It has been proposed that Semantic Web technologies would be key enablers in achieving context-aware computing in our everyday environments. In our vision of semantic technology empowered smart spaces, the whole interaction model is based on the sharing of semantic data via common blackboards. This approach allows smart space applications to take full advantage of semantic technologies. Because of its novelty, there is, however, a lack of solutions and methods for developing semantic smart space applications according to this vision. In this paper, we present solutions to the most relevant challenges we have faced when developing context-aware computing in smart spaces. In particular the paper describes (1) methods for utilizing semantic technologies with resource restricted-devices, (2) a solution for identifying real world objects in semantic technology empowered smart spaces, (3) a method for users to modify the behavior of context-aware smart space applications, and (4) an approach for content sharing between autonomous smart space agents. The proposed solutions include ontologies, system models, and guidelines for building smart spaces with the M3 semantic information sharing platform. To validate and demonstrate the approaches in practice, we have implemented various prototype smart space applications and tools.

Accurate short-term forecasts of building energy consumption are necessary for profitable demand response. Short-term forecasting methods can be roughly classified into physics-based modelling and data-based modelling. Both of these approaches have their advantages and disadvantages and it would be therefore ideal to combine them. This paper proposes a novel approach that allows us to combine the best parts of physics-based modelling and machine learning while avoiding many of their drawbacks. A key idea in the approach is to provide a variety of building parameters as input for an Artificial Neural Network (ANN) and train the model with data from a large group of simulated buildings. The hypothesis is that this forces the ANN model to learn the underlying simulation model-based physics, and thus enables the ANN model to be used in place of the simulator. The advantages of this type of model is the combination of robustness and accuracy from a high-detail physics-based model with the inference speed, ease of deployment, and support for gradient based optimization provided by the ANN model. To evaluate the approach, an ANN model was developed and trained with simulated data from 900–11,700 buildings, including equal distribution of office buildings, apartment buildings, and detached houses. The performance of the ANN model was evaluated with a test set consisting of 60 buildings (20 buildings for each category). The normalized root mean square errors (NRMSE) were on average 0.050, 0.026, 0.052 for apartment buildings, office buildings, and detached houses, respectively. The results show that the model was able to approximate the simulator with good accuracy also outside of the training data distribution and generalize to new buildings in new geographical locations without any building specific heat demand data.

Demand-side flexibility management is a key enabler of the transformation towards the high penetration of renewable energy resources. We present a flexibility-management system called Flex4Grid, which is designed to provide a low-cost solution for residential consumers wishing to participate in power-grid balancing. The Flex4Grid system continuously forecasts the need for flexibility in a power grid and informs consumers about the flexibility-management periods. Consumers can provide their flexibility to an aggregator in exchange for a reward, which depends on the selected incentive scheme. The automation of the flexibility-management events is provided by interfacing with devices and the system via the Z-Wave and Open Platform Communication Unified Architecture (OPC UA) technologies. The Flex4Grid system has been deployed in three pilots in Slovenia and Germany. A large-scale pilot in Celje, Slovenia, with 1047 participants, was used to collect statistical data regarding how consumers participate in the flexibility-management events. A critical peak-pricing incentive scheme was used in the Celje pilot. The smaller German pilots with a total of 185 participants were used for testing the technical capabilities of the system. User-satisfaction surveys were performed in all three pilots. The results indicate that the proposed approach is appropriate for engaging consumers in flexibility-management events. On average, the pilots’ participants reduced their load by 10% during a peak event. The overall scores of the user-satisfaction survey were 3.4 and 3.9 on a 5-point Likert scale for the German and Slovenian pilots, respectively. These are good results for a prototype system; however, improvements to the stability and usability of the system are required.

The current approaches for energy consumption optimisation in buildings are mainly reactive or focus on scheduling of daily/weekly operation modes in heating. Machine Learning (ML)-based advanced control methods have been demonstrated to improve energy efficiency when compared to these traditional methods. However, placing of ML-based models close to the buildings is not straightforward. Firstly, edge-devices typically have lower capabilities in terms of processing power, memory, and storage, which may limit execution of ML-based inference at the edge. Secondly, associated building information should be kept private. Thirdly, network access may be limited for serving a large number of edge devices. The contribution of this paper is an architecture, which enables training of ML-based models for energy consumption prediction in private cloud domain, and transfer of the models to edge nodes for prediction in Kubernetes environment. Additionally, predictors at the edge nodes can be automatically updated without interrupting operation. Performance results with sensor-based devices (Raspberry Pi 4 and Jetson Nano) indicated that a satisfactory prediction latency (~7–9 s) can be achieved within the research context. However, model switching led to an increase in prediction latency (~9–13 s). Partial evaluation of a Reference Architecture for edge computing systems, which was used as a starting point for architecture design, may be considered as an additional contribution of the paper.

Optimal management of demand-side flexibility in buildings is important for properly integrating large amounts of intermittent generation from windmills and photovoltaics. This paper proposes a novel Energy Management Agent (EMA) concept that can optimize building’s energy costs with respect to external prices while at the same time allow building’s flexibility to be used via explicit demand response. The EMA combines Artificial Neural Networks (ANN) and model predictive control for modelling and optimization of building’s flexibility. It continuously manages building’s flexibility with respect to external prices and provides forecasts of the load and available flexibility for a defined time window. A proof-of-concept (PoC) of the EMA is implemented for controlling a heat pump in an apartment, located in Oulu, Finland. Two ANN-based models were implemented for modelling the energy consumption of the heat pump and the indoor temperature of the apartment. Monte Carlo Tree Search based planning and control was implemented for finding optimal control policies with the ANNs. The EMA PoC was evaluated in 16-week period between 11 November 2019 - 1 March, 2020. When compared to a fixed setpoint control strategy, the EMA achieved 14.8 % lower costs under Nord Pool spot prices for Finland. At the same time, it was also able to accurately follow the 24h load plans (NRMSE was 0.050) and activate the offered flexibilities (NRMSE was 0.074).

This paper presents a novel deep learning architecture for short-term load forecasting of building energy loads. The architecture is based on a simple base learner and multiple boosting systems that are modelled as a single deep neural network. The architecture transforms the original multivariate time series into multiple cascading univariate time series. Together with sparse interactions, parameter sharing and equivariant representations, this approach makes it possible to combat against overfitting while still achieving good presentation power with a deep network architecture. The architecture is evaluated in several short-term load forecasting tasks with energy data from an office building in Finland. The proposed architecture outperforms state-of-the-art load forecasting model in all the tasks.