Longer Prediction Horizons Research Articles

Comprehensive Optimal Network Scheduling Strategies for Wireless Control Systems

Although wireless control is one of the key technologies for future industries, most wireless networks are only used for monitoring. When wireless networks are applied to transmit control commands, the uncertain link qualities and limited network resources may destroy the performance of multi-loop control systems. Hence, it is critical to allocate these resources to optimize the control performance as the network condition changes and plants evolve. This article presents comprehensive optimal scheduling strategies for wireless control systems based on adaptive dynamic programming. First, we propose an effective adaptive dynamic programming scheduling (ADPS) strategy to solve the optimal scheduling problem based on the single-step control performance at runtime while significantly reducing computational complexity. Moreover, to overcome the “short-sightedness” of single-step performance prediction, we extend ADPS to ADPS-m ( m ulti-step prediction), which optimizes multi-step performance by incorporating a longer-horizon evolution of the plants. Furthermore, we propose ADPS-H ( H eterogeneous flow scheduling) to support heterogeneous flows with different data rates and sizes and ADPS-H-m ( m ulti-step prediction for H eterogeneous flow scheduling), which schedules heterogeneous flows in a longer prediction horizon. We prove that all these scheduling strategies can achieve optimality and stability under mild assumptions. Extensive experiments integrating TOSSIM and MATLAB/Simulink are performed to evaluate all of the proposed methods in case studies of four- and ten-loop control systems. The simulation results demonstrate that these strategies can effectively improve the control performance at lower computing costs under both cyber and physical disturbances. Under the noise level of \(-\) 76 dBm, for the four-loop case, ADPS achieves the same control performance as the linear programming while saving 99.5% of the execution time. ADPS-m further improves the control performance by up to 27.0% compared with ADPS at the prediction horizon of 3, and ADPS-H-m improves the performance by up to 32.3% and 8.4% compared with round-robin and ADPS-H, respectively. The ten-loop case indicates the effectiveness and scalability of the proposed approaches.

Forecasting wastewater flows and pollutant loads: A comparison of data-driven models within the urban water system framework

The ability to forecast key features of wastewater treatment plant (WWTP) influent, is emerging as an important tool to enable advanced WWTP operational strategies. Various data driven models have been reported in the scientific literature for WWTP influent forecast however, there has been no quantitative comparison of them against each other. The present study is the first to undertake this comparison utilising a high-frequency reference dataset generated from the Urban Water System model. Specifically, the performance of autoregressive models was compared against time-delay networks, nonlinear autoregressive networks and long short-term memory networks. Time-delay networks were generally found to outperform the other tested methods, although the reliability of the generated forecasts decreases as the prediction horizon exceeds one hour. While longer prediction horizons would be desirable, there is a trade-off between model accuracy and overall optimisation of plant operation. This study also highlights the challenge of dealing with both concentration and flow variations suggesting the need for future research to analyse the separate impacts of flow and concentration variations on model performance.

A Comparative Analysis of ARX and ANFIS Models for Tumor Growth Prediction Under Single and Multi-agent Chemotherapy.

Despite the advances in oncology and cancer treatment over the past decades, cancer remains one of the deadliest diseases. This study focuses on further understanding the complex nature of cancer by using mathematical tumor modeling to understand, capture as best as possible, and describe its complex dynamics under chemotherapy treatment. Focusing on autoregressive with exogenous inputs, i.e., ARX, and adaptive neuro-fuzzy inference system, i.e., ANFIS, models, this work investigates tumor growth dynamics under both single and combination anticancer agent chemotherapy treatments using chemotherapy treatment data on xenografted mice. Four ARX and ANFIS models for tumor growth inhibition were developed, estimated, and evaluated, demonstrating a strong correlation with tumor weight data, with ANFIS models showing superior performance in handling the multi-agent tumor growth complexities. These findings suggest potential clinical applications of the ANFIS models through further testing. Both types of models were also tested for their prediction capabilities across different chemotherapy schedules, with accurate forecasting of tumor growth up to five days in advance. The use of adaptive prediction and sliding (moving) data window techniques allowed for continuous model updating, ensuring more robust predictive capabilities. However, long-term forecasting remains a challenge, with accuracy declining over longer prediction horizons. While ANFIS models showed greater reliability in predictions, the simplicity and rapid deployment of ARX models offer advantages in situations requiring immediate approximations. Future research with larger, more diverse datasets and by exploring varying model complexities is recommended to improve the models' reliability and applicability in clinical decision-making, thereby aiding the development of personalized chemotherapy regimens.

Risk calculator for incident atrial fibrillation across a range of prediction horizons

The increasing burden of atrial fibrillation (AF) emphasizes the need to identify high-risk individuals for enrolment in clinical trials of AF screening and primary prevention. We aimed to develop prediction models to identify individuals at high-risk of AF across prediction horizons from 6-months to 10-years. We used secondary-care linked primary care electronic health record data from individuals aged ≥30 years without known AF in the UK Clinical Practice Research Datalink-GOLD dataset between January 2, 1998 and November 30, 2018; randomly divided into derivation (80%) and validation (20%) datasets. Models were derived using logistic regression from known AF risk factors for incident AF in prediction periods of 6 months, 1-year, 2-years, 5-years, and 10-years. Performance was evaluated using in the validation dataset with bootstrap validation with 200 samples, and compared against the CHA2DS2-VASc and C2HEST scores. Of 2,081,139 individuals in the cohort (1,664,911 in the development dataset, 416,228 in the validation dataset), the mean age was 49.9 (SD 15.4), 50.7% were women, and 86.7% were white. New cases of AF were 7,386 (0.4%) within 6 months, 15,349 (0.7%) in 1 year, 38,487 (1.8%) in 5 years, and 79,997 (3.8%) by 10 years. Valvular heart disease and heart failure were the strongest predictors, and association of hypertension with AF increased at longer prediction horizons. The optimal risk models incorporated age, sex, ethnicity, and 8 comorbidities. The models demonstrated good-to-excellent discrimination and strong calibration across prediction horizons (AUROC, 95%CI, calibration slope: 6-months, 0.803, 0.789-0.821, 0.952; 1-year, 0.807, 0.794-0.819, 0.962; 2-years, 0.815, 0.807-0.823, 0.973; 5-years, 0.807, 0.803-0.812, 1.000; 10-years 0.780, 0.777-0.784, 1.010), and superior to the CHA2DS2-VASc and C2HEST scores. The models are available as a web-based FIND-AF calculator. The FIND-AF models demonstrate high discrimination and calibration across short- and long-term prediction horizons in 2 million individuals. Their utility to inform trial enrolment and clinical decisions for AF screening and primary prevention requires further study.

Forecasting influenza incidence as an ordinal variable using machine learning

Background: Many mechanisms contribute to the variation in the incidence of influenza disease, such as strain evolution, the waning of immunity and changes in social mixing. Although machine learning methods have been developed for forecasting, these methods are used less commonly in influenza forecasts than statistical and mechanistic models. In this study, we applied a relatively new machine learning method, Extreme Gradient Boosting (XGBoost), to ordinal country-level influenza disease data. Methods: We developed a machine learning forecasting framework by adopting the XGBoost algorithm and training it with surveillance data for over 32 countries between 2010 and 2018 from the World Health Organisation’s FluID platform. We then used the model to predict incidence 1- to 4-week ahead. We evaluated the performance of XGBoost forecast models by comparing them with a null model and a historical average model using mean-zero error (MZE) and macro-averaged mean absolute error (mMAE). Results: The XGBoost models were consistently more accurate than the null and historical models for all forecast time horizons. For 1-week ahead predictions across test sets, the mMAE of the XGBoost model with an extending training window was reduced by 78% on average compared to the null model. Although the mMAE increased with longer prediction horizons, XGBoost models showed a 62% reduction in mMAE compared to the null model for 4-week ahead predictions. Our results highlight the potential utility of machine learning methods in forecasting infectious disease incidence when that incidence is defined as an ordinal variable. In particular, the XGBoost model can be easily extended to include more features, thus capturing complex patterns and improving forecast accuracy. Conclusion: Given that many natural extreme phenomena are often described on an ordinal scale when informing planning and response, these results motivate further investigation of using similar scales for communicating risk from infectious diseases.

Assessment of Seasonal Stochastic Local Models for Glucose Prediction without Meal Size Information under Free-Living Conditions.

Accurate blood glucose (BG) forecasting is key in diabetes management, as it allows preventive actions to mitigate harmful hypoglycemic/hyperglycemic episodes. Considering the encouraging results obtained by seasonal stochastic models in proof-of-concept studies, this work assesses the methodology in two datasets (open-loop and closed-loop) recorded in free-living conditions. First, similar postprandial glycemic profiles are grouped together with fuzzy C-means clustering. Then, a seasonal stochastic model is identified for each cluster. Finally, real-time BG forecasting is performed by weighting each model's prediction. The proposed methodology (named C-SARIMA) is compared to other linear and nonlinear black-box methods: autoregressive integrated moving average (ARIMA), its variant with input (ARIMAX), a feed-forward neural network (NN), and its modified version (NN-X) fed by BG, insulin, and carbohydrates (timing and dosing) information for several prediction horizons (PHs). In the open-loop dataset, C-SARIMA grants a median root-mean-squared error (RMSE) of 20.13 mg/dL (PH = 30) and 27.23 mg/dL (PH = 45), not significantly different from ARIMA and NN. Over a longer PH, C-SARIMA achieves an RMSE = 31.96 mg/dL (PH = 60) and RMSE = 33.91 mg/dL (PH = 75), significantly outperforming the ARIMA and NN, without significant differences from the ARIMAX for PH ≥ 45 and the NN-X for PH ≥ 60. Similar results hold on the closed-loop dataset: for PH = 30 and 45 min, the C-SARIMA achieves an RMSE = 21.63 mg/dL and RMSE = 29.67 mg/dL, not significantly different from the ARIMA and NN. On longer PH, the C-SARIMA outperforms the ARIMA for PH > 45 and the NN for PH > 60 without significant differences from the ARIMAX for PH ≥ 45. Although using less input information, the C-SARIMA achieves similar performance to other prediction methods such as the ARIMAX and NN-X and outperforming the CGM-only approaches on PH > 45min.

Multi-Purpose, Multi-Step Deep Learning Framework for Network-Level Traffic Flow Prediction

This study proposes a data fusion and deep learning (DL) framework that learns high-level traffic features from network-level images to predict large-scale, multi-route, speed and volume of connected vehicles (CVs). We present a scalable and parallel method of processing statewide CVs’ trajectory data that leads to real-time insights on the micro-scale in time and space (two-dimensional (2D) arrays) on graphics processing unit (GPUs) using the Nvidia rapids framework and dask parallel cluster, which provided a 50× speed-up in the data extraction, transform and load (ETL). A UNet model is then applied to perform feature extraction and multi-route speed and volume channels over a multi-step prediction horizon. The accuracy and robustness of the proposed model are evaluated by taking different road types, times of day and image snippets and comparing the model to benchmarks: Convolutional Long–Short-Term Memory (ConvLSTM) and a historical average (HA). The results show that the proposed model outperforms benchmarks with an average improvement of 15% over ConvLSTM and 65% over the HA. Comparing the image snippets from each prediction model to the actual image shows that image textures were highly similar in UNet to the benchmark models used. UNet’s dominance in performing image predictions was also evident in multi-step forecasting, where the increase in errors was relatively minimal over longer prediction horizons.

Cryptocurrency News Website with Prediction

We analyze the predictability of the bitcoin market across prediction horizons ranging from 1 to 60 min. In doing so, we test various machine learning models and find that, while all models outperform a random classifier, recurrent neural networks and gradient boosting classifiers are especially well-suited for the examined prediction tasks. We use a comprehensive feature set, including technical, blockchain-based, sentiment-/interest-based, and asset-based features. Our results show that technical features remain most relevant for most methods, followed by selected blockchain- based and sentiment-/interest-based features. Additionally, we find that predictability increases for longer prediction horizons. Although a quantile-based long-short trading strategy generates monthly returns of up to 39% before transaction costs, it leads to negative returns after taking transaction costs into account due to the particularly short holding periods. Cryptocurrencies, which the Bitcoin is the most remarkable one, have allured substantial awareness up to now, and they have encountered enormous instability in their price. While some studies utilize conventional statistical and econometric ways to uncover the driving variables of Bitcoin's prices, experimentation on the advancement of predicting models to be used as decision support tools in investment techniques is.

Spatiotemporal prediction of COVID-19 cases using inter- and intra-county proxies of human interactions

Measurements of human interaction through proxies such as social connectedness or movement patterns have proved useful for predictive modeling of COVID-19, which is a challenging task, especially at high spatial resolutions. In this study, we develop a Spatiotemporal autoregressive model to predict county-level new cases of COVID-19 in the coterminous US using spatiotemporal lags of infection rates, human interactions, human mobility, and socioeconomic composition of counties as predictive features. We capture human interactions through 1) Facebook- and 2) cell phone-derived measures of connectivity and human mobility, and use them in two separate models for predicting county-level new cases of COVID-19. We evaluate the model on 14 forecast dates between 2020/10/25 and 2021/01/24 over one- to four-week prediction horizons. Comparing our predictions with a Baseline model developed by the COVID-19 Forecast Hub indicates an average 6.46% improvement in prediction Mean Absolute Errors (MAE) over the two-week prediction horizon up to 20.22% improvement in the four-week prediction horizon, pointing to the strong predictive power of our model in the longer prediction horizons.

An Advanced Angular Velocity Error Prediction Horizon Self-Tuning Nonlinear Model Predictive Speed Control Strategy for PMSM System

In nonlinear model predictive control (NMPC), higher accuracy can be obtained with a shorter prediction horizon in steady-state, better dynamics can be obtained with a longer prediction horizon in a transient state, and calculation burden is proportional to the prediction horizon which is usually pre-selected as a constant according to dynamics of the system with NMPC. The minimum calculation and prediction accuracy are hard to ensure for all operating states. This can be improved by an online changing prediction horizon. A nonlinear model predictive speed control (NMPSC) with advanced angular velocity error (AAVE) prediction horizon self-tuning method has been proposed in which the prediction horizon is improved as a discrete-time integer variable and can be adjusted during each sampling period. A permanent magnet synchronous motor (PMSM) rotor position control system with the proposed strategy is accomplished. Tracking performances including rotor position Integral of Time-weighted Absolute value of the Error (ITAE), the maximal delay time, and static error are improved about 15.033%, 23.077%, and 10.294% respectively comparing with the conventional NMPSC strategy with a certain prediction horizon. Better disturbance resisting performance, lower weighting factor sensitivities, and higher servo stiffness are achieved. Simulation and experimental results are given to demonstrate the effectiveness and correctness.

Pedestrian Crossing Intention Prediction at Red-Light Using Pose Estimation

Pedestrians’ red-light crossing can present a threat to traffic safety. Among all the existing work related to pedestrian’s red-light crossing, there are few studies using trajectory data in time sequence. This paper uses pose estimation (keypoint detection) to generate pedestrians’ variables from CCTV videos. Four machine learning models are used to predict pedestrians’ crossing intention at intersections’ red-light. The best model achieves an accuracy of 0.920 and AUC value of 0.849, with data from three intersections. Different prediction horizons (up to 4 sec) are used. With longer prediction horizons, the sample size gets smaller, which partially leads to worse model performance. However, the performance with prediction horizon up to 2 sec is still good (AUC value as 0.841). It is found that keypoint variables such as the angles between ankle and knee (left side) and elbow and shoulder (right side) are important. This model can be further implemented in the Infrastructure-to-Vehicle (I2V) applications and thus prevent accidents due to pedestrians’ red-light crossing by issuing warnings to drivers.

A user-friendly algorithm for adaptive closed-loop phase-locked stimulation

BackgroundClosed-loop phase-locked stimulation experiments are rare due to the unavailability of user-friendly algorithms and devices. Our goal is to provide an algorithm for the detection of oscillatory activity in local field potentials (LFPs) and phase prediction, which is user-friendly and robust to non-stationarities in LFPs of behaving animals. New methodWe propose an algorithm that only requires specification of the frequency range within which oscillatory episodes are tracked. Frequency-specific detection thresholds and filter parameters are adjusted automatically based on the short-time LFP power spectrum. Estimates of instantaneous frequency and instantaneous phase are used for phase extrapolation, taking advantage of Bayesian estimation. We used real LFP signals, recorded from a variety of different species and different brain areas, as well as artificial LFP signals with known properties to assess the detection and prediction performance of our algorithm and three previously published reference algorithms under various conditions. Results and comparison with existing methodsOur algorithm, while significantly more user-friendly than previous approaches, provides a solid detection and prediction performance over a wide range of realistic conditions and, in many cases, has a longer prediction horizon than the reference algorithms. Due to its ability to adjust to changes in the signal, the algorithm is well-prepared to deal with non-stationarities in oscillation frequency, even in the presence of multiple oscillation components. ConclusionsWe have created a universal algorithm for oscillation detection and phase prediction, which performs well and is user-friendly at the same time, making closed-loop phase-locked stimulation experiments easier to accomplish.

Innovate or die: Corporate innovation and bankruptcy forecasts

We investigate the relationship between a firm’s innovation performance and its probability of bankruptcy. Estimating the discrete hazard model with a comprehensive set of bankruptcies spanning the period of 1980–2009, we find several previously neglected innovation-based variables are important determinants of bankruptcy probability, especially for firms belonging to technology-intensive industries. R&D productivity demonstrates persistent significance across different prediction horizons while the predictive power of patent count becomes larger and more significant at longer prediction horizons. We also find that a firm’s organization capital intensity correlates positively with future bankruptcy.

Model predictive control of microgrids for real-time ancillary service market participation

This study develops two model predictive control approaches to optimize microgrid dispatch, one with participation in real-time ancillary service markets and the other without participation. Results are compared to a baseline logic-based control with case study data taken from a grid-tied 326 kW solar photovoltaic, 634 kW/634 kWh battery, and 350 kW diesel generator microgrid portfolio designed to serve an office building. Annual performance evaluations show that model predictive control algorithms can reduce operating expenses by up to 13.73% when compared to logic-based controls, and through participation in ancillary service markets, model predictive control can reduce net operating expenses by up to 23.47%. Revenue from ancillary service equated to 12.03% of operating costs, with approximately two-thirds of revenue from spinning reserve and one-third from non-spinning reserve. Model predictive control with ancillary services maintained battery state of charge an average of 38.78% higher than batteries dispatched by model predictive control without market participation. This reduced battery cycling losses, minimized battery operation and maintenance expenses, and improved battery lifetime. Sensitivity analyses indicate that model predictive control with more granular time steps and longer prediction horizons changes the dispatch schedule to further reduce operating costs. Intraday simulations indicate that both model predictive control algorithms can adapt to differences in environmental conditions and pricing signals to minimize operational costs. This generalizable finding suggests the inherent modularity, scalability, and robustness of the proposed algorithms can benefit a variety of microgrid configurations and use cases.

Bayesian neural networks for flight trajectory prediction and safety assessment

Safety, as the most important concern in civil aviation, needs to be maintained at an acceptable level at all times in the air transportation system. This paper aims to increase en-route flight safety through the development of deep learning models for trajectory prediction, where model prediction uncertainty is characterized following a Bayesian approach. The proposed methodology consists of four steps. In the first step, a large volume of raw messages in Flight Information Exchange Model (FIXM) format streamed from Federal Aviation Administration are processed with a distributed computing engine Apache Spark to extract trajectory information in an efficient manner. In the second step, two types of deep learning models are trained to predict flight trajectory from different perspectives. Specifically, deep feedforward neural networks (DNN) are trained to make a one-step-ahead prediction on the deviation along latitude and longitude between the actual flight trajectory and target flight trajectory. In parallel, deep Long Short-Term Memory (LSTM) neural networks are trained to make longer-term predictions on the flight trajectory over several subsequent time instants. The DNN model is more accurate but has a single-step prediction horizon, whereas the LSTM model is less accurate but longer prediction horizon. Therefore, in the third step, the two different types of deep learning models are blended together to create a multi-fidelity prediction. After quantifying the discrepancy between the two model predictions in the current time instant, the DNN prediction is used to correct the LSTM prediction of flight trajectory along subsequent time instants accordingly. The multi-fidelity approach is expanded to multiple flights, and is then used to assess safety based on horizontal and vertical separation distance between two flights. Computational results illustrate the promising performance of the blended model in predicting the flight trajectory and assessing en-route flight safety.

Prediction Horizons Optimized Nonlinear Predictive Control for Permanent Magnet Synchronous Motor Position System

In this article, a kind of predictive control (PC) for permanent magnet synchronous motor (PMSM) rotor position control is introduced, and more stable and rapid performances are achieved comparing with other PCs. Nonlinear predictive control (NLPC) is shown to outperform the linear in terms of performance for nonlinear controlled plants because of containing nonlinear part and longer prediction horizons. A novel direct speed NLPC scheme for PMSM is presented based on this strategy. The response and reference signals are predicted with Taylor series expansion to guarantee the accuracy of the system. Prediction horizons are selected according to Shannon with its maximum value to eliminate steady-state error and delay between actual and reference signals. Furthermore, the proposed method is designed and analyzed in detail, and tested by simulations and experiments. Results with respect to the rotor position trace tracking, control performance, and weighting factors sensitivity demonstrate the efficiency of the proposed method.

Seasonal Local Models for Glucose Prediction in Type 1 Diabetes.

Linear empirical dynamic models have been widely used for blood glucose prediction and risks prevention in people with type 1 diabetes. More accurate blood glucose prediction models with longer prediction horizon (PH) are desirable to enable warnings to patients about imminent blood glucose changes with enough time to take corrective actions. In this study, a blood glucose prediction method is developed by integrating the predictions of a set of seasonal local models (each of them corresponding to different glucose profiles observed along historical data). In the modeling step, the number of sets and their corresponding glucose profiles characteristics are obtained by clustering techniques (Fuzzy C-Means). Then, Box-Jenkins methodology is used to identify a seasonal model for each set. Finally, blood glucose predictions of local models are integrated using different techniques. The proposed method is tested by using 18 60-h closed-loop experiments (including different exercise types and artificial pancreas strategies) and achieving mean absolute percentage error (MAPE) of 2.94%, 3.89%, 5.41%, 6.29% and 8.66% for 15-, 30-, 45-, 60-, and 90-min PHs, respectively.

Finite-Control-Set Model Predictive Control for DFIG Wind Turbines

This paper presents a time efficient finite-control-set model predictive control (FCS-MPC) scheme for the doubly fed induction generator system. In this scheme, the switching states of the rotor side converter are directly taken as control inputs. This way, the optimized control action can be directly applied to the converter. Compared with the existing FCS-MPC approaches, the salient feature of the proposed scheme is the reduction of the computation time. By introducing a set of augmented decision variables, the original intractable binary quadratic programming problem in FCS-MPC can be analytically transformed to a binary linear programming problem, which can be solved efficiently. By this means, the computation time of the proposed scheme is much less than that of the existing schemes. This reduction in computation time enables FCS-MPC with longer prediction horizons, thus yielding better control performance. Note to Practitioners —This paper was motivated by the problem of improving the control performance for the doubly fed induction generator (DFIG). Because of its advantages, such as high energy efficiency, low acoustic noise, variable speed operation, and reduced converter rating, DFIG has emerged as a promising solution for the wind power generation. However, existing DFIG controllers have limitations, such as difficult to tune the parameters, inefficient to handle constraints, and do not consider the discrete operation of the power converters. To overcome these limitations, this paper proposes a new DFIG control scheme based on finite-control-set model predictive control (FCS-MPC), which has the abilities of online optimal control, explicitly handling constraints, and directly generating the switching signals for the power converters. A computationally efficient algorithm is proposed to reformulate the FCS-MPC problem as a linear program, thereby significantly reducing the computational efforts and rendering the proposed scheme more practical for implementation. Simulation results validate the effectiveness of the proposed control scheme. This scheme can be implemented in the field-programmable gate array, which will be our future work.

Feature selection‐based approach for urban short‐term travel speed prediction

This study proposes a feature selection-based approach to identify reasonable spatial-temporal traffic patterns related to the target link, in order to improve the online-prediction performance. The prediction task is composed of two steps: one hybrid intelligent algorithm-based feature selector (FS) is proposed to optimise original state vectors, which are designed empirically during the offline process and optimised state vectors are employed to carry out the online prediction. Numerical experiments by three non-parametric algorithms are conducted with taxis' global positioning system data in an urban road network of Changsha, China. It is concluded that: (i) under optimised state vectors, the prediction accuracies improve or almost maintain the same; (ii) K-nearest neighbour (KNN) with the simplest state vectors obtains the greatest improvement of prediction performance; (iii) although the performance improvement of e-support vector regression is limited with optimised state vectors, it always outperforms backward-propagation neural network and KNN; and (iv) three non-parametric approaches with optimised state vectors outperform auto-regressive integrated moving average in relatively longer prediction horizons. In conclusion, such FS-based approach is able to improve or guarantee the prediction performance under the remarkably reduced model complexity, and is a promising methodology for short-term traffic prediction.

Model Predictive Control of Distributed Air-Conditioning Loads to Compensate Fluctuations in Solar Power

Flexible loads such as residential air-conditioners (ACs) can be directly controlled to provide demand-side regulation and balance services in electricity grids. Large aggregations of ACs offer a resource akin to that of a distributed energy storage system, which may be used to compensate fluctuations in the power output of local renewable energy generation. This paper formulates distributed and centralized model predictive control (MPC) strategies to balance fluctuations in solar power generation by directly controlling the aggregate demand of clusters of distributed residential ACs. The proposed receding-horizon control strategies rely on a new second-order linear time-varying model for the aggregate demand response of a population of heterogeneous ACs to changes in thermostat setpoints under varying ambient temperature. The performance of the proposed MPC strategies is analyzed in a numerical simulation study implementing AC demand tracking of 1-min fluctuations in actual photovoltaic capacity based on persistence and sky imager short-term solar forecasts. The results show that distributed and centralized MPC strategies achieve comparable performance, with better performance of persistence forecasts in a shorter prediction horizon, and better performance with sky imager forecasts on a longer prediction horizon.