Semantische Suche

In dieser Bachelor-Arbeit wurden Werkzeuge und wissenschaftliche Arbeiten zur fortlaufenden Integration untersucht und eine Werkzeugkette aufgebaut, die dieses Prinzip umsetzt. Sie automatisiert neben der Softwareerstellung auch die Testausführung, die Bereitstellung und die Performanzmessung auf verschiedenen Zielsystemen und führt alle Zwischenergebnisse an einer Stelle zusammen. In Zusammenarbeit mit Agilent Technologies wurde eine Benutzerstudie durchgeführt, die aufzeigt, dass die Werkzeugkette eine Funktionalitäts- bzw. Leistungsaussage zu Quellcodeänderungen innerhalb weniger Minuten nach dem Einchecken ermöglicht, was sonst typischerweise Tage bis Wochen benötigt.

Energy time series (ETS) generally feature many components and are gathered at a high temporal resolution. Hence, it is required to reduce the data in order to allow analysis or further processing of the time series. However, existing data reduction methods do not account for energy-related characteristics of ETS and thus may lead to unsatisfying results.

In this work, we present a range of state-of-the art approaches for time series reduction (TSR) in the context of energy time series. The aim is to identify representative time slices from the multivariate energy time series without any prior knowledge about the inherent structure of the data. We rely on unsupervised approaches, i.e., clustering algorithms, to derive these representatives. For validation purpose, we apply the proposed reduction methods in two distinct approaches:

First, we use the TSR method to reduce the run time of energy system optimization models (ESM). ESM produce predictions and recommendations for the future energy system on the basis of historical data. As the model complexity and execution time of the ESM increases dramatically with the temporal resolution of the input data, reducing the input data without impacting the quality of predictions allows analysis at scales that are out of reach otherwise. In particular, we will study the Perseus-EU model. Our analysis show the extent to which each TSR method can reduce run times without degrading the quality of the prediction significantly.

The second application relates to the compression of ETS emerging from grid measurement data. Measurements from sensors installed in the energy grid collect observations in a high temporal resolution but are often highly redundant. Hence, while the storage requirements are high, the collected time series only contain few interesting and representative observations. Here, we use TSR methods to reduce the multivariate time series to a set of representative time slices. We show that amount of redundant observations can be greatly reduced in that way while preserving rare and interesting observations.

Recent research has shown that various hardware components such as wireless network interface cards (WNIC), cellular network interface cards or GPS modules have power states, that is, the power consumption behavior of a hardware component depends on the current state. Power models that consider power states (stateful power models) can be modeled as Power State Machines (PSM). For systems with multiple power states, stateful models proved to be more accurate than models that do not consider power states (stateless models).

Manually generating PSMs is time-consuming and limits the practicability of PSMs. Therefore, in this thesis, we explore the possibility of automatically generating PSMs. The contribution of this thesis is twofold: (1) We introduce an automated measurementbased profiling approach (2) and we introduce a step-based approach, which, provided with profiling data, automatically extracts PSMs along with tail states and state transitions.

We evaluate the automated PSM extraction in a case study on an offloading speech recognition system. We compare the power consumption prediction accuracy of the generated PSM with the prediction accuracy of a stateless regression based model. Because we measure the power consumption of the whole system, we use along with all WiFi power models the same CPU power model in order to predict the power consumption of the whole system. We find that a slightly adapted version of the generated PSM predicts the power consumption with a mean error of approx. 3% and an error of approx. 2% in the best case. In contrast, the regression model produces a mean error of approx. 19% and an error of approx. 18% in the best case.