommodations and food services to the Regional GDP. Here, researchers use an approach in quantitative dynamics modeling that is often used in time series analysis, which is is the Autoregressive Exogenous modeling structure input (ARX). In general, AR is the Auto-Regressive nature of a model, while X is an exogenous input. The initial population of the study is 296 airports, which consists of all those managed by PT. Angkasa Pura I, PT. Angkasa Pura II, and the UPT Air Transportation in the Region. Then, the population is filtered further to 151 airports in Indonesia. The study shows that the variable of departing aircraft, departing passengers, arriving passengers, baggage unloading, cargo unloading, and cargo loading, each had a contribution to the GDP variable. The departing passenger variable had a positive impact on the increase of the variable of accommodation and food services of the Regional GDP

Forecasting of Building Electrical Energy Consumption Using Nonlinear Autoregressive Exogenous Input (NARX) Neural Network Time Series (NNTS) Model

Forecasting of Building Electrical Energy Consumption Using Nonlinear Autoregressive Exogenous Input (NARX) Neural Network Time Series (NNTS) Model

Initialization Approach for Decoupling Polynomial NARX Models Using Tensor Decomposition

The Nonlinear Auto-regressive eXogenous input (NARX) model has been widely used in nonlinear system identification. It’s chief disadvantages are that it is a black-box model that suffers from the curse of dimensionality, in that the number of parameters increases rapidly with the nonlinearity degree. One approach to dealing with these problems involves decoupling the nonlinearity, but this requires solving a non-convex optimization problem. Solving non-convex optimization problems has always been challenging due to the possibility of getting trapped in a sub-optimal local optima. As a result, these kinds of optimization problems are sensitive to the initial solution. Providing an appropriate initial solution can increase the likelihood of finding the globally optimal solution. In this paper, an initialization technique that uses the polynomial coefficients in a full, albeit low order, NARX model is proposed. This technique generates a tensor from the coefficients in the from full polynomial NARX model and applies a tensor factorization in order to generate an appropriate starting point for decoupled polynomial NARX model optimization problem. The proposed technique is applied to nonlinear benchmark problem and the results are promising.

Advanced Adaptive Fault Diagnosis and Tolerant Control for Robot Manipulators

In this paper, an adaptive Takagi–Sugeno (T–S) fuzzy sliding mode extended autoregressive exogenous input (ARX)–Laguerre proportional integral (PI) observer is proposed. The proposed T–S fuzzy sliding mode extended-state ARX–Laguerre PI observer adaptively improves the reliability, robustness, estimation accuracy, and convergence of fault detection, estimation, and identification. For fault-tolerant control in the presence of uncertainties and unknown conditions, an adaptive fuzzy sliding mode estimation technique is introduced. The sliding surface slope gain is significant to improve the system’s stability, but the sliding mode technique increases high-frequency oscillation (chattering), which reduces the precision of the fault diagnosis and tolerant control. A fuzzy procedure using a sliding surface and actual output position as inputs can adaptively tune the sliding surface slope gain of the sliding mode fault-tolerant control technique. The proposed robust adaptive T–S fuzzy sliding mode estimation extended-state ARX–Laguerre PI observer was verified with six degrees of freedom (DOF) programmable universal manipulation arm (PUMA) 560 robot manipulator, proving qualified efficiency in detecting, isolating, identifying, and tolerant control for faults inherent in sensors and actuators. Experimental results showed that the proposed technique improves the reliability of the fault detection, estimation, identification, and tolerant control.

Prediction of short and medium term PM10 concentration using artificial neural networks

Purpose There has been a growing concern about air quality because in recent years, industrial and vehicle emissions have resulted in unsatisfactory human health conditions. There is an urgent need for the measurements and estimations of particulate pollutants levels, especially in urban areas. As a contribution to this issue, the purpose of this paper is to use data from measured concentrations of particulate matter and meteorological conditions for the predictions of PM10. Design/methodology/approach The procedure included daily data collection of current PM10 concentrations for the city of São Carlos-SP, Brazil. These data series enabled to use an estimator based on artificial neural networks. Data sets were collected using the high-volume sampler equipment (VFA-MP10) in the period ranging from 1997 to 2006 and from 2014 to 2015. The predictive models were created using statistics from meteorological data. The models were developed using two neural network architectures, namely, perceptron multilayer (MLP) and non-linear autoregressive exogenous (NARX) inputs network. Findings It was observed that, over time, there was a decrease in the PM10 concentration rates. This is due to the implementation of more strict environmental laws and the development of less polluting technologies. The model NARX that used as input layer the climatic variables and the PM10 of the previous day presented the highest average absolute error. However, the NARX model presented the fastest convergence compared with the MLP network. Originality/value The presentation of a given PM10 concentration of the previous day improved the performance of the predictive models. This paper brings contributions with the NARX model applications.

Identification of Neuronal Networks from Calcium Oscillation Data

Abstract Neuronal network reconstruction is an important problem in neuroscience as it helps understanding neuronal circuit and function. With the advancements of the calcium imaging technique, the dynamic activity of hundreds of neurons can be observed, which also provides a foundation for modelling and inferring network connectivity directly from data. In this paper, the dynamic behavior of each neuron is described using a non-linear autoregressive exogenous input (NARX) model. Since the neurons are inter-connected, the NARX models contain inputs which depend on outputs of the surrounding neurons. The complete model is expressed in a linear-in-parameter form and the network dynamics are identified by obtaining a block-sparse solution given the calcium oscillation data. The solution to the identification problem also provides the network topology simultaneously, i.e., the physical connections between the neurons. To demonstrate the accuracy of the proposed method, two experimental case studies are considered and the results are compared with other network identification methods.

Integration of a Parsimonious Hydrological Model with Recurrent Neural Networks for Improved Streamflow Forecasting

This study applied a GR4J model in the Xiangjiang and Qujiang River basins for rainfall-runoff simulation. Four recurrent neural networks (RNNs)—the Elman recurrent neural network (ERNN), echo state network (ESN), nonlinear autoregressive exogenous inputs neural network (NARX), and long short-term memory (LSTM) network—were applied in predicting discharges. The performances of models were compared and assessed, and the best two RNNs were selected and integrated with the lumped hydrological model GR4J to forecast the discharges; meanwhile, uncertainties of the simulated discharges were estimated. The generalized likelihood uncertainty estimation method was applied to quantify the uncertainties. The results show that the LSTM and NARX better captured the time-series dynamics than the other RNNs. The hybrid models improved the prediction of high, median, and low flows, particularly in reducing the bias of underestimation of high flows in the Xiangjiang River basin. The hybrid models reduced the uncertainty intervals by more than 50% for median and low flows, and increased the cover ratios for observations. The integration of a hydrological model with a recurrent neural network considering long-term dependencies is recommended in discharge forecasting.