-
Reinforcement Learning for Efficient Design and Control Co-optimisation of Energy Systems
Authors:
Marine Cauz,
Adrien Bolland,
Nicolas Wyrsch,
Christophe Ballif
Abstract:
The ongoing energy transition drives the development of decentralised renewable energy sources, which are heterogeneous and weather-dependent, complicating their integration into energy systems. This study tackles this issue by introducing a novel reinforcement learning (RL) framework tailored for the co-optimisation of design and control in energy systems. Traditionally, the integration of renewa…
▽ More
The ongoing energy transition drives the development of decentralised renewable energy sources, which are heterogeneous and weather-dependent, complicating their integration into energy systems. This study tackles this issue by introducing a novel reinforcement learning (RL) framework tailored for the co-optimisation of design and control in energy systems. Traditionally, the integration of renewable sources in the energy sector has relied on complex mathematical modelling and sequential processes. By leveraging RL's model-free capabilities, the framework eliminates the need for explicit system modelling. By optimising both control and design policies jointly, the framework enhances the integration of renewable sources and improves system efficiency. This contribution paves the way for advanced RL applications in energy management, leading to more efficient and effective use of renewable energy sources.
△ Less
Submitted 28 June, 2024;
originally announced June 2024.
-
Privacy-preserving methods for smart-meter-based network simulations
Authors:
Jordan Holweger,
Lionel Bloch,
Christophe Ballif,
Nicolas Wyrsch
Abstract:
Smart-meters are a key component of energy transition. The large amount of data collected in near real-time allows grid operators to observe and simulate network states. However, privacy-preserving rules forbid the use of such data for any applications other than network operation and billing. Smart-meter measurements must be anonymised to transmit these sensitive data to a third party to perform…
▽ More
Smart-meters are a key component of energy transition. The large amount of data collected in near real-time allows grid operators to observe and simulate network states. However, privacy-preserving rules forbid the use of such data for any applications other than network operation and billing. Smart-meter measurements must be anonymised to transmit these sensitive data to a third party to perform network simulation and analysis. This work proposes two methods for data anonymisation that enable the use of raw active power measurements for network simulation and analysis. The first is based on an allocation of an externally sourced load database. The second consists of grou** smart-meter data with similar electric characteristics, then performing a random permutation of the network load-bus assignment. A benchmark of these two methods highlights that both provide similar results in bus-voltage magnitude estimation concerning ground-truth voltage.
△ Less
Submitted 7 October, 2021;
originally announced October 2021.
-
Unsupervised algorithm for disaggregating low-sampling-rate electricity consumption of households
Authors:
Jordan Holweger,
Marina Dorokhova,
Lionel Bloch,
Christophe Ballif,
Nicolas Wyrsch
Abstract:
Non-intrusive load monitoring (NILM) has been extensively researched over the last decade. The objective of NILM is to identify the power consumption of individual appliances and to detect when particular devices are on or off from measuring the power consumption of an entire house. This information allows households to receive customized advice on how to better manage their electrical consumption…
▽ More
Non-intrusive load monitoring (NILM) has been extensively researched over the last decade. The objective of NILM is to identify the power consumption of individual appliances and to detect when particular devices are on or off from measuring the power consumption of an entire house. This information allows households to receive customized advice on how to better manage their electrical consumption. In this paper, we present an alternative NILM method that breaks down the aggregated power signal into categories of appliances. The ultimate goal is to use this approach for demand-side management to estimate potential flexibility within the electricity consumption of households. Our method is implemented as an algorithm combining NILM and load profile simulation. This algorithm, based on a Markov model, allocates an activity chain to each inhabitant of the household, deduces from the whole-house power measurement and statistical data the appliance usage, generate the power profile accordingly and finally returns the share of energy consumed by each appliance category over time. To analyze its performance, the algorithm was benchmarked against several state-of-the-art NILM algorithms and tested on three public datasets. The proposed algorithm is unsupervised; hence it does not require any labeled data, which are expensive to acquire. Although better performance is shown for the supervised algorithms, our proposed unsupervised algorithm achieves a similar range of uncertainty while saving on the cost of acquiring labeled data. Additionally, our method requires lower computational power compared to most of the tested NILM algorithms. It was designed for low-sampling-rate power measurement (every 15 min), which corresponds to the frequency range of most common smart meters.
△ Less
Submitted 20 August, 2019;
originally announced August 2019.
