Skip to main content

Über dieses Buch

This book constitutes the revised selected papers from the 6th ECML PKDD Workshop on Data Analytics for Renewable Energy Integration, DARE 2018, held in Dublin, Ireland, in September 2018.

The 9 papers presented in this volume were carefully reviewed and selected for inclusion in this book and handle topics such as time series forecasting, the detection of faults, cyber security, smart grid and smart cities, technology integration, demand response, and many others.



Mathematical Optimization of Design Parameters of Photovoltaic Module

We propose a method for finding the most appropriate photovoltaic (hereafter PV) module and the return on investment for specific household needs, leveraging mathematical optimization. Based on electricity consumption and location of the household, the algorithm finds PV module design parameters using Covariance Matrix Adaptation Evolution Strategy (hereafter CMA-ES). According to these computed design parameters, the algorithm finds the most similar PV module from the dataset of PV modules using Euclidean distance. Subsequently, savings and costs are calculated for the recommended PV system. When calculating the return on investment, the algorithm takes into account the price of electricity, initial investment and battery replacements. The objective function is defined to maximize net savings during the lifetime of the PV system, which is considered to be 20 years. Heuristics of the battery takes into account the aging of batteries and the loss of energy due to storage. Battery life is set at 10 years, as we are considering using high quality batteries. In order to extend battery life, the algorithm counts with the charge level restrictions.
Dávid Kubík, Jaroslav Loebl

Fused Lasso Dimensionality Reduction of Highly Correlated NWP Features

Two problems when using Numerical Weather Prediction features in Machine Learning are the high dimensionality inherent to the current high-resolution models, and the high correlation of the features, which can affect the performance of learning machines as Multilayer Perceptron (MLP). In this work we propose to reduce the dimension of the problem using a supervised Fused Lasso model, which generates meta-features corresponding to the average of the groups with constant coefficients. The Fused Lasso problem is defined in terms of the feature correlation graph and tries to retain features with the stronger connections. As shown experimentally, training the models over the correlation graph-based reduced dataset allows to decrease the overall computational time while preserving almost the same error in the case of Support Vector Regressors and even improving the error of the MLPs, if the original dimension is high.
Alejandro Catalina, Carlos M. Alaíz, José R. Dorronsoro

Sampling Strategies for Representative Time Series in Load Flow Calculations

Power system analysis algorithms increasingly use time series with a high temporal resolution to assess operational and planning aspects of the power grid. By using time series with high temporal resolution, information is getting more detailed, but at the same time, the computational costs of the algorithms increase. With the help of our algorithm, we create representative time series that have similar characteristics to the original time series. With the help of these representative time series, it is possible to reduce the computational cost of power system analysis algorithms having nearly the same results as with the original time series. In this work, we improve our previous algorithm with the help of specialized sampling strategies. Furthermore, we provide a new method to compare power analysis results achieved with the representative time series to the original time series.
Janosch Henze, Stephan Kutzner, Bernhard Sick

Probabilistic Graphs for Sensor Data-Driven Modelling of Power Systems at Scale

The growing complexity of the power grid, driven by increasing share of distributed energy resources and by massive deployment of intelligent internet-connected devices, requires new modelling tools for planning and operation. Physics-based state estimation models currently used for data filtering, prediction and anomaly detection are hard to maintain and adapt to the ever-changing complex dynamics of the power system. A data-driven approach based on probabilistic graphs is proposed, where custom non-linear, localised models of the joint density of subset of system variables can be combined to model arbitrarily large and complex systems. The graphical model allows to naturally embed domain knowledge in the form of variables dependency structure or local quantitative relationships. A specific instance where neural-network models are used to represent the local joint densities is proposed, although the methodology generalises to other model classes. Accuracy and scalability are evaluated on a large-scale data set representative of the European transmission grid.
Francesco Fusco

Renewable Energy Integration: Bayesian Networks for Probabilistic State Estimation

Increased availability of renewable energy sources, along with novel techniques for power flow control, open up a broad range of interesting challenges and opportunities in power flow optimization. This promises reduced power generation costs through better integration of renewable energy generators into the Smart Grid. Unfortunately, renewable generators are fundamentally variable and uncertain. This uncertainty motivates our study of probabilistic state estimation techniques in this paper. Specifically, we use probabilistic graphical models in the form of Bayesian networks to compute probabilities of power system states, thus enabling improved power flow control. Key differences between our probabilistic state estimation results as reported in this paper and similar previous efforts include: our use of Bayesian probabilistic but exact (rather than Monte Carlo) state estimation techniques; auto-generation of Bayesian networks for probabilistic state estimation; integration with corrective Security-Constrained Optimal Power Flow; and application to Distributed Flexible AC Transmission Systems. We present novel models and algorithms for probabilistic state estimation using auto-generated Bayesian networks compiled to junction trees, and report on experimental results that illustrate the scalability of our methods.
Ole J. Mengshoel, Priya K. Sundararajan, Erik Reed, Dongzhen Piao, Briana Johnson

Deep Learning for Wave Height Classification in Satellite Images for Offshore Wind Access

Measuring wave heights has traditionally been associated with physical buoy tools that aim to measure and average multiple wave heights over a period of time. With our method, we demonstrate a process of utilizing large-scale satellite images to classify a wave height with a continuous regressive output using a corresponding input for close shore sea. We generated and trained a convolutional neural network model that achieved an average loss of 0.17 m (Fig. 8). Providing an inexpensive and scalable approach for uses in multiple sectors, with practical applications for offshore wind farms.
Ryan J. Spick, James A. Walker

Machine Learning as Surrogate to Building Performance Simulation: A Building Design Optimization Application

Increasing Heating, Ventilation, and Air conditioning (HVAC) efficiency is critically important as the building sector accounts for about 40% of the world’s primary energy consumption. Building Performance Simulation (BPS) can be used to model the relationship between building characteristics and energy consumption and to facilitate optimization efforts. However, BPS is computationally intensive and only a limited set of building configurations can be evaluated. Machine learning techniques provide an alternative method of modeling energy consumption. While not as accurate, they can be used to perform a “first pass” evaluation of large numbers of building configurations and hence to identify promising candidates for subsequent analysis. This paper presents an initial proof-of-concept implementation of this idea. A machine learning algorithm is trained on a dataset generated using BPS, and is combined with a Genetic Algorithm (GA) based optimization to evaluate tens of thousands of building configurations in terms of energy consumption, producing designs that are very close to the optimum.
Sokratis Papadopoulos, Wei Lee Woon, Elie Azar

Clustering River Basins Using Time-Series Data Mining on Hydroelectric Energy Generation

Hydropower is a significant renewable energy type with a considerable share in energy generation worldwide. As with the other common means of energy generation, hydropower is critical for the reliability and quality of electricity supply. Maintaining the reliability and quality of supply enables meeting the electricity demand of the loads adequately and efficient use of the energy resources, in addition to decreasing the related financial and environmental losses. In this paper, we target at the problem of basin clustering which is crucial for hydrological and electrical analyses regarding hydropower plants. We propose an approach based on time-series data mining on generation data of a large number of run-of-river type plants as well as of a number of representative storage type plants, in order to cluster the river basins in Turkey and present the clustering results with the related discussions. Based on these results, a new basin map is proposed which will be beneficial for enhanced hydrological and electrical analyses on hydropower and thereby for the maintenance of supply reliability and quality.
Yusuf Arslan, Dilek Küçük, Sinan Eren, Aysenur Birturk

Short-Term Electricity Consumption Forecast Using Datasets of Various Granularities

It is widely known that the generation and consumption of electricity should be balanced for secure operation and maintenance of the electricity grid. In order to help achieve this balance in the grid, the renewable energy resources such as wind and stream-flow should be forecast at high accuracies on the generation side, and similarly, electricity consumption should be forecast using a high-performance system. In this paper, we deal with short-term electricity consumption forecast in Turkey, and conduct various ANN-based experiments using real consumption data. The experiments are carried out on datasets of various scales in order to arrive at a learning system that uses, as the training dataset, a convenient subset of large quantities of field data. Thereby, the performance of system can be improved in addition to decreasing the time for the training stage, so that the resulting system can be efficiently used in operational settings. The performance evaluation results of these experiments to forecast electricity consumption in Nigde province of Turkey are presented together with the related discussions. This study provides an important baseline of findings, upon which other learning systems and training settings can be tested, improved, and compared with each other.
Yusuf Arslan, Aybike Şimşek Dilbaz, Seyda Ertekin, Pinar Karagoz, Aysenur Birturk, Sinan Eren, Dilek Küçük

Intelligent Monitoring of Transformer Insulation Using Convolutional Neural Networks

The ability to monitor and detect potential faults in smart grid system components is extremely valuable. In this paper, we demonstrate the use of machine learning techniques for condition monitoring in power transformers. Our objective is to classify the three different types of Partial Discharge (PD), the identify of which is highly correlated with insulation failure. Measurements from Acoustic Emission (AE) sensors are used as input data. Two broad machine learning based approaches are considered - the conventional method which uses a predefined feature set (Fourier based), and deep learning where features are learned automatically from the data. The performance of deep learning compares very favorably to the traditional approach, which includes ensemble learning and support vector machines, while eliminating the need for explicit feature extraction from the input AE signals. The results are particularly encouraging as manual feature extraction is a subjective process that may require significant redesign when confronted with new operating conditions and data types. In contrast, the ability to automatically learn feature sets from the raw input data (AE signals) promises better generalization with minimal human intervention.
Wei Lee Woon, Zeyar Aung, Ayman El-Hag

Nonintrusive Load Monitoring Based on Deep Learning

This paper presents a novel nonintrusive load monitoring method based on deep learning. Unlike the existing work based on convolutional neural network and recurrent neural network with fully connected layers, this paper develops a deep neural network based on sequence-to-sequence model and attention mechanism to perform nonintrusive load monitoring. The overall framework can be divided into three layers. In the first layer, the input active power time sequence is embedded into a group of high dimensional vectors. In the second layer, the vectors are encoded by a bi-directional LSTM layer, and the N encoded vectors are added up to form a dynamic context vector according to its weights calculated by the attention mechanism. In the third layer, an LSTM-based decoder utilizes the dynamic context vector to calculate the disaggregated power consumption at every time step. The proposed method is trained and tested on REFITPowerData dataset. The results show that compared to the state-of-the-art methods, the proposed method significantly increases the accuracy of the estimation for the disaggregated power value and decreases the misjudge rate by 10% to 20%.
Ke Wang, Haiwang Zhong, Nanpeng Yu, Qing Xia

Urban Climate Data Sensing, Warehousing, and Analysis: A Case Study in the City of Abu Dhabi, United Arab Emirates

With the ever increasing observations and measurements of geo-sensor networks, satellite imageries, geo-spatial data of location based services (LBS) and location-based social networks has become a serious challenge for data management and analysis systems. In urban micro-climate, we need to deal with various types of data such as: environmental data measurements, Wi-Fi data and so on. The format and the nature of data coming from different sensors such as temperature, humidity, thermal cameras, wind sensors, and others within an urban area varies. Therefore, there is a need for a unified platform to store these data efficiently using new technologies for which, we have come up with implementation of OLAP cubes. Furthermore, additional analytics for assessing urban thermal comfort can also be derived based on behavioural patterns of people. Therefore, outdoor Wi-Fi usage statistics is used as a proxy for the amount of time people spend outdoors, to correlate outdoor thermal conditions to perceived thermal comfort. Some interesting obervations are made in our study.
Prajowal Manandhar, Prashanth Reddy Marpu, Zeyar Aung


Weitere Informationen

Premium Partner

BranchenIndex Online

Die B2B-Firmensuche für Industrie und Wirtschaft: Kostenfrei in Firmenprofilen nach Lieferanten, Herstellern, Dienstleistern und Händlern recherchieren.



Best Practices für die Mitarbeiter-Partizipation in der Produktentwicklung

Unternehmen haben das Innovationspotenzial der eigenen Mitarbeiter auch außerhalb der F&E-Abteilung erkannt. Viele Initiativen zur Partizipation scheitern in der Praxis jedoch häufig. Lesen Sie hier  - basierend auf einer qualitativ-explorativen Expertenstudie - mehr über die wesentlichen Problemfelder der mitarbeiterzentrierten Produktentwicklung und profitieren Sie von konkreten Handlungsempfehlungen aus der Praxis.
Jetzt gratis downloaden!