Fiche individuelle

Microgrids play a crucial role in modernizing the power grid by facilitating the integration of renewable energy sources. However, these sources exhibit high intermittency and stochastic behaviour, leading to challenges in effectively managing a microgrid amidst varying load demand and unexpected grid events. To address these uncertainties, local advanced control methods that leverage real-time data and enhanced computing capabilities are required. Frequency droop controllers generate additional power faster than the allocated power reserve to achieve an instantaneous balance of the electrical system. The automatic frequency restoration consists in making other generators participate gradually after 30 seconds. This paper proposes an intelligent control technique designed to enhance the static frequency droop controller, aiming to achieve active power balancing while minimizing CO2 emissions and operating costs. Consequently, an artificial neural network-based adaptive module is developed to anticipate and substitute the diesel-based reserve with a low carbon footprint reserve. This module considers new influencing input factors to anticipate and adjust the power setpoints of a stationary storage unit. The effectiveness of the proposed method is demonstrated on an islanded AC microgrid. The real-time simulation is conducted and validated on Opal-RT simulator, showing improved active power balancing while reducing both costs and CO2 emissions.

Renewable energy sources like photovoltaic systems (PV) often result in significant active power imbalances due to their variable nature. Despite their predictability, the uncertainty in their power production increases the complexity of planned power generation and necessitates the activation of costly and polluting power reserves commonly provided by conventional generators. This paper proposes an intelligent control technique to predict and replace diesel-based power reserves by faster, more economical and environmentally friendly reserves provided by a battery energy storage system (BESS). Following a power imbalance, a frequency droop controller ensures the primary reserve support from a BESS within a few seconds. Building upon the assumption that the variations of PV power are persistent within a 10-minute window, a secondary reserve for the next 5 minutes is commanded from an artificial neural network (ANN)-based adaptive module. This module predicts battery power references with a 1-minute time increment based on the PV and diesel generator power measured 5 minutes earlier. The proposed method's efficacy is showcased through its application on an islanded AC microgrid. Real-time simulations are performed using the Opal-RT simulator, revealing enhanced power balancing along with reductions in both operating costs and CO2 emissions.

The cost of energy and climate change due to the emission of greenhouse gases have yielded a large penetration of renewable energy sources in the grid and new types of loads (electric vehicles, . . .). The increasing variability and uncertainty in sources and demand lead to significant impacts on small electrical distribution networks. Thus, a stable, and reliable operation in local areas becomes more difficult to achieve. Adapted grid control according to the observed or expected variability in power flows and in model parameters of the electrical system, is required while maintaining the costs and CO2 emissions to low levels. This paper presents a state of the art on adaptive techniques and new self-learning methods for achieving grid balancing.