Scientific direction Development of key enabling technologies
Transfer of knowledge to industry

PhD : selection by topics

Technological challenges >> Solar energy for energy transition
3 proposition(s).

See all positions [+]

Diagnostic and prognostic tools for inverters and PV modules using machine-learning approaches

Département des Technologies Solaires (LITEN)

Laboratoire des Systèmes PV Appliqués

01-09-2021

SL-DRT-21-0347

sylvain.lespinats@cea.fr

Solar energy for energy transition (.pdf)

Framework: In the current context of climate change, the issue of energy is central from a societal point of view and from a political or economic point of view. Solar production, which is a renewable alternative to carbon-based energies, is growing exponentially and this rise in power will probably continue in the years to come. One of the best ways to lower the financial and environmental cost of solar power plants is automatic diagnostics, which can detect and correct plant failures and thus increase their performance. Basically, photovoltaic power plants are made up of modules connected to an inverter. The modules produce direct current which is converted into alternating current by the inverter for transport on the distribution network. The failures and aging of these devices are the main source of non-trivial failures. For example, the lifetime of a power plant is generally estimated at 20 or 30 years, to be compared to only 10 years for an inverter. It is very common for the current and voltage upstream and downstream of the inverter to be monitored. These data are generally supplemented by meteorological measurements (irradiance and temperature in particular). These data are however under-exploited. In the case of the behavior of the modules, it is mostly due to the strong correlation with various factors including daily and seasonal phenomena, weather conditions, relative position of the sun, non-linear interactions between the different modules, aging continue, break, etc. In the case of inverters, the difficulties are mainly due to the strong dependence on the operating conditions and on the noise level of the measurement largely higher than the signal (as encountered in the context of the detection of gravitational waves by the LIGO project). Objective: From these data we want to provide a close monitoring of photovoltaic plants, diagnose failures and anticipate them. In that goal, based, on the one hand, on the very large amount of data which can counter the signal-to-noise problem, and on machine-learning on the other hand, we will isolate the different explanatory components. Firstly, the modules and inverters will be considered separately. Secondly, we will consider the system as a whole. In the past, the LSPV (CEA) and LAMA UMR 5127 (Savoie Mont Blanc University) laboratories have collaborated on the development of dimensionality reduction methods. These methods (probably to be adapted) make it possible to explore the datasets in order to extract behaviors that can be linked to various modes of operation and aging. This step will allow definoing classes for regression / classification methods. The final goal is a diagnostic tools deployable onto power plant monitoring systems. Desired profile: We are looking for a student in mathematics interested in applications in the field of renewable energy and electronics or a student in engineering sciences passionate about mathematics. Experience in electronics is not necessary, but the candidate may operate measurements in laboratory under the supervision of electronics and photovoltaic engineers to produce data or confirm behavior. The tools used may include dimension reduction methods, statistics (descriptions and tests), time series analysis, SVMs, neural networks or tensor methods.

Download the offer (.zip)

Multiphysical design of high-voltage power semiconductor modules for renewable energy conversion

Département des Technologies Solaires (LITEN)

Laboratoire Systèmes PV

01-09-2021

SL-DRT-21-0387

jeremy.martin@cea.fr

Solar energy for energy transition (.pdf)

Research and development around silicon carbide (SiC) power semiconductors provides samples that can withstand voltages up to 15kV. These devices switch at very high speeds (e.g. 120kV / µs for a 10kV SiC MOSFET or 180kV / µs for a 15kV SiC IGBT). Overall, the performances of these semiconductors are exceptional, and drastically reduces the switching losses compared to Silicon equivalents. The implementation of these switches is on the other hand very delicate and calls upon methodologies of multiphysics design in transversal disciplinary fields. It is, from the scientific literature addressed a number of scientific and technological obstacles that we can list: -Minimization of parasitic inductors of power modules (<5nH) -Integration of EMC shielding to collect disturbing impulse currents -Cooling of SiC chips so the size is very small compared to a Si equivalent -Management of partial discharges and dielectric materials -Influence of dV / dt on the aging of materials (in DC, at 50Hz, and in pulse) -Reflection phenomena (electromagnetic wave) The proposed work consists of studying and proposing a power module architecture integrating innovations making it possible to address the implementation of SiC chips up to 10kV. The teams from the CEA in Toulouse specialists in high power 3D packaging will provide their skills in assembly technologies for the production of complex power modules. The CEA teams at INES campus (Nat. Inst. of Solar Energy)located at the Bourget du Lac (Savoy) will provide their high voltage measurement and prototyping means as well as their knowledge in power module design (finite element simulation). Researchers from G2ELAB in Grenoble in cooling of power modules and dielectric science will use their knowledge as well as their experimental platforms.

Download the offer (.zip)

Perovskite materials: influence of nanocrystallization processes on the performances of PK cells for tandem integration

Département des Technologies Solaires (LITEN)

Laboratoire des Cellules Tandem

01-10-2021

SL-DRT-21-0526

noella.lemaitre@cea.fr

Solar energy for energy transition (.pdf)

After several decades of intense development, silicon-based PV is a well-established and mature technology that now nears its practical efficiency limit. One strategy to overcome this limit is to use multi-junctions solar cells coupling a silicon subcell with a topcell based on a higher band gap absorber (1,6 - 1,7 eV vs 1.12 eV for silicon) exhibiting high efficiency. Lead halide perovskites (with ABX3 structure) can fulfill these requirements. Such materials can be integrated via solution processing at low temperature and yield theoretical efficiencies well beyond 30% when combined in a tandem device with a silicon subcell. To this end, controlling perovskite crystallization from precursors solution is obviously of prime importance. Also, the development of upscalable techniques to process the perovskite is one of the main practical challenge still to be met to ensure a practical deployment of the technology. The strategy currently developed within CEA relies on a gas-quenching of the precursors wet film to trigger perovskite crystallization. The main goal of this PhD will be to study in-depth the perovskite crystallization when processed in such conditions, to pave the way towards successful integration in tandem devices.

Download the offer (.zip)

See all positions