Search

We present a three-step method to perform system identification and optimal control of non-linear systems. Our approach is mainly data driven and does not require active excitation of the system to perform system identification. In particular, it is designed for systems for which only historical data under closed-loop control are available and where historical control commands exhibit low variability. In a first step, simple simulation models of the system are built and run under various conditions. In a second step, a neural network architecture is extensively trained on the simulation outputs to learn the system physics, and retrained with historical data from the real system with stopping rules. These constraints avoid overfitting that arise by fitting closed-loop controlled systems. By doing so, we obtain one (or many) system model(s), represented by this architecture, and whose behaviour can be chosen to match more or less the real system. Finally, state-of-the-art reinforcement learning with a variant of domain randomization and distributed learning is used for optimal control of the system. We first illustrate the model identification strategy with a simple example, the pendulum with external torque. We then apply our method to model and optimize the control of a large building facility located in Switzerland. Simulation results demonstrate that this approach generates stable functional controllers which outperform on comfort and energy benchmark rule-based controllers., Comment: 15 pages, Accepted for publication in TNNLS

To improve the accuracy of bifacial gain estimation, recent radiative models of solar energy systems have abandoned the traditional assumption of isotropic ground-reflected radiance. However, surface reflectance itself is still commonly considered as a constant — partly because of the recommendations of previous evaluations of reflectance models. This paper presents the findings of a new model evaluation based on a large database of measurements from 26 sites, which are representative of major land covers and climates. Both novel and previously reviewed formulations are validated with the data. On a global average, data-based estimation reduces mean absolute error by 22%, 29%, and 39% with constant, univariate, and bivariate models, respectively, compared to literature-based estimates. Only at the urban and snow-and-ice sites does time-variant estimation not notably improve accuracy. Arid sites tend to favour univariate models based on solar elevation, and diffuse fraction adds little value as the second predictor. By contrast, bivariate estimation clearly improves accuracy at vegetated and water sites. When considering the best-performing model for each site, the global average mean absolute error is 11%. Two novel formulations, univariate and bivariate, provide superior performance at many sites. The proposed 3-parameter bivariate model is one of the top performers at 19 out of the 26 considered sites.

Accurate forecasting of solar power generation with fine temporal and spatial resolution is vital for the operation of the power grid. However, state-of-the-art approaches that combine machine learning with numerical weather predictions (NWP) have coarse resolution. In this paper, we take a graph signal processing perspective and model multi-site photovoltaic (PV) production time series as signals on a graph to capture their spatio-temporal dependencies and achieve higher spatial and temporal resolution forecasts. We present two novel graph neural network models for deterministic multi-site PV forecasting dubbed the graph-convolutional long short term memory (GCLSTM) and the graph-convolutional transformer (GCTrafo) models. These methods rely solely on production data and exploit the intuition that PV systems provide a dense network of virtual weather stations. The proposed methods were evaluated in two data sets for an entire year: 1) production data from 304 real PV systems, and 2) simulated production of 1000 PV systems, both distributed over Switzerland. The proposed models outperform state-of-the-art multi-site forecasting methods for prediction horizons of six hours ahead. Furthermore, the proposed models outperform state-of-the-art single-site methods with NWP as inputs on horizons up to four hours ahead., 10 pages, 7 figures, accepted for publication in IEEE Transactions on Sustainable Energy

Accurate forecasting of photovoltaic (PV) and wind production is crucial for the integration of more renewable energy sources into the power grid. To address the limited resolution and costs of methods based on numerical weather predictions (NWP), we take PV production data as main input for forecasting. Since PV power is affected by weather and cloud dynamics, we model spatio-temporal correlations between production data by representing PV systems as nodes of a dynamic graph and embedding production data, geographical information and clear-sky irradiance as signals on that graph. We introduce a new temporal-spatial multi -windows graph attention network (TSM-GAT) for predicting future PV power production. TSM-GAT can adapt to the dynamics of the problem, by learning different graphs over time. It consists of temporal attention with an overlapping-window mechanism that finds the temporal correlations and spatial attention with a multi-window mechanism, which captures different dynamical spatio-temporal correlations for different parts of the forecasting horizon. Thus, it is possible to interpret which PV stations have the most influence when making a prediction for short-, medium-and long-term intra-day forecasts. TSM-GAT outperforms multi-site state-of-the-art models for four to six hours ahead predictions, with average NRMSE 12.4% and 10.5% on a real and synthetic dataset, respectively. Furthermore, it outperforms state-of-the-art models that use NWP as inputs for up to five hours ahead predictions. TSM-GAT yields predicted signals with a closer shape to ground truth than state-of-the-art models, which indicates that it is better at capturing cloud motion and may lead to better generalization capabilities.

The increasing penetration of renewable energy sources creates a challenge for the stability of current power systems due to their intermittent and stochastic nature. This paper presents the field results of an efficient demand response solution for controlling and adjusting the electric demand of buildings in an energy district through the activation of their thermal mass while respecting the occupants’ thermal comfort constraints. This multilevel control approach aims to support grid flexibility during peak times by constraining the energy exchange with the grid and increasing the self-consumption of the district. The results show a great potential for increasing the self-consumption up to 37% for offices, as well as improving the indoor environment, based on real data collected from a case study in Greece.

The high-penetration of Distributed Energy Resources (DER) in low voltage distribution grids, mainly photovoltaics (PV), might lead to overvoltage in the point of common coupling, thus, limiting the entrance of renewable sources to fulfill the requirements from the network operator. Volt-var is a common control function for DER power converters that is used to enhance the stability and reliability of the voltage in the distribution system. In this study, a centralized algorithm provides local volt-var control parameters to each PV inverter, which are based on the electrical grid characteristics. Because accurate information of grid characteristics is typically not available, the parametrization of the electrical grid is done using a local power meter data and a voltage sensitivity matrix. The algorithm has different optimization modes that take into account the minimization of voltage deviation and line current. To validate the effectiveness of the algorithm and its deployment in a real infrastructure, the solution has been tested in an experimental setup with PV emulators under laboratory conditions. The volt-var control algorithm successfully adapted its parameters based on grid topology and PV inverter characteristics, achieving a voltage reduction of up to 25% of the allowed voltage deviation. Peer Reviewed Objectius de Desenvolupament Sostenible::7 - Energia Assequible i No Contaminant Objectius de Desenvolupament Sostenible::7 - Energia Assequible i No Contaminant::7.2 - Per a 2030, augmentar substancialment el percentatge d’energia renovable en el con­junt de fonts d’energia Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles Objectius de Desenvolupament Sostenible::11 - Ciutats i Comunitats Sostenibles::11.6 - Per a 2030, reduir l’impacte ambiental negatiu per capita de les ciutats, amb especial atenció a la qualitat de l’aire, així com a la gestió dels residus municipals i d’altre tipus

The present deliverable presents the activities done within T2.3 on the development of DC-driven compression heat pumps for both the HYBUILD Continental system and the HYBUILD Mediterranean system. The heat pumps integrate the RPW-HEX (latent storage) described in D2.2 and have been modified to allow the operation in DC.

Operating power systems with large amounts of renewables requires predicting future photovoltaic (PV) production with fine temporal and spatial resolution. State-of-the-art techniques combine numerical weather predictions with statistical post-processing, but their resolution is too coarse for applications such as local congestion management. In this paper we introduce computing methods for multi-site PV forecasting, which exploit the intuition that PV systems provide a dense network of simple weather stations. These methods rely entirely on production data and address the real-life challenges that come with them, such as noise and gaps. Our approach builds on graph signal processing for signal reconstruction and for forecasting with a linear, spatio-temporal autoregressive (ST-AR) model. It also introduces a data-driven clear-sky production estimation for normalization. The proposed framework was evaluated over one year on both 303 real PV systems under commercial monitoring across Switzerland, and 1000 simulated ones based on high-resolution weather data. The results demonstrate the performance and robustness of the approach: with gaps of four hours on average in the input data, the average daytime NRMSE over a six-hour forecasting horizon (in 15 min steps) and over all systems is 13.8% and 9% for the real and synthetic data sets, respectively.

Power consumption in buildings show non-linear behaviors that linear models cannot capture whereas recurrent neural networks (RNNs) can. This ability makes RNNs attractive alternatives for the model-predictive control (MPC) of buildings. However RNN models lack mathematical regularity which makes their use challenging in optimization problems. This work therefore systematically investigates whether using RNNs for building control provides net gains in an MPC framework. It compares the representation power and control performance of two architectures: a fully non-linear RNN architecture and a linear state-space model with non-linear regressor. The comparison covers five instances of each architecture over two months of simulated operation in identical conditions. The error on the one-hour forecast of temperature is 69% lower with the RNN model than with the linear one. In control the linear state-space model outperforms by 10% on the objective function, shows 2.8 times higher average temperature violations, and needs a third of the computation time the RNN model requires. This work therefore demonstrates that in their current form RNNs do improve accuracy but on balance well-designed linear state-space models with non-linear regressors are best in most cases of MPC.

A full-scale pilot line for producing bifacial solar modules based on silicon heterojunction technology was set up as part of the European H2020 AMPERE project. In this study, the measurements from two outdoor test benches were used to evaluate the performance of the new modules under real operating conditions. The performance was benchmarked to that of commercial modules. A new bifacial irradiance model was validated based on the data and used to correct the bifacial gain estimates for differences in rear irradiance between the strings under comparison. The new modules increased the yield by 1-4% compared to the commercial alternatives.

Accurately estimating the state of charge of batteries is an essential part of battery management systems. Indeed, it conditions the ability to maintain the battery in a safe region with respect to ageing, and to ensure service availability when needed. This paper presents Bestimator™, CSEM’s novel estimation algorithm. This algorithm builds on a robust, interpretable electrical circuit model of the battery, and uses an extended Kalman filter to estimate the state of charge and correct parameters. Benchmarking against state-of-the-art estimators shows that Bestimator™ outperforms them by immediately reaching and maintaining over time an estimation error below 2%.

Despite a continuous reduction in battery prices, investment in storage systems carries a high risk. Indeed, many applications lack an adequate test protocol and method to predict the performance of batteries in operations. In this paper, we focus on lithium-ion battery energy storage systems (BESS), for which we introduce a novel method. With this method, predicting the efficiency of lithium-ion BESS becomes possible for any application from a single set of measurements. We show that evaluating the battery resistance as a function of the C-rate provides general “capability charts” based on which the efficiency can be evaluated under arbitrary operation profiles. Application-specific test procedures therefore become unnecessary. We have applied the novel procedure on three different lithium-ion cell technologies and illustrate its accuracy in the example of grid-connected BESS used for primary control reserve.

Power fluctuations, especially in LV grids, are increasing due to the continuing penetration of renewable energy sources which are naturally variable and characterized by fast transients of the energy production. In order to improve the power quality aspects, the authors show a test bench emulating a DC micro-grid comprising a PV system, various loads and an energy storage system. This micro-grid is interfaced to the AC public distribution grid through a single bidirectional converter. The design of the whole system has the following goals: i) supply the connected DC loads with a stabilized voltage; ii) guarantee the PV source operation in MPP in all irradiation conditions and dynamics, iii) limit the ramp-rate of the power exchanged with the AC grid and iv) improve the on.-site consumption of the locally produced energy. The feasibility of the solution is demonstrated by using off-the-shelves components for the hardware and developing an ad hoc software for the controller. The effectiveness of the solution is shown by experimental measurements of the power flows through the system in different conditions of generation and consumption.

Renewable energy sources (RES) are changing the operating conditions of power systems. In this context, battery energy storage systems (BESSs) could be a viable technical solution to maximize the local consumption of RES generation or to provide ancillary services to the grid. However, the lack of technical information about the real field performance is affecting the diffusion of BESS on the market. Here we present a technological comparison between different lithium-based storage technologies in terms of energy density, power density and efficiency based on the IEC 62660-1 standard for automotive. However, the standard measurements are not sufficient to describe the cell performance in the case of stationary applications. We therefore introduce two additional test protocols to evaluate the energy density as a function of the power density, and the efficiency as a function of the charge/discharge current rates. We discuss how these procedures improve the evaluation of performance and how they improve the system design and cost estimation for BESS in stationary applications.

Silicon nanowires (SiNWs) provide new opportunities for developing a new generation of thin film Si solar cells with enhanced light trapping and increased overall performance. Here, we report on the fabrication of SiNW-based thin film solar cells directly on top of low cost TCO/glass substrates in an all-in-situ process. The SiNWs are obtained on ITO (or SnO 2 ) substrates via vapor–liquid–solid growth mechanism, with the nano-scaled In (or Sn) catalyst droplets prepared by using H 2 plasma superficial reduction of ITO (or SnO 2 ) in a plasma enhanced chemical vapor deposition system (PECVD). We demonstrate that the morphology and compositional properties of the SiNWs, as well as the catalyst remnant in the SiNWs, can be effectively controlled by tuning the growth temperature and plasma conditions. The enhanced light trapping and absorption effects have been achieved by growing SiNWs directly on top of the TCO substrates, with the absorption edge downshifting steadily to ∼1.1 eV, indicating that the crystalline core of the SiNWs also participates in the light absorption. According to a real time monitoring using an in-situ MM-16 ellipsometer, the enhanced light trapping/absorbing effects can be attributed to the growth of long, sharp and straight SiNW. Prototype SiNWs-based thin film solar cells are successfully demonstrated.

The synthesis of silicon nanowires by Ga-assisted plasma enhanced chemical vapor deposition (PECVD) has been recently demonstrated. In the present work, we study in detail the structural characteristics of the synthesized nanowires. High resolution transmission electron microscopy (HRTEM) analysis reveals the existence of various types of structural defects, which can be classified mainly according to the orientation into axial twins, lateral twins, and transverse twins. We compare our results with previous studies of Si nanowires synthesized with other catalyst metals. Understanding both the origin and the effects of the observed defects is important for technological applications. The presence of twinned domains changes locally the structure of the material. As a consequence, one should find a different local density of states and band gap, which should result in a variation of the carrier transport and optical properties of the nanowires.

International audience; Nano-structured silicon thin films have been grown on tin-doped indium oxide (ITO) by Plasma-Enhanced Chemical Vapor Deposition (PECVD) at temperatures lower than 200ºC. Nanometer-scaled aggregates of metal (copper or gold) obtained from evaporated layers were necessary to initiate the nano-structuring growth. Different deposition conditions have been investigated. The highest aspect ratio was obtained with copper and high-pressure plasmas with SiH4 diluted in H2. The metals help dissociating silane so the deposition starts faster on the aggregates than around them, which leads to the nano-structuration. It is likely that the metal remains confined at the interface between ITO and silicon and do not diuse in the silicon layer.

International audience; The formation of gold islands on a transparent conductive oxide by annealing a thin evaporated layer is studied by SEM. The droplets of metal formed this way may catalyze the growth of nanowires or nanotubes to be used in optoelectronic devices. The effect of the initial thickness (between 1 nm and 4 nm) and the annealing temperature (between 70 °C and 650 °C) is investigated. A qualitative description of the evolution is given and quantitative data are obtained by automated image analysis. Number density, area and effective diameter of the aggregates depend heavily on the initial thickness but very slightly on the annealing temperature. Three phenomena are considered to explain the evolution: interfacial stress due to mismatch in the linear expansion coefficient, Ostwald ripening and kinetic aggregation. By comparing the experimental size distributions to the theoretical ones, the relevance of the last two is discussed. As an application, the gold catalyzed CVD growth of silicon nanowires on zinc oxide is demonstrated.

Smart grids offer an indisputable business opportunity for system operators and energy traders to engage in demand response programs. Hereby these actors may profit from trading flexibility provided at the prosumer side on the energy markets. This paper discusses the system design challenges for an information and communication technology system infrastructure facilitating DR for residential prosumers. It presents the SEMIAH framework for providing a scalable infrastructure for residential DR built on a component-based architecture with the virtual power plant at the hearth of the system. The paper examines the possible impact of deploying an automated residential DR program on the quality and stability of a low voltage grid.

Increasing penetration of PV sources in LV grids is introducing new challenges to the quality of voltage. Variability of solar radiation can cause power ramps of up to the nominal power of the plant in one second, which directly induces voltage fluctuations on distribution feeders. To mitigate this impact on power quality, we investigate the introduction of dc micro-grids interfaced to the ac grid through four-quadrant inverter. The introduced control strategy is able to manage the entire system without any communication layer and to ensure at the same time: i) ramp-rate control of the power exchanged with the ac grid; ii) voltage stability on both dc and ac grids; iii) maintaining the state of charge of storage devices connected to the dc grid. In the paper the effectiveness of the proposed strategy is shown with numerical simulations considering real profiles of solar irradiation and load power.

Hybrid concepts based on a nanoscale combination of organic and inorganic semiconductors are a promising way to enhance the cost efficiency of solar cells through a better use of the solar spectrum, a higher aspect ratio of the interface, and the good processability of polymers. A new type of solar cells has been investigated. It is based on a heterojunction between regio-regular poly(3-hexylthiophene) as an organic electron donor and silicon as an inorganic electron acceptor. In a first step towards nano-structured devices, cells made of flat thin films of these materials have been studied as a model case of the heterojunction. The materials were characterized through ellipsometry and absorption spectroscopy. The devices were studied by means of their spectral response and their I - V characteristics. By combining these results, the contribution of each layer and the mechanisms of photocurrent generation are explained. The best cells to-date show a power conversion efficiency of 1.6% under AM 1.5 illumination, with a V oc of 0.704 V and a J sc of 4.22 mA/cm 2 .

31st European Photovoltaic Solar Energy Conference and Exhibition; 2945-2951, Integration in the power system is becoming a limiting factor to the further development of photovoltaics. Proper quantification is needed to evaluate both issues and solutions; the share of annual electricity demand is widely used but we found that some of the metrics which are related to power rather than energy better reflect the impact on networks. Barriers to wider deployment of PV into power grids can be split between local technical issues (voltage levels, harmonics distortion, reverse power flows and transformer loading) and system-wide issues (intermittency, reduction of system resilience). Many of the technical solutions to these issues rely on the inverters as actuators (e.g., for control of active and reactive power) or as interfaces (e.g., for local storage). This role requires further technical standardisation and needs to be taken into account in the planning of power networks. Forecasting, storage, and combination with other renewable sources are interdependent solutions to solve the intermittency issue. Finally, we found that PV is also an opportunity to reduce some investment required to upgrade existing power networks. Through integration with micro-grids and hybrid generators, it can form the basis of novel power systems.

Low-tilt PV systems are increasingly common on flat roofs, where they enable a high power density. Yet there little is known about their requirements in terms of operations and maintenance, especially soiling. Here we present a case study on a 410 kWp system in Switzerland with a 5° tilt. Nineteen months of inverter data have been carefully analysed and correlated with consolidated weather data from four weather stations. Unlike systems on sloped roofs, which are deemed self-cleaning in Europe, this system demonstrates a clear and lasting benefit from active cleaning, with a 6% gain in energy yield.

Distributed renewable energy sources (RES), such as photovoltaic systems (PV), can contribute to a reduction in the power quality in distribution grids due to higher power oscillations both in absorption and generation. In this work, by means of a statistical approach, the quality of supply at the point of common connection is evaluated for different PV sizes and also for different levels of aggregated loads sharing a single PV installation as can happen in multi-occupancy building. The role of incentive schemes in Switzerland and UK in driving the choice for the installation of a specific PV size is evaluated.

Plasma-enhanced low temperature growth (300 degrees C) of silicon nanowires (SiNWs) and hierarchical structures via a vapor-liquid-solid (VLS) mechanism are investigated. The SiNWs were grown using tin and indium as catalysts prepared by in situ H(2) plasma reduction of SnO(2) and ITO substrates, respectively. Effective growth of SiNWs at temperatures as low as 240 degrees C have been achieved, while tin is found to be more ideal than indium in achieving a better size and density control of the SiNWs. Ultra-thin (4-8 nm) silica nanowires, sprouting from the dendritic nucleation patterns on the catalyst's surface, were also observed to form during the cooling process. A kinetic growth model has been proposed to account for their formation mechanism. This hierarchical structure combines the advantages of the size and position controllability from the catalyst-on-top VLS-SiNWs and the ultra-thin size from the catalyst-on-bottom VLS-ScNWs.

Silicon nanowires have been grown with gallium as catalyst by plasma enhanced chemical vapor deposition. The morphology and crystalline structure has been studied by electron microscopy and Raman spectroscopy as a function of growth temperature and catalyst thickness. We observe that the crystalline quality of the wires increases with the temperature at which they have been synthesized. The crystalline growth direction has been found to vary between111and112, depending on both the growth temperature and catalyst thickness. Gallium has been found at the end of the nanowires, as expected from the vapor-liquid-solid growth mechanism. These results represent good progress towards finding alternative catalysts to gold for the synthesis of nanowires.

We report an in-plane solid-liquid-solid (IPSLS) mode for obtaining self-avoiding lateral silicon nanowires (SiNW) in a reacting-gas-free annealing process, where the growth of SiNWs is guided by liquid indium drops that transform the surrounding a-SiratioH matrix into crystalline SiNWs. The SiNWs can be approximately mm long, with the smallest diameter down to approximately 22 nm. A high growth rate of >10(2) nm/s and rich evolution dynamics are revealed in a real-time in situ scanning electron microscopy observation. A qualitative growth model is proposed to account for the major features of this IPSLS SiNW growth mode.

International audience; In situ generation of indium catalyst droplets and subsequent growth of crystalline silicon nanowires on ITO by plasma-enhanced CVD are reported, and the wurtzite (Si-IV) phase is clearly evidenced in some wires

Diazonium salts coupled with vinylic monomers were able to functionalize carpets of carbon nanotubes by electro-activation in aqueous media. Hence, nano-sized carbon surfaces were shown to react with active radical species in solution through a new covalent grafting process called Surface Electroinitiated Emulsion Polymerization (SEEP), where diazonium salts are used both to initiate the polymerization of vinylic monomers in solution and to form a primer grafted polyphenylene-like layer on the carbon surface. SEM and TEM analyses revealed that ultrathin polymer films are grafted on the sidewalls of carbon nanotubes. XPS spectroscopy further confirmed the functionalization of multi-walled carbon nanotubes. (© 2008 WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim)

We here propose an all-in situ method for growing vapor-liquid-solid (VLS) silicon nanowires (SiNWs) directly on SnO(2) substrates in a plasma-enhanced chemical vapor deposition system. The tin catalysts are formed by a well-controlled H(2) plasma treatment of the SnO(2) layer. The lowest temperature for the tin-catalyzed VLS SiNWs growth in a silane plasma is ∼250 °C. The effects of substrate temperature and H(2) dilution of silane on the morphology and compositional evolution of the SiNWs were systematically investigated. The catalyst content in the SiNWs can be effectively controlled by the deposition temperature. Moreover, enhanced absorption (down to ∼1.1 eV) is achieved due to the strong light trapping and anti-reflection effects in the straight and long tapered SiNWs.

32nd European Photovoltaic Solar Energy Conference and Exhibition; 1444-1448, Forecasting and monitoring technologies for photovoltaics are required on different spatial and temporal scales by multiple actors, from the owners of PV systems to transmission system operators. In this paper the Grid integration working group of the European Technology and Innovation Platform – Photovoltaics (ETIP PV) reviews the different use cases for these technologies, their current status, and the need for future developments. Power system operations require a real-time view of PV production for managing power reserves and for feeding shortterm forecasts. They also require forecasts on all timescales from the short (for dispatching purposes), where statistical models work best, to the very long (for infrastructure planning), where physics-based models are more accurate. Power system regulations are driving the development of these techniques. This application also provides a good basis for a cost/benefit analysis since the forecasting error can be linked to the prices charged for energy imbalance.

The photovoltaics industry is adopting bifacial systems which offer improved energy harvesting compared to monofacial ones. This stems from the collection of light reflected by the surroundings on the rear side of the modules, leading to system-level gains typically ranging from 5% to 25%(rel). The question arises, however, whether this bifacial gain also applies to two-terminal (2T) tandem solar cells, since the series-connected subcells must generate equivalent currents to achieve an optimal performance. Using comprehensive simulations based on realistic device characteristics and typical meteorological data at different locations, we demonstrate that 2T tandem solar cells can indeed benefit from bifaciality provided that their design is tailored to this configuration. The top cell needs to absorb significantly more light than when designed for monofacial operation. In the geographical locations and system configurations simulated here, a broad performance optimum is found at a mismatch value of 6 mA center dot cm(-2) under front illumination, corresponding to a top cell generating similar to 23 mA center dot cm(-2) compared to similar to 17 mA center dot cm(-2) in the bottom one. With such specific design, bifacial tandems can yield up to 1.2 times the energy output of bifacial single-junction devices across a wide range of locations.