## WP5 – Smart Grid based services

This task addresses key areas pointed out by the IEA and the EU, focusing on the development of integrated modeling, control and assessment tools to support the design and implementation of Microgrid (MG) and Multi-Microgrid (MMG) management strategies integrating micro and mini-generation technologies, distributed storage solutions, EV (including V2G mode) and load management techniques (following the demand-side management strategies developed in WP3). Resilience of these local electrical grids will be a key concern for the definition of these control solutions. These strategies will focus on four areas:

- To the extent possible, charge EV batteries with energy from renewables, thereby contributing to increase the amount of renewables that can be safely integrated in urban environment;
- Evaluate periodically the possibility of the MG/MMG pass from interconnected to islanded operation, providing information about the resilience of electrical networks integrating local generation;
- Evaluate the impact of the EV charging management algorithm mentioned above, together with the building demand response schemes developed in WP3, at the distribution network level and, in the event of major scale and area addressed, on the high voltage distribution and transmission levels;
- Design new markets and business models that enable new stakeholders (such as aggregators) to use the urban local load / generation management models, in order to provide ancillary services (e.g. reserves, congestion relief) to DSO and, when applicable, to TSO.

WP5 will include a testbed component and 6 major tasks, as follows:

#### Task 1

### Solar forecasting

Development of statistical forecasting algorithms for solar power generation for the very short-term horizon (up to six hours-ahead). These algorithms will run at a hierarchical distributed control and management architecture and will combine spatiotemporal information from sensors distributed in the urban MG (e.g., temperature, direct and horizontal irradiance, cloud-index, metered generation of PV panels). In order to handle this large volume of data, component-wise gradient-boosting algorithms, with feature selection ability, will be used and adapted to this problem. In addition, neural networks algorithms will also be explored for short-term forecasting.

#### Task 2

### EV charging optimization

Development of an algorithm to maximize the share of renewables in the energy used to charge EV. This will run at the hierarchical distributed control and management architecture level and will be fed with the load and solar power generation forecasts from the previous point and EV mobility data from WP4. It will then, taking into account the electricity network technical constraints, calculate for the next six hours the availability of EV charging points and the periods when EV should charge in order to maximize the usage of renewable energy. Finally, this information will be broadcasted to the EV in the neighborhood, allowing them to book the charging point in case of being interested.

#### Task 3

### Evaluation of the MG/MMG interconnection

Assessment of the conditions to pass from interconnected to islanded operation. To this end, several information will be taken into account, such as load and solar power generation forecasts, availability of storage devices and EV (including V2G model) and the responsiveness of other load in the network (building load that will be modelled in WP3). We will develop an algorithm to gather and process this information, having as output the information about the feasibility of passing from interconnected to islanded operation. The algorithm will run periodically in the upper layer of control, and will help to identify which feeders will be capable of operating under islanding conditions. The implementation of this resilience functionality will require a set of rules to be obtained from off-‐line studies, defining the conditions for which MG/MMG can operate in islanding mode, considering the composition of their controllable elements. We will test this second objective in a microgrid pilot installation (installed at INESC Porto Laboratory facilities) to demonstrate the technical feasibility of this integrative approach.

#### Task 4

### Assessment of the EV impact in the upstream networks

This task is related with a combined assessment of the impacts that the EV charging management algorithm, together with the building active demand response schemes developed in WP3, will provoke in the grids. To this end, these algorithms will be implemented in a set of representative urban distribution grids and their impacts in system operation will be sequentially analyzed on a bottom-up approach, until reaching the upper nodes of the distribution network and the respective node of the transmission network.

#### Task 5

### Design of new market models

We will design new market models envisaging the exploitation of controllable loads for ancillary services provision. In order to participate in the market, residential customers should be clustered by aggregating entities that will exploit their flexibility to sell system services. These new market agents (i.e. aggregators) represent a new and more effective ancillary service (reserve) resource for the DSO and TSO, but at the same time it is necessary to handle the uncertainty associated to its availability. These new market models are stochastic in order to tackle the inherent uncertainty by minimizing expected operational costs and the risk of reserve shortage, which increases the network operators’ confidence in using this new demand-‐side resource.

#### Task 6

### USD integration

The algorithms developed will be included and simulated in the USD. This simulation can demonstrate the feasibility of such an integrative approach as well as aid the calibration of the models and algorithms developed. The solar power forecast algorithm, together with the building demand calculated with the load management algorithms to be developed in WP3, are the inputs of the optimization algorithm for EV charging. The output from the optimization algorithm are broadcasted to the EV inside USD and modelled in WP4.