Kepplinger, Peter
Refine
Year of publication
Document Type
- Article (20)
- Conference Proceeding (18)
- Doctoral Thesis (1)
Institute
Keywords
EU regulations get stricter from 2028 on by imposing net-zero energy building (NZEB) standards on new residential buildings including on-site renewable energy integration. Heat pumps (HP) using thermal building mass, and Model Predictive Control (MPC) provide a viable solution to this problem. However, the MPC potential in NZEBs considering the impact on indoor comfort have not yet been investigated comprehensively. Therefore, we present a co-simulative approach combining MPC optimization and IDA ICE building simulation. The demand response (DR) potential of a ground-source HP and the long-term indoor comfort in an NZEB located in Vorarlberg, Austria over a one year period are investigated. Optimization is performed using Mixed-Integer Linear Programming (MILP) based on a simplified RC model. The HP in the building simulation is controlled by power signals obtained from the optimization. The investigation shows reductions in electricity costs of up to 49% for the HP and up to 5% for the building, as well as increases in PV self-consumption and the self-sufficiency ratio by up to 4% pt., respectively, in two distinct optimization scenarios. Consequently, the grid consumption decreased by up to 5%. Moreover, compared to the reference PI controller, the MPC scenarios enhanced indoor comfort by reducing room temperature fluctuations and lowering the average percentage of people dissatisfied by 1% pt., resulting in more stable indoor conditions. Especially precooling strategies mitigated overheating risks in summer and ensured indoor comfort according to EN 16798-1 class II standards.
The food industry faces significant challenges in managing operational costs due to its high energy intensity and rising energy prices. Industrial food-processing facilities, with substantial thermal capacities and large demands for cooling and heating, offer promising opportunities for demand response (DR) strategies. This study explores the application of deep reinforcement learning (RL) as an innovative, data-driven approach for DR in the food industry. By leveraging the adaptive, self-learning capabilities of RL, energy costs in the investigated plant are effectively decreased. The RL algorithm was compared with the well-established optimization method Mixed Integer Linear Programming (MILP), and both were benchmarked against a reference scenario without DR. The two optimization strategies demonstrate cost savings of 17.57% and 18.65% for RL and MILP, respectively. Although RL is slightly less efficient in cost reduction, it significantly outperforms in computational speed, being approximately 20 times faster. During operation, RL only needs 2ms per optimization compared to 19s for MILP, making it a promising optimization tool for edge computing. Moreover, while MILP’s computation time increases considerably with the number of binary variables, RL efficiently learns dynamic system behavior and scales to more complex systems without significant performance degradation. These results highlight that deep RL, when applied to DR, offers substantial cost savings and computational efficiency, with broad applicability to energy management in various applications.
Optimal scheduling of electric vehicle charging is a non-trivial problem associated with multiple sources of uncertainties. These uncertainties are often neglected in demand-side management studies assuming perfect predictions, which in practice is unrealistic. In this paper, we propose a model predictive control framework for scheduling the charging of residential electric vehicles to account for the uncertainties associated. The evaluations are performed considering a decentralized demand-side management algorithm proposed in the literature, which aims to exploit the flexibility of electric vehicles to fill the valleys in the demand curve in order to flatten the aggregated load. The performance of the method is evaluated in response to the uncertainty in the non-elastic load, aggregated electric vehicle demand, and electric vehicle user behavior. The results show that the variance in the demand is reduced by a factor of 4.8 in the proposed model predictive-based method in the presence of all three uncertainties considered relative to the uncontrolled charging. Under perfect prediction, the reduction is a factor of 7.5, thereby indicating that the method is a viable solution against uncertainties. Moreover, the study provides an overview of the degree of overestimations in the desired outcomes realized under the assumptions of perfect predictions for the different uncertain parameters, demonstrating that the most significant impact arises from uncertainty in mobility usage.
Demand side management holds potential for improving energy efficiency and cutting energy consumption within the food industry. This research introduces a demand response approach tailored for an industrial food processing facility, utilizing a chilled water buffer as active thermal energy storage and the plant building as passive thermal energy storage. The plant building and production process are modeled using transient thermal energy balances and the demand side management problem is formulated as a linear program. Model predictive control is employed to manage uncertainties in the optimization process. A simulated case study of an Austrian food processing plant shows reductions in electrical power consumption by up to 18%, electricity costs by up to 24%, and peak load by up to 36% in three distinct optimization scenarios. Simple prediction approaches via averaging historical data already lead to nearly optimal results concerning energy consumption and cost reduction. Highly accurate predictions are necessary for peak load reduction, as considering the simple prediction method only roughly a third of the potential reductions are achieved.
Domestic hot water heaters are considered to be easily integrated as flexible loads for demand response. While literature grows on reproducible simulation and lab tests, real-world implementation in field tests considering state estimation and demand prediction-based model predictive control approaches is rare. This work reports the findings of a field test with 16 autonomous smart domestic hot water heaters. The heaters were equipped with a retrofittable sensor/actuator setup and a real-time price-driven model predictive control unit, which covers state estimation, demand prediction, and optimization of switching times. With the introduction of generic performance indicators (specific costs and thermal efficiency), the results achieved in the field are compared by simulations to standard control modes (instantaneous heating, hysteresis, night-only switching). To evaluate how model predictive control performance depends on the user demand prediction and state estimation accuracy, simulations assuming perfect predictions and state estimations are conducted based on the data measured in the field. Results prove the feasible benefit of RTP-based model predictive control in the field compared to a hysteresis-based standard control regarding cost reduction and efficiency increase but show a strong dependency on the degree of utilization.
Renewable energy communities (REC) are legal entities defined by the European Clean Energy Package to own and share energy resources. A common goal of RECs is to have a high self-consumption. This work focuses on how self-consumption can be increased by means of intelligent control of flexibilities based on a use case featuring a community battery energy storage (CBES). Many simulation studies for the control and optimization of distributed energy resources (DER) including community storage exist, focusing on various aspects, however, often neglecting important effects for a real implementation due to simplified assumptions: (A) The control or optimization routine is run open loop. Thereby, the influence of prediction and modeling errors is ignored, so that the results obtained represent only an upper limit for the effectiveness. (B) The availability of data is not considered appropriately which might render certain control approaches unpractical. In practice, this might be a limiting factor for forecasting methods and control methods in general. Especially, as RECs are of a distributed nature and the implementation of potentially costly data acquisition and information infrastructure could limit the economic rentability significantly. We investigate a realistic use case, constructed based on the current Austrian legal definition of RECs. Here, recent smart meter data of the REC participants, as it is made available by the distribution system operator (DSO) for settlement purposes. We utilize this data for control. Our realistic simulation model consists of several residential loads, a community PV system, two private PV systems and a CBES which is controlled to boost the RECs self-consumption. The model was implemented within the MOSAIK co-simulation framework. Due to the enforced modularity of this framework, the controller is integrated into the simulation model in a closed loop manner, addressing limitation (A) of similar studies. In addition, the data exchange between the different actors, such as the grid operator and the REC operator, is modeled, which clearly addresses limitation (B). A promising method for the control problem of the described REC is model predictive control (MPC) and its solution by means of mixed integer linear programming (MILP). To control the community storage system using MPC, an accurate prediction of the residual load is required. This forecast is created using a machine learning model (LSTM model), which is pre-trained on synthetic load profiles to improve the prediction with limited data availability. Results show that MPC can handle the model and prediction uncertainties generally well. Nevertheless, the consideration of forecasting and modelling uncertainties in the simulation shows that the achievable self-consumption rates can be significantly reduced. Utilizing simplified forecasting and considering modelling uncertainty reduces the self-consumption by up to 14.5% points in the use case considered. These losses can, however, be contained by accurate prediction methods. With the consideration of modelling uncertainties for the CBES model alone, we find that a discrepancy between real storage capacity and controller model capacity of 20% reduces the self-consumption by 3.9% points. As the CBES can be considered a comparably “easy to model” component, this highlights the need for closed loop simulation in the controller development for RECs.
PV hosting capacity quantification is currently of importance for distribution system operators to manage upcoming PV installation requests, especially within low-voltage grids. The violation-mitigation-based (VMB) method presented a novel downward approach to quantify the PV hosting capacity in low-voltage grids, having the advantage of also quantifying possible expansions in the hosting capacity. The VMB method, however, requires typically hundreds of power flow simulations to quantify the hosting capacity in a low-voltage grid. In this paper, we address this issue by introducing sensitivity matrices to optimize the downward process of quantification. Evaluated in 79 low-voltage feeders from Austria, the enhanced VMB method reduces the number of iterations required from a median of 1764 to 7 while obtaining the same or higher hosting capacity values. The reduced computational effort permits the application of the method proposed at a large-scale. The application of the method proposed on the standard CIGRE residential low-voltage grid is also presented for possible comparison with other methods.
Grey Box models provide an important approach for control analysis in the Heating, Ventilation and Air Conditioning (HVAC) sector. Grey Box models consist of physical models where parameters are estimated from data. Due to the vast amount of component models that can be found in literature, the question arises, which component models perform best on a given system or dataset? This question is investigated systematically using a test case system with real operational data. The test case system consists of a HVAC system containing an energy recovery unit (ER), a heating coil (HC) and a cooling coil (CC). For each component, several suitable model variants from the literature are adapted appropriately and implemented. Four model variants are implemented for the ER and five model variants each for the HC and CC. Further, three global optimization algorithms and four local optimization algorithms to solve the nonlinear least squares system identification are implemented, leading to a total of 700 combinations. The comparison of all variants shows that the global optimization algorithms do not provide significantly better solutions. Their runtimes are significantly higher. Analysis of the models shows a dependency of the model accuracy on the number of total parameters.
Activation of heat pump flexibilities is a viable solution to support balancing the grid via Demand Side Management measures and fulfill the need for flexibility options. Aggregators as interface between prosumers, distribution system operators and balance responsible parties face the challenge due to data privacy and technical restrictions to transform prosumer information into aggregated available flexibility to enable trading thereof. Thereby, literature lacks a generic, applicable and widely accepted flexibility estimation method for heat pumps,which incorporates reduced sensor and system information, system- and demand-dependent behaviour. In this paper, we adapt and extend a method from literature, by incorporating domain knowledge to overcome reduced sensor and system information. We apply data of five real-world heat pump systems, distinguish operation modes, estimate power and energy flexibility of each single heat pump system, proof transferability of the method, and aggregate the flexibilities available to showcase a small HP pool as a proof of concept.
