No full-text available
To read the full-text of this research,
you can request a copy directly from the authors.
Machine-learning-enabled smart cities
Abstract
A smart city leverages the Internet of Things (IoT) and sensors to collect the available wealth of raw data from various urban surroundings. This huge volume of unstruc- tured data gleaned in real time needs to be effectively analysed and utilised to spot trends, which can give city planners the information that is highly responsive to the needs of the citizens. The massive amount of information presented in this data is difficult to be viewed and processed by humans. Here, the information retrieval via machine learning (ML) helps in extracting knowledge or structured data from the unstructured form by recognising the underlying pattern. It produces a sum- marised tabular output in a relational database, which helps one to optimise the given set of services for enhanced functioning and sustainability of the city, such as pre- dicting parking spots for drivers, helping first responders, and locating dangerous intersections.
The factors responsible for surging interest in ML are powerful computational processing and cost-effective data storage options, which allow training models that gain experience by quickly and accurately analysing huge chunks of complex data. ML combined with the IoT helps to realise the vision of a more livable and resilient city that is capable of quickly responding to the critical challenges prompted by an outrageous urban population, encompassing traffic congestion, environment deterioration, sanitation issues, energy crises, thwart crime, healthcare, and many more. It can automate municipal operations and advance smart city initiatives at large.
In this chapter, a comprehensive list of applications is curated to understand the nuts-and-bolts of ML in the domain of the smart city. The chapter walks through the recent applied examples alongside familiarising with the key research developments in the context of ML-assisted smart cities. Ultimately, the chapter concludes by men- tioning the major challenges faced by the implication of ML as a smart city use case. On that account, we are focusing on various examples of ML in a smart city.
The future Internet of Things (IoT) will have a deep economical, commercial and social impact on our lives. The participating nodes in IoT networks are usually resource-constrained, which makes them luring targets for cyber attacks. In this regard, extensive efforts have been made to address the security and privacy issues in IoT networks primarily through traditional cryptographic approaches. However, the unique characteristics of IoT nodes render the existing solutions insufficient to encompass the entire security spectrum of the IoT networks. Machine Learning (ML) and Deep Learning (DL) techniques, which are able to provide embedded intelligence in the IoT devices and networks, can be leveraged to cope with different security problems. In this paper, we systematically review the security requirements, attack vectors, and the current security solutions for the IoT networks. We then shed light on the gaps in these security solutions that call for ML and DL approaches. We also discuss in detail the existing ML and DL solutions for addressing different security problems in IoT networks. We also discuss several future research directions for ML-and DL-based IoT security.
Electroencephalogram (EEG) is one of the fundamental tools for analyzing the behaviorof brain and particularly helpful for treatment of epilepsy and detection of associated seizures. For long-term recording of EEG signals, current research is heading towards simple, unobtrusive and ambulatorydevices with a small number of channels. The primary contribution of this paper is to assess the performancedifference between the seizure detection results using features from all channels versus only the channelsin/around the temporal region. For this purpose, we develop a supervised seizure detection algorithm thatuses time domain features extracted sequentially for every1-second epoch. By using this algorithm, weobtained sensitivity values of0.95and0.92, specificity values of0.99and0.99and false positive per hourvalues as0.16and0.21for all23channels and10temporal region channels, respectively. These resultsshow that restricting the EEG analysis to temporal region results only in a graceful and gradual degradationof classifier performance. We conclude that EEG ambulatory devices with a montage local to the temporalregion could demonstrate satisfactory performance. This presents a promising way forward for the use ofambulatory devices with compact wearable design.
Smart cities are aimed to efficiently manage growing urbanization, energy consumption, maintain a green environment, improve the economic and living standards of their citizens, and raise the people’s capabilities to efficiently use and adopt the modern information and communication technology (ICT). In the smart cities concept, ICT is playing a vital role in policy design, decision, implementation, and ultimate productive services. The primary objective of this review is to explore the role of artificial intelligence (AI), machine learning (ML), and deep reinforcement learning (DRL) in the evolution of smart cities. The preceding techniques are efficiently used to design optimal policy regarding various smart city-oriented complex problems. In this survey, we present in-depth details of the applications of the prior techniques in intelligent transportation systems (ITSs), cyber-security, energy-efficient utilization of smart grids (SGs), effective use of unmanned aerial vehicles (UAVs) to assure the best services of 5G and beyond 5G (B5G) communications, and smart health care system in a smart city. Finally, we present various research challenges and future research directions where the aforementioned techniques can play an outstanding role to realize the concept of a smart city.
Internet-of-Things (IoT) refers to a massively heterogeneous network formed through smart devices connected to the Internet. In the wake of disruptive IoT with a huge amount and variety of data, Machine Learning (ML) and Deep Learning (DL) mechanisms will play a pivotal role to bring intelligence to the IoT networks. Among other aspects, ML and DL can play an essential role in addressing the challenges of resource management in large-scale IoT networks. In this article, we conduct a systematic and in-depth survey of the ML-and DL-based resource management mechanisms in cellular wireless and IoT networks. We start with the challenges of resource management in cellular IoT and low-power IoT networks, review the traditional resource management mechanisms for IoT networks, and motivate the use of ML and DL techniques for resource management in these networks. Then, we provide a comprehensive survey of the existing ML-and DL-based resource allocation techniques in wireless IoT networks and also techniques specifically designed for HetNets, MIMO and D2D communications, and NOMA networks. To this end, we also identify the future research directions in using ML and DL for resource allocation and management in IoT networks.
With various developments, the concept of the smart city has attracted great attention all over the world. To many, it is a good intelligent response to the needs of people’s livelihoods, environmental protection, public safety, etc. A weather-smart grid is an important component of the smart city, and the health of the weather-smart grid will directly affect the health of the smart city. Efficient and accurate predictions about air quality levels can provide a reliable basis for societal decisions, safety for smart transportation, and weather-related disaster prevention and preparation. To improve the time performance and accuracy of prediction with a large amount of data, this paper proposes an improved decision tree method. Based on an existing method, the model is improved in two aspects: the feature attribute value and the weighting of the information gain. Both accuracy and computational complexity are improved. The experimental results demonstrate that the improved model has great advantages in terms of the accuracy and computational complexity compared with the traditional methods. Additionally, it is more efficient in addressing classification and prediction with a large amount of air quality data. Moreover, it has good prediction ability for future data.
The volume of internet users is increasingly causing transactions on e-commerce to increase as well. We observe the quantity of fraud on online transactions is increasing too. Fraud prevention in e-commerce shall be developed using machine learning, this work to analyze the suitable machine learning algorithm, the algorithm to be used is the Decision Tree, Naive Bayes, Random Forest, and Neural Network. Data to be used is still unbalance. Synthetic Minority Over-sampling Technique (SMOTE) process is to be used to create balance data. Result of evaluation using confusion matrix achieve the highest accuracy of the neural network by 96 percent, random forest is 95 percent, Naïve Bayes is 95 percent, and Decision tree is 91 percent. Synthetic Minority Over-sampling Technique (SMOTE) is able to increase the average of F1-Score from 67.9 percent to 94.5 percent and the average of G-Mean from 73.5 percent to 84.6 percent.
Loan lending has been playing a significant role in the financial world throughout the years. Although it is quite profitable and beneficial for both the lenders and the borrowers. It does, however, carry a great risk, which in the domain of loan lending is referred to as Credit risk. Industry experts and Researchers around the world assign individuals with numerical scores known as credit scores to measure the risk and their creditworthiness. Throughout the years machine learning algorithms have been used to calculate and predict credit risk by evaluating an individual's historical data. This study reviews the present literature on models predicting risk assessment that use machine learning algorithms.
Dealing with air pollution is one of the major environmental challenges in a smart city environment. Real-time monitoring of pollution data enables the metropolitans to analyze the current traffic situation of the city and take their decisions accordingly. Deployment of the Internet of things (IoT) based sensors has considerably changed the dynamics of predicting air quality. Existing research has used different machine learning tools for pollution prediction; however, comparative analysis of these techniques is often required to have a better understanding of their processing time for multiple datasets. In this paper, we have performed the pollution prediction using four advanced regression techniques and have presented a comparative study to analyze the best model for accurately predicting the air quality with reference to data size and processing time. We have used Apache Spark for conducting experiments and performing pollution estimation using multiple available data sets. Mean Absolute error (MAE) and Root Mean Square Error (RMSE) have been used as evaluation criteria for the comparison of regression models. Furthermore, the processing time of each technique by standalone learning and by fitting the hyperparameter tuning on Apache Spark has also been calculated to find the best-fitted model in terms of processing time and least error rate.
This paper conducts a comprehensive study on the application of big data and machine learning in the electrical power grid, introduced through the emergence of the next-generation power system – the smart grid (SG). Connectivity lies at the core of this new grid infrastructure, which is provided by the internet of things (IoT). This connectivity, and constant communication required in this system, also introduced a massive data volume that demands techniques far superior than conventional methods for proper analysis and decision-making. IoT integrated SG system can provide efficient load forecasting and data acquisition technique, along with cost effectiveness. Big data analysis and machine learning techniques are essential to reap these benefits. In the complex connected system of SG, cyber security becomes a critical issue; IoT devices and their data turning into major targets of attacks. Such security concerns and their solutions are also included in this study. Key information obtained through literature review are tabulated in corresponding sections to provide a clear synopsis; and the findings of this rigorous review are listed to give a concise picture of this area of study and promising future fields of academic and industrial research, with current limitations with viable solutions along with their effectiveness.
The smart building concept aims to use smart technology to reduce energy consumption, as well as to improve comfort conditions and users' satisfaction. It is based on the use of smart sensors and software to follow both outdoor and indoor conditions for the control of comfort, and security devices for the optimization of energy consumption. This paper presents a data-based model for indoor temperature forecasting, which could be used for the optimization of energy device use. The model is based on an artificial neural network (ANN), which is validated on data recorded in an old building. The novelty of this work consists of the methodology proposed for the development of a simplified model for indoor temperature forecasting. This methodology is based on the selection of pertinent input parameters after a relevance analysis of a large set of input parameters, including solar radiation outdoor temperature history, outdoor humidity, indoor facade temperature, and humidity. It shows that an ANN-based model using outdoor and facade temperature sensors provides good forecasting of indoor temperatures. This model can be easily used in the optimal regulation of buildings' energy devices.
With the growing penetration of wind power into electric grids, improving wind speed prediction accuracy has become particularly valuable for the exploitation of wind power. In this paper, a novel hybrid strategy based on a three-phase signal decomposition (TPSD) technique, feature extraction (FE) and weighted regularized extreme learning machine (WRELM) is developed for multi-step ahead wind speed prediction. The TPSD including seasonal separation algorithm (SSA), fast ensemble empirical mode decomposition (FEEMD) and variational mode decomposition (VMD) is proposed for the first time to handle the complex and irregular natures of wind speed comprehensively. The FE process is used to capture the useful features of wind speed fluctuations and determine the optimal inputs for a prediction model. The WRELM is employed as a basic predictor for building the prediction model by these selected features. Four real wind speed prediction cases are utilized to evaluate the proposed model, and experimental results verify the effectiveness of the proposed model compared with the benchmark models.
This paper presents a two-level control scheme for the energy management of an electricity prosumer node equipped with controllable loads, local generation, and storage devices. The main control objective is to optimize the prosumer’s energy bill by means of intelligent load shifting and storage control. A generalized tariff model including both volumetric and capacity components is considered, and user preferences as well as all technical constraints are respected. Simulations based on real household consumption data acquired with a sampling period of 1 s are discussed. The proposed control scheme bestows the prosumer node with the flexibility needed to support smart grid use cases such as bill optimization (i.e., local energy trading), control of the profile at the point of connection with the grid, demand response, and reaction to main supply faults (e.g., islanding operation), etc.
The aim of this study is to design and implement soft computing techniques called Support Vector Machine (SVM) and Multilayer Perceptron (MLP) for great management of energy generation based on experimental work. Solar energy could be utilized through thermal systems or Photovoltaics (PV) and it is renewable energy source, environmental friendly and proven globally for a long time. The SVM and MLP models are consist of two inputs layers and one layer output. The inputs of SVM network are solar radiation and time, while the output is the PV current. The inputs of MLP network are solar radiation and ambient temperature, while the output is the PV current. The practical implementation of the proposed SVM model is achieved a final MSE of (0.026378744) in training phase and (0.035615759) in cross validation phase. Besides, MLP is achieved a final MSE of (0.005804253) in the training phase and it is achieved (0.010523501) in cross validation phase. The final MSE of cross validation with standard deviation is (0.000527668). The experiments achieved in the predicting model a value of determination factor (R² = 0.9844388787) for SVM and (R² = 0.9701310549) for MLP which indicates the predicting model is very close to the regression line and a well data fitting to the statistical model. Besides, the proposed model achieved less MSE in comparison with other related work.
Patient telemonitoring results in an aggregation of significant amounts of information about patient disease trajectory. However, the potential use of this information for early prediction of exacerbations in adult asthma patients has not been systematically evaluated. The aim of this study was to explore the utility of telemonitoring data for building machine learning algorithms that predict asthma exacerbations before they occur. The study dataset comprised daily self-monitoring reports consisting of 7001 records submitted by adult asthma patients during home telemonitoring. Predictive modeling included preparation of stratified training datasets, predictive feature selection, and evaluation of resulting classifiers. Using a 7-day window, a naive Bayesian classifier, adaptive Bayesian network, and support vector machines were able to predict asthma exacerbation occurring on day 8, with sensitivity of 0.80, 1.00, and 0.84; specificity of 0.77, 1.00, and 0.80; and accuracy of 0.77, 1.00, and 0.80, respectively. Our study demonstrated that machine learning techniques have significant potential in developing personalized decision support for chronic disease telemonitoring systems. Future studies may benefit from a comprehensive predictive framework that combines telemonitoring data with other factors affecting the likelihood of developing acute exacerbation. Approaches implemented for advanced asthma exacerbation prediction may be extended to prediction of exacerbations in patients with other chronic health conditions.
Abstract:
In recent years, the consumption of electricity has increased considerably in the industrial, commercial and residential sectors. This has prompted a branch of research which attempts to overcome this problem by applying different information and communication technologies, turning houses and buildings into smart environments. In this paper, we propose and design an energy saving technique based on the relationship between the user's activities and electrical appliances in smart home environments. The proposed method utilizes machine learning techniques to automatically recognize the user's activities, and then a ranking algorithm is applied to relate activities and existing home appliances. Finally, the system gives recommendations to the user whenever it detects a waste of energy. Tests on a real database show that the proposed method can to save up to 35% of electricity in a smart home.
This paper empirically shows that the combined effect of applying the selected feature subsets and optimized parameters on machine learning techniques significantly improves the accuracy for solar power prediction. To provide evidence, experiments are carried on in two phases. For all the experiments the machine learning techniques namely Least Median Square (LMS), Multilayer Perceptron (MLP) and Support Vector Machine (SVM) are used. In the first phase five well-known wrapper feature selection methods are used to obtain the prediction accuracy of machine learning techniques with selected feature subsets and default parameter settings. The experiments from the first phase demonstrate that holding the default parameters, LMS, MLP and SVM provides better prediction accuracy (i.e. reduced MAE and MASE) with selected feature subsets rather than without selected feature subsets. After getting improved prediction accuracy from the first phase, the second phase continues the experiments to optimize machine learning parameters and the prediction accuracy of those machine learning techniques are re-evaluated through adopting both the optimized parameter settings and selected feature subsets. The comparison between the results of two phases clearly shows that the later phase (i.e. machine learning techniques with selected feature subsets and optimized parameters) provides substantial improvement in the accuracy for solar power prediction than the earlier phase (i.e. machine learning techniques with selected feature subsets and default parameters). Experiments are carried out using reliable and real life historical meteorological data. The machine learning accuracy of solar radiation prediction is justified in terms of statistical error measurement and validation metrics. Experimental results of this paper facilitate to make a concrete verdict that providing more attention and effort towards the feature subset selection and machine learning parameter optimization (e.g. combined effect of selected feature subsets and optimized parameters on prediction accuracy which is investigated in this paper) can significantly contribute to improve the accuracy of solar power prediction.
Smart meters have been deployed in many countries across the world since early 2000s. The smart meter as a key element for the smart grid is expected to provide economic, social, and environmental benefits for multiple stakeholders. There has been much debate over the real values of smart meters. One of the key factors that will determine the success of smart meters is smart meter data analytics, which deals with data acquisition, transmission, processing, and interpretation that bring benefits to all stakeholders. This paper presents a comprehensive survey of smart electricity meters and their utilization focusing on key aspects of the metering process, different stakeholder interests, and the technologies used to satisfy stakeholder interests. Furthermore, the paper highlights challenges as well as opportunities arising due to the advent of big data and the increasing popularity of cloud environments.
Of the many kinds of renewable energy, wind power is low in cost and non-polluting, so it is especially well-suited to Taiwan. The Mai Liao Wind Farm is the most important wind farm in Taiwan, and forecasting the wind power output for national sustainable development continues to be a challenging research feature. In this study, we attempt to forecast the wind power data collected from the Mai Liao Wind Farm. Our forecast model is based on a Multi-Layer Perceptron Artificial Neural Network (MLP) model using the data collected at the Mai Liao Wind Farm over a period of five years from September 2002 to August 2007. We proposed a new algorithm, namely improved Simplified Swarm Optimization (iSSO), which improves Simplified Swarm Optimization (SSO) by justifying the weights and bias in training the MLP. The proposed iSSO combines Principal Component Analysis (PCA), Autocorrelation Function (AF) and Partial Autocorrelation Function (PAF) for the selection of features which increases the efficiency of the proposed model. The experimental results demonstrate that the performance of iSSO outperforms the other six most popular algorithms.
Smart grids are becoming popular with the advent of sophisticated smart meters. They allow utilities to optimize energy consumption during peak hours by applying various demand response techniques including voluntary curtailment, direct control and price incentives. To sustain the curtailment over long periods of time of up to several hours utilities need to make fast and accurate consumption predictions on a large set of customers based on a continuous flow of real time data and huge historical data sets. Given the numerous consumption patterns customers exhibit, different prediction methods need to be used to reduce the prediction error. The straightforward approach of testing each customer against every method is unfeasible in this large volume and high velocity environment. To this aim, we propose a neural network based approach for automatically selecting the best prediction method per customer by relying only on a small subset of customers. We also introduce two historical averaging methods for consumption prediction that take advantage of the variability of the data and continuously update the results based on a sliding window technique. We show that once trained, the proposed neural network does not require frequent retraining, ensuring its applicability in online scenarios such as the sustainable demand response.
The performance of prediction models is often based on "abstract metrics"
that estimate the model's ability to limit residual errors between the observed
and predicted values. However, meaningful evaluation and selection of
prediction models for end-user domains requires holistic and
application-sensitive performance measures. Inspired by energy consumption
prediction models used in the emerging "big data" domain of Smart Power Grids,
we propose a suite of performance measures to rationally compare models along
the dimensions of scale independence, reliability, volatility and cost. We
include both application independent and dependent measures, the latter
parameterized to allow customization by domain experts to fit their scenario.
While our measures are generalizable to other domains, we offer an empirical
analysis using real energy use data for three Smart Grid applications:
planning, customer education and demand response, which are relevant for energy
sustainability. Our results underscore the value of the proposed measures to
offer a deeper insight into models' behavior and their impact on real
applications, which benefit both data mining researchers and practitioners.
Solar irradiation prediction is an important problem in geosciences with direct applications in renewable energy. Recently, a high number of machine learning techniques have been introduced to tackle this problem, mostly based on neural networks and support vector machines. Gaussian process regression (GPR) is an alternative nonparametric method that provided excellent results in other biogeophysical parameter estimation. In this letter, we evaluate GPR for the estimation of solar irradiation. Noting the nonstationary temporal behavior of the signal, we develop a particular time-based composite covariance to account for the relevant seasonal signal variations. We use a unique meteorological data set acquired at a radiometric station that includes both measurements and radiosondes, as well as numerical weather prediction models. We show that the so-called temporal GPR outperforms ten state-of-the-art statistical regression algorithms (even when including time information) in terms of accuracy and bias, and it is more robust to the number of predictions used.
Wind farms are producing a considerable portion of the world renewable energy. Since the output power of any wind farm is highly dependent on the wind speed, the power extracted from a wind park is not always a constant value. In order to have a non-disruptive supply of electricity, it is important to have a good scheduling and forecasting system for the energy output of any wind park. In this paper, a new hybrid swarm technique (HAP) is used to forecast the energy output of a real wind farm located in Binaloud, Iran. The technique consists of the hybridization of the ant colony optimization (ACO) and particle swarm optimization (PSO) which are two meta-heuristic techniques under the category of swarm intelligence. The hybridization of the two algorithms to optimize the forecasting model leads to a higher quality result with a faster convergence profile. The empirical hourly wind power output of Binaloud Wind Farm for 364 days is collected and used to train and test the prepared model. The meteorological data consisting of wind speed and ambient temperature is used as the inputs to the mathematical model. The results indicate that the proposed technique can estimate the output wind power based on the wind speed and the ambient temperature with an MAPE of 3.513%
In 2006, the Public Interest Energy Research Program (PIER) Demand Response Research Center (DRRC) at Lawrence Berkeley National Laboratory initiated research into Automated Demand Response (OpenADR) applications in California industry. The goal is to improve electric grid reliability and lower electricity use during periods of peak demand. The purpose of this research is to begin to define the relationship among a portfolio of actions that industrial facilities can undertake relative to their electricity use. This ?electricity value chain? defines energy management and demand response (DR) at six levels of service, distinguished by the magnitude, type, and rapidity of response. One element in the electricity supply chain is OpenADR, an open-standards based communications system to send signals to customers to allow them to manage their electric demand in response to supply conditions, such as prices or reliability, through a set of standard, open communications. Initial DRRC research suggests that industrial facilities that have undertaken energy efficiency measures are probably more, not less, likely to initiate other actions within this value chain such as daily load management and demand response. Moreover, OpenADR appears to afford some facilities the opportunity to develop the supporting control structure and to"demo" potential reductions in energy use that can later be applied to either more effective load management or a permanent reduction in use via energy efficiency. Under the right conditions, some types of industrial facilities can shift or shed loads, without any, or minimal disruption to operations, to protect their energy supply reliability and to take advantage of financial
incentives.1 In 2007 and 2008, 35 industrial facilities agreed to implement OpenADR, representing a total capacity of nearly 40 MW. This paper describes how integrated or centralized demand management and system-level network controls are linked to OpenADR systems. Case studies of refrigerated warehouses and wastewater treatment facilities are used to illustrate OpenADR load reduction potential. Typical shed and shift strategies include: turning off or operating compressors, aerator blowers and pumps at reduced capacity, increasing temperature set-points or pre-cooling cold storage areas and over-oxygenating stored wastewater prior to a DR event. This study concludes that understanding industrial end-use processes and control capabilities is a key to support reduced service during DR events and these capabilities, if DR enabled, hold significant promise in reducing the electricity demand of the industrial sector during utility peak periods.
Artificial Intelligence (AI) is the intelligence that can be shown by machines, in contrast to the natural way of intelligence displayed by humans or the way in which we can create intelligent machines that work and react like humans. AI is basically a term used when a machine behaves like a human, in activities such as problem-solving or learning, which is also known as Machine Learning. There are so many applications of AI that we use in our day to day lives without even knowing it. Such as, Siri, Alexa, Self-driven cars, Robotics, Gaming etc.
Objectives:
This study sought to determine whether machine learning can be used to better identify patients at risk for death or congestive heart failure (CHF) rehospitalization after percutaneous coronary intervention (PCI).
Background:
Contemporary risk models for event prediction after PCI have limited predictive ability. Machine learning has the potential to identify complex nonlinear patterns within datasets, improving the predictive power of models.
Methods:
We evaluated 11,709 distinct patients who underwent 14,349 PCIs between January 2004 and December 2013 in the Mayo Clinic PCI registry. Fifty-two demographic and clinical parameters known at the time of admission were used to predict in-hospital mortality and 358 additional variables available at discharge were examined to identify patients at risk for CHF readmission. For each event, we trained a random forest regression model (i.e., machine learning) to estimate the time-to-event. Eight-fold cross-validation was used to estimate model performance. We used the predicted time-to-event as a score, generated a receiver-operating characteristic curve, and calculated the area under the curve (AUC). Model performance was then compared with a logistic regression model using pairwise comparisons of AUCs and calculation of net reclassification indices.
Results:
The predictive algorithm identified a high-risk cohort representing 2% of all patients who had an in-hospital mortality of 45.5% (95% confidence interval: 43.5% to 47.5%) compared with a risk of 2.1% for the general population (AUC: 0.925; 95% confidence interval: 0.92 to 0.93). Advancing age, CHF, and shock on presentation were the leading predictors for the outcome. A high-risk group representing 1% of all patients was identified with 30-day CHF rehospitalization of 8.1% (95% confidence interval: 6.3% to 10.2%). Random forest regression outperformed logistic regression for predicting 30-day CHF readmission (AUC: 0.90 vs. 0.85; p = 0.003; net reclassification improvement: 5.14%) and 180-day cardiovascular death (AUC: 0.88 vs. 0.81; p = 0.02; net reclassification improvement: 0.02%).
Conclusions:
Random forest regression models (machine learning) were more predictive and discriminative than standard regression methods at identifying patients at risk for 180-day cardiovascular mortality and 30-day CHF rehospitalization, but not in-hospital mortality. Machine learning was effective at identifying subgroups at high risk for post-procedure mortality and readmission.
Recent global smart city efforts resemble the establishment of electricity networks when electricity was first invented, which meant the start of a new era to sell electricity as a utility. A century later, in the smart era, the network to deliver services goes far beyond a single entity like electricity. Supplemented by a well-established Internet infrastructure that can run an endless number of applications, abundant processing and storage capabilities of clouds, resilient edge computing, and sophisticated data analysis like machine learning and deep learning, an already-booming Internet of Things movement makes this new era far more exciting.
In this article, we present a multi-faceted survey of machine intelligence in modern implementations. We partition smart city infrastructure into application, sensing, communication, security, and data planes and put an emphasis on the data plane as the mainstay of computing and data storage. We investigate (i) a centralized and distributed implementation of data plane’s physical infrastructure and (ii) a complementary application of data analytics, machine learning, deep learning, and data visualization to implement robust machine intelligence in a smart city software core. We finalize our article with pointers to open issues and challenges.
Making houses more inclusive, safer, resilient and sustainable is an important requirement that must be achieved in every society. Gas leakage and fires in smart houses are serious issues that are causing people’s death and properties losses. Currently, preventing and alerting systems are widely available. However, they are generally individual units having elementary functions without adequate capabilities of multi-sensing and interaction with the existing Machine-to-Machine (M2M) home network along with the outside networks such as Internet. Indeed, this communication paradigm will be clearly the most dominant in the near future for M2M home networks. In this paper, we are proposing an efficient system model to integrate the gas leakage and fire detection system into a centralized M2M home network using low cost devices. Then, through machine learning approach, we are involving a data mining method with the sensed information and detect the abnormal air state changes in hidden patterns for early prediction of the risk incidences. This work will help to enhance safety and protect property in smart houses.
Liver cancer is quite common type of cancer among individuals worldwide. Hepatocellular carcinoma (HCC) is the malignancy of liver cancer. It has high impact on individual's life and investigating it early can decline the number of annual deaths. This study proposes a new machine learning approach to detect HCC using 165 patients. Ten well-known machine learning algorithms are employed. In the preprocessing step, the normalization approach is used. The genetic algorithm coupled with stratified 5-fold cross-validation method is applied twice, first for parameter optimization and then for feature selection. In this work, support vector machine (SVM) (type C-SVC) with new 2level genetic optimizer (genetic training) and feature selection yielded the highest accuracy and F1-Score of 0.8849 and 0.8762 respectively. Our proposed model can be used to test the performance with huge database and aid the clinicians.
In this paper, a big data-based approach is proposed for the security improvement of an unplanned microgrid islanding (UMI). The proposed approach contains two major steps: the first step is big data analysis of wide-area monitoring to detect a UMI and locate it; the second step is particle swarm optimization (PSO)-based stability enhancement for the UMI. First, an optimal synchrophasor measurement device selection (OSMDS) and matching pursuit decomposition (MPD)-based spatial-temporal analysis approach is proposed to significantly reduce the volume of data while keeping appropriate information from the synchrophasor measurements. Second, a random forest-based ensemble learning approach is trained to detect the UMI. When combined with grid topology, the UMI can be located. Then the stability problem of the UMI is formulated as an optimization problem and the PSO is used to find the optimal operational parameters of the UMI. An eigenvalue-based multiobjective function is proposed, which aims to improve the damping and dynamic characteristics of the UMI. Finally, the simulation results demonstrate the effectiveness and robustness of the proposed approach.
The limited available fossil fuels and the call for sustainable environment have brought about new technologies for the high efficiency in the use of fossil fuels and introduction of renewable energy. Smart grid is an emerging technology that can fulfill such demands by incorporating advanced information and communications technology (ICT). The pervasive deployment of the advanced ICT, especially the smart metering, will generate big energy data in terms of volume, velocity, and variety. The generated big data can bring huge benefits to the better energy planning, efficient energy generation, and distribution. As such data involve end users' privacy and secure operation of the critical infrastructure, there will be new security issues. This paper is to survey and discuss new findings and developments in the existing big energy data analytics and security. Several taxonomies have been proposed to express the intriguing relationships of various variables in the field.
In this paper, an online power system transient stability assessment (TSA) problem is mapped as a two-class classification problem and a novel data mining algorithm the core vector machine (CVM) is proposed to solve the problem based on phasor measurement units (PMUs) big data. First of all, an offline training, online application framework is proposed, which contained four sub-steps, namely features selection, offline training, online application, and assessment evaluation. First, 24 features are selected to present the system status. Then in the offline training procedure, the PMU big data is generated by time domain simulation, and a CVM model is trained and tested. In the online application procedure, an interface between PMU data center and feature calculation program is set up to collect real time specific PMU big data and the CVM trained is applied to the TSA problem. Last but not least, the evaluation indices are calculated. Compared with other support vector machines, the proposed CVM based assessment algorithm has the higher precision, meanwhile, it has the least time consumption and space complexity. As long as online PMU big data are received, TSA can be done simultaneously. Case studies on the IEEE New England 39-bus system, and real systems in China and the U.S., exhibit the speed and effectiveness of the proposed algorithm.
Due to the increasingly significant energy crisis nowadays, the exploitation and utilization of new clean energy gains more and more attention. As an important category of renewable energy, wind power generation has become the most rapidly growing renewable energy in China. However, the intermittency and volatility of wind power has restricted the large-scale integration of wind turbines into power systems. High-precision wind power forecasting is an effective measure to alleviate the negative influence of wind power generation on the power systems. In this paper, a novel combined model is proposed to improve the prediction performance for the short-term wind power forecasting. Variational mode decomposition is firstly adopted to handle the instability of the raw wind power series, and the subseries can be reconstructed by measuring sample entropy of the decomposed modes. Then the base models can be established for each subseries respectively. On this basis, the combined model is developed based on the optimal virtual prediction scheme, the weight matrix of which is dynamically adjusted by a self-adaptive multi-strategy differential evolution algorithm. Besides, a probabilistic interval prediction model based on quantile regression averaging and variational mode decomposition-based hybrid models is presented to quantify the potential risks of the wind power series. The simulation results indicate that: (1) the normalized mean absolute errors of the proposed combined model from one-step to three-step forecasting are 4.34%, 6.49% and 7.76%, respectively, which are much lower than those of the base models and the hybrid models based on the signal decomposition techniques; (2) the interval forecasting model proposed can provide reliable and excellent prediction results for a certain expectation probability, which is an effective and reliable tool for the short-term wind power probabilistic interval prediction.
Photovoltaic generation of electricity is an important renewable energy source, and large numbers of relatively small photovoltaic systems are proliferating around the world. Today it is widely acknowledged by power producers, utility companies and independent system operators that it is only through advanced forecasting, communications and control that these distributed resources can collectively provide a firm, dispatchable generation capacity to the electricity market. One of the challenges of realizing such a goal is the precise forecasting of the output of individual photovoltaic systems, which is affected by a lot of factors. This paper introduces our short-term solar irradiance forecasting algorithms based on machine learning methodologies, Hidden Markov Model and SVM regression. A series of experimental evaluations are presented to analyze the relative performance of the techniques in order to show the importance of these methodologies. The Matlab interface, the Weather Forecasting Platform, has been used for these evaluations. The experiments are performed using the dataset generated by Australian Bureau of Meteorology. The experimental results show that our machine learning based forecasting algorithms can precisely predict future 5–30 min solar irradiance under different weather conditions.
This paper presents the use of a Support Vector Machine load predictive energy management system to control the energy flow between a solar energy source, a supercapacitor-battery hybrid energy storage combination and the load. The supercapacitor-battery hybrid energy storage system is deployed in a solar energy system to improve the reliability of delivered power. The combination of batteries and supercapacitors makes use of complementary characteristic that allow the overlapping of a battery’s high energy density with a supercapacitors’ high power density. This hybrid system produces a straightforward benefit over either individual system, by taking advantage of each characteristic. When the supercapacitor caters for the instantaneous peak power which prolongs the battery lifespan, it also minimizes the system cost and ensures a greener system by reducing the number of batteries. The resulting performance is highly dependent on the energy controls implemented in the system to exploit the strengths of the energy storage devices and minimize its weaknesses. It is crucial to use energy from the supercapacitor and therefore minimize jeopardizing the power system reliability especially when there is a sudden peak power demand. This study has been divided into two stages. The first stage is to obtain the optimum SVM load prediction model, and the second stage carries out the performance comparison of the proposed SVM-load predictive energy management system with conventional sequential programming control (if-else condition). An optimized load prediction classification model is investigated and implemented. This C-Support Vector Classification yields classification accuracy of 100% using 17 support vectors in 0.004866 s of training time. The Polynomial kernel is the optimum kernel in our experiments where the C and g values are 2 and 0.25 respectively. However, for the load profile regression model which was implemented in the K-step ahead of load prediction, the radial basis function (RBF) kernel was chosen due to the highest squared correlation coefficient and the lowest mean squared error. Results obtained shows that the proposed SVM load predictive energy management system accurately identifies and predicts the load demand. This has been justified by the supercapacitor charging and leading the peak current demand by 200 ms for different load profiles with different optimized regression models. This methodology optimizes the cost of the system by reducing the amount of power electronics within the hybrid energy storage system, and also prolongs the batteries’ lifespan as previously mentioned.
It is well known that many real-world systems can be described by complex networks with the nodes and the edges representing the individuals and their communications, respectively. Based on recent advances in complex networks, this paper aims to provide some new methodologies to study some fundamental problems in smart grids. In particular, it summarises some results for networks properties, distributed control and optimisation, and pinning control in complex networks and tries to reveal how these new technologies can be applied in smart grids.
To improve real-time control performance and reduce possible negative impacts of photovoltaic (PV) systems, an accurate forecasting of PV output is required, which is an important function in the operation of an energy management system (EMS) for distributed energy resources. In this paper, a weather-based hybrid method for 1-day ahead hourly forecasting of PV power output is presented. The proposed approach comprises classification, training, and forecasting stages. In the classification stage, the self-organizing map (SOM) and learning vector quantization (LVQ) networks are used to classify the collected historical data of PV power output. The training stage employs the support vector regression (SVR) to train the input/output data sets for temperature, probability of precipitation, and solar irradiance of defined similar hours. In the forecasting stage, the fuzzy inference method is used to select an adequate trained model for accurate forecast, according to the weather information collected from Taiwan Central Weather Bureau (TCWB). The proposed approach is applied to a practical PV power generation system. Numerical results show that the proposed approach achieves better prediction accuracy than the simple SVR and traditional ANN methods.
This paper presents a new approach based on wavelet design and machine learning applied to passive islanding detection of distributed generation. Procrustes analysis is used to determine the filter coefficients of a newly designed wavelet. To automate the classification process, machine learning algorithms are used to develop appropriate models. The IEEE 13-bus standard test distribution system simulated in PSCAD/EMTDC is used as a test bed to assess the performance of the proposed approach. The numerical results demonstrating the effectiveness of the proposed approach are discussed and conclusions are drawn.
This paper discusses the performance of a novel Coral Reefs Optimization – Extreme Learning Machine (CRO–ELM) algorithm in a real problem of global solar radiation prediction. The work considers different meteorological data from the radiometric station at Murcia (southern Spain), both from measurements, radiosondes and meteorological models, and fully describes the hybrid CRO–ELM to solve the prediction of the daily global solar radiation from these data. The algorithm is designed in such a way that the ELM solves the prediction problem, whereas the CRO evolves the weights of the neural network, in order to improve the solutions obtained. The experiments carried out have shown that the CRO–ELM approach is able to obtain an accurate prediction of the daily global radiation, better than the classical ELM, and the Support Vector Regression algorithm.
We propose an ensemble short-term wind power forecasting model that is based on our novel approaches and advanced forecasting techniques in the modern literature. The performance of our model has been verified by forecasting wind power up to 48 hours ahead at seven wind farms for one and a half years. Our model ranked fourth in the Power and Energy Society (PES) wind power forecasting competition. The forecasting model uses 52 Neural Network (NN) sub-models and five Gaussian Process (GP) sub-models in parallel. For 48 hours, the NN sub-models forecast the future wind power based on historical wind power data and forecasted wind information. In parallel, for the first five hours, five GP sub-models are used to forecast wind power using only historical wind power in order to provide accurate wind power forecasts to NN sub-models. These models provide various forecasts for the same hour, so the optimal forecast should be decided from overlapped forecasts by the decision process.
