There is, I think, something ethereal about i —the square root of minus one. I remember first hearing about it at school. It seemed an odd beast at that time—an intruder hovering on the edge of reality.

Usually familiarity dulls this sense of the bizarre, but in the case of i it was the reverse: over the years the sense of its surreal nature intensified. It seemed that it was impossible to write mathematics that described the real world in …

I and i

Data Mining Practical Machine Learning Tools and Techniques

As the power system is facing a transition toward a more intelligent, flexible, and interactive system with higher penetration of renewable energy generation, load forecasting, especially short-term load forecasting for individual electric customers plays an increasingly essential role in the future grid planning and operation. Other than aggregated residential load in a large scale, forecasting an electric load of a single energy user is fairly challenging due to the high volatility and uncertainty involved. In this paper, we propose a long short-term memory (LSTM) recurrent neural network-based framework, which is the latest and one of the most popular techniques of deep learning, to tackle this tricky issue. The proposed framework is tested on a publicly available set of real residential smart meter data, of which the performance is comprehensively compared to various benchmarks including the state-of-the-arts in the field of load forecasting. As a result, the proposed LSTM approach outperforms the other listed rival algorithms in the task of short-term load forecasting for individual residential households.

Short-Term Residential Load Forecasting Based on LSTM Recurrent Neural Network

https://hal.archives-ouvertes.fr/hal-01009386/document

Energy efficiency in wireless sensor networks: A top-down survey

The statistical analysis of compositional data

Power curves constructed from wind speed and active power output measurements provide an established method of analyzing wind turbine performance. In this paper, it is proposed that operational data from wind turbines are used to estimate bivariate probability distribution functions representing the power curve of existing turbines so that deviations from expected behavior can be detected. Owing to the complex form of dependency between active power and wind speed, which no classical parameterized distribution can approximate, the application of empirical copulas is proposed; the statistical theory of copulas allows the distribution form of marginal distributions of wind speed and power to be expressed separately from information about the dependency between them. Copula analysis is discussed in terms of its likely usefulness in wind turbine condition monitoring, particularly in early recognition of incipient faults such as blade degradation, yaw, and pitch errors.

/pdf/wind-turbine-condition-assessment-through-power-curve-copula-35lqgt2x53.pdf

Wind Turbine Condition Assessment Through Power Curve Copula Modeling

Aspirations of grid independence could be achieved by residential power systems connected only to small highly variable loads if overall demand on the network can be accurately anticipated. Absence of the diversity found on networks with larger load cohorts or consistent industrial customers makes such overall load profiles difficult to anticipate on even a short term basis. Here, existing forecasting techniques are employed alongside enhanced classification/clustering models in proposed methods for forecasting demand in a bottom up manner. A Markov chain based sampling technique derived from practice theory of human behavior is proposed as a means of providing a forecast with low computational effort and reduced historical data requirements. The modeling approach proposed does not require seasonal adjustments or environmental data. Forecast and actual demand for a cohort of residential loads over a five-month period are used to evaluate a number of models, as well as demonstrate a significant performance improvement if utilized in an ensemble forecast.

/pdf/incorporating-practice-theory-in-sub-profile-models-for-4aub0y59o4.pdf

Incorporating Practice Theory in Sub-Profile Models for Short Term Aggregated Residential Load Forecasting

Anticipating load characteristics on low voltage circuits is an area of increased concern for Distribution Network Operators with uncertainty stemming primarily from the validity of domestic load profiles. Identifying customer behavior makeup on a LV feeder ascertains the thermal and voltage constraints imposed on the network infrastructure; modeling this highly dynamic behavior requires a means of accommodating noise incurred through variations in lifestyle and meteorological conditions. Increased penetration of distributed generation may further worsen this situation with the risk of reversed power flows on a network with no transformer automation. Smart Meter roll-out is opening up the previously obscured view of domestic electricity use by providing high resolution advance data; while in most cases this is provided historically, rather than real-time, it permits a level of detail that could not have previously been achieved. Generating a data driven profile of domestic energy use would add to the accuracy of the monitoring and configuration activities undertaken by DNOs at LV level and higher which would afford greater realism than static load profiles that are in existing use. In this paper, a linear Gaussian load profile is developed that allows stratification to a finer level of detail while preserving a deterministic representation.

/pdf/enhanced-load-profiling-for-residential-network-customers-3o7y4cxaru.pdf

Enhanced Load Profiling for Residential Network Customers

Future power systems could benefit considerably from having a continuous real-time estimate of system inertia. If realized, this could provide reference inputs to proactive control and protection systems which could enhance not only system stability but also operational economics through, for example, more informed ancillary reserve planning using knowledge of prevailing system conditions and stability margins. Performing these predictions in real time is a significant challenge owing to the complex stochastic and temporal relationships between available measurements. This paper proposes a statistical model capable of estimating system inertia in real time through observed steady-state and relatively small frequency variations; it is trained to learn the features that inter-relate steady-state averaged frequency variations and system inertia, using historical system data demonstrated over two consecutive years. The proposed algorithm is formulated as a Gaussian Mixture Model with temporal dependence encoded as Markov chains. Applied to the U.K. power system, it produces an optimized mean squared error within 0.1 s $^{2}$ for 95% of the daily estimation if being calibrated on a half-hourly basis and maintains robustness through measurement interruptions of up to a period of two hours.

/pdf/switching-markov-gaussian-models-for-dynamic-power-system-3ed1rbb4zv.pdf

Switching Markov Gaussian Models for Dynamic Power System Inertia Estimation

Increasing prevalence of dc sources and loads has resulted in dc distribution being reconsidered at a microgrid level. However, in comparison to ac systems, the lack of a natural zero crossing has traditionally meant that protecting dc systems is inherently more difficult—this protection issue is compounded when attempting to diagnose and isolate fault conditions. One such condition is the series arc fault, which poses significant protection issues as their presence negates the logic of overcurrent protection philosophies. This paper proposes the IntelArc system to accurately diagnose series arc faults in dc systems. IntelArc combines time–frequency and time-domain extracted features with hidden Markov models (HMMs) to discriminate between nominal transient behavior and arc fault behavior across a variety of operating conditions. Preliminary testing of the system is outlined with results showing that the system has the potential for accurate, generalized diagnosis of series arc faults in dc systems.

/pdf/diagnosis-of-series-dc-arc-faults-a-machine-learning-2bdlxyz9qi.pdf

Bruce Stephen

Papers

Wind Turbine Condition Assessment Through Power Curve Copula Modeling

Incorporating Practice Theory in Sub-Profile Models for Short Term Aggregated Residential Load Forecasting

Enhanced Load Profiling for Residential Network Customers

Switching Markov Gaussian Models for Dynamic Power System Inertia Estimation

Diagnosis of Series DC Arc Faults—A Machine Learning Approach