Christoph Doblander

Bio: Christoph Doblander is an academic researcher from Technische Universität München. The author has contributed to research in topics: Complex event processing & Bandwidth (computing). The author has an hindex of 7, co-authored 22 publications receiving 256 citations.

Household Electricity Demand Forecasting -- Benchmarking State-of-the-Art Methods

Abstract: The increasing use of renewable energy sources with variable output, such as solar photovoltaic and wind power generation, calls for Smart Grids that effectively manage flexible loads and energy storage. The ability to forecast consumption at different locations in distribution systems will be a key capability of Smart Grids. The goal of this paper is to benchmark state-of-the-art methods for forecasting electricity demand on the household level across different granularities and time scales in an explorative way, thereby revealing potential shortcomings and find promising directions for future research in this area. We apply a number of forecasting methods including ARIMA, neural networks, and exponential smoothening using several strategies for training data selection, in particular day type and sliding window based strategies. We consider forecasting horizons ranging between 15 minutes and 24 hours. Our evaluation is based on two data sets containing the power usage of individual appliances at second time granularity collected over the course of several months. The results indicate that forecasting accuracy varies significantly depending on the choice of forecasting methods/strategy and the parameter configuration. Measured by the Mean Absolute Percentage Error (MAPE), the considered state-of-the-art forecasting methods rarely beat corresponding persistence forecasts. Overall, we observed MAPEs in the range between 5 and >100%. The average MAPE for the first data set was ~30%, while it was ~85% for the other data set. These results show big room for improvement. Based on the identified trends and experiences from our experiments, we contribute a detailed discussion of promising future research.

Energy informatics: Current and future research directions

Abstract: Due to the increasing importance of producing and consuming energy more sustainably, Energy Informatics (EI) has evolved into a thriving research area within the CS/IS community. The article attempts to characterize this young and dynamic field of research by describing current EI research topics and methods and provides an outlook of how the field might evolve in the future. It is shown that two general research questions have received the most attention so far and are likely to dominate the EI research agenda in the coming years: How to leverage information and communication technology (ICT) to (1) improve energy efficiency, and (2) to integrate decentralized renewable energy sources into the power grid. Selected EI streams are reviewed, highlighting how the respective research questions are broken down into specific research projects and how EI researchers have made contributions based on their individual academic background.

Household electricity demand forecasting: benchmarking state-of-the-art methods

Abstract: We benchmark state-of-the-art methods for forecasting electricity demand on the household level. Our evaluation is based on two data sets containing the power usage on the individual appliance level. Our results indicate that without further refinement the considered advanced state-of-the-art forecasting methods rarely beat corresponding persistence forecasts. Therefore, we also provide an exploration of promising directions for future research.

Forecasting household electricity demand with complex event processing: insights from a prototypical solution

Abstract: The increasing use of renewable energy is leading to a paradigm shift in operating electrical grids. Production is moving away from centralized power plants to decentralized sources like solar panels and windmills. One consequence of this development is the need for managing supply and demand in local distribution grids in a "smart" way, which also implies the capability to forecast the demand for electric power closer to the end consumer and on shorter time scales than today. In this paper, we describe a system prototype for electricity demand forecasting based on highly disaggregated data from sensors deployed in homes and evaluate its performance both with respect to forecasting accuracy and ICT resource requirements. The data we use for our evaluation was collected in a pilot trial. Our system prototype combines complex event processing with state-of-the-art forecasting capabilities. For short-term forecasts, we observed average error reductions of up to 98 percentage points compared to average demand profiles. Our experiments also show the applicability of our approach at large scale. We were able to run the forecasting service for 1,000 households in parallel on one off-the-shelf server.

Adaptive middleware for real-time prescriptive analytics in large scale power systems

Abstract: The increased digitalization of power systems poses both opportunities and challenges for system operators. GPS time-synchronized high-resolution data streams emanating from measurement devices distributed over a wide area enable the detection of disturbances and the real-time monitoring of consequences as they are evolving, such as undamped oscillations. Processing these data streams is not possible with state-of-the-art SCADA systems that poll data asynchronously at much lower time intervals. Moreover, real-time analysis on fresh streaming data at the enterprise level is an unresolved challenge. In this paper we propose an adaptive middleware concept that can make better use of available data processing resources by enabling distributed computation both on the enterprise and on the field level. We apply the concept of linked data to provide a map for moving the computation to the data it requires for analysis. If based on the IEC 61850 standard semantic data model, the linked data concept additionally yields location and domain awareness that can be leveraged for real-time prescriptive analytics in the field. Another advantage of the proposed adaptive middleware is the abstraction of computational resources: Analytical programs can be written once and then be used to process historical data residing on servers on the enterprise level as well on the distributed devices that originated the data to enable fast analysis of events as they are unfolding.

Deep Learning for Household Load Forecasting—A Novel Pooling Deep RNN

Abstract: The key challenge for household load forecasting lies in the high volatility and uncertainty of load profiles. Traditional methods tend to avoid such uncertainty by load aggregation (to offset uncertainties), customer classification (to cluster uncertainties) and spectral analysis (to filter out uncertainties). This paper, for the first time, aims to directly learn the uncertainty by applying a new breed of machine learning algorithms—deep learning. However, simply adding layers in neural networks will cap the forecasting performance due to the occurrence of over-fitting. A novel pooling-based deep recurrent neural network is proposed in this paper which batches a group of customers’ load profiles into a pool of inputs. Essentially the model could address the over-fitting issue by increasing data diversity and volume. This paper reports the first attempts to develop a bespoke deep learning application for household load forecasting and achieved preliminary success. The developed method is implemented on Tensorflow deep learning platform and tested on 920 smart metered customers from Ireland. Compared with the state-of-the-art techniques in household load forecasting, the proposed method outperforms ARIMA by 19.5%, SVR by 13.1% and classical deep RNN by 6.5% in terms of RMSE.

A blockchain-based smart grid: towards sustainable local energy markets

Abstract: The increasing amount of renewable energy sources in the energy system calls for new market approaches to price and distribute the volatile and decentralized generation. Local energy markets, on which consumers and prosumers can trade locally produced renewable generation directly within their community, balance generation and consumption locally in a decentralized approach. We present a comprehensive concept, market design and simulation of a local energy market between 100 residential households. Our approach is based on a distributed information and communication technology, i.e. a private blockchain, which underlines the decentralized nature of local energy markets. Thus, we provide energy prosumers and consumers with a decentralized market platform for trading local energy generation without the need of a central intermediary. Furthermore, we present a preliminary economic evaluation of the market mechanism and a research agenda for the technological evaluation of blockchain technology as the local energy market’s main information and communication technology.