From home energy management system local flexibility to low-voltage predictive grid management

This study presents Integrid's project framework to manage low voltage (LV) electrical networks, aiming to avoid both technical and quality constraints, induced by the ever-increasing amount of flexible resources spread all over the grid. These assets cover a large amount of renewable-based energy generation to electrical vehicles and energy storage units. For this to be possible, new advanced tools were developed to exploit the benefits of the so-called distributed energy resources, while overcoming limitations on the metering and communication infrastructures. Hence, this study describes the approach taken to perform the active management of LV networks, without a perfect level of observability, exploiting the flexibility provided by the distribution system operator's resources combined with the one offered by private consumers through the home energy management systems. Additionally, some results followed by a brief discussion are presented, enforcing the success of the developed tools. The algorithms within these tools allow to forecast both microgeneration, available flexibility and load profiles, as well as to estimate the network's state, at different time frames.


Introduction
InteGrid project aims to demonstrate the feasibility of smart distribution networks, coping with a high amount of renewable energy sources (RESs), and making use of the available DERs flexibility for various functions / business cases at different levels. The project includes the development and the demonstration of a LV state estimation (LVSE) and a LV control (LVC) tools, making use of smart meter data and a complementarity of multiple technologies like grid asset data, offering different types, degrees and levels of flexibility, such as Demand Response (DR) at domestic level and energy storage (at a utility scale and domestic scale). This paper highlights one of the key interfaces of the new energy value chain, the interaction between end-consumer and the remaining energy actors, enabled by the digitalization of the power systems, which have favoured a multilevel deployment of smart devices and systems. It will also address InteGrid's low voltage ecosystem and the approach chosen for DR through a fully integrated architecture that covers not only the DSO environment but also LV consumers, individually or aggregated, and considering not only residential consumers but also commercial and public facilities. This complex ecosystem, presented in Figure 1, needs a set of different software modules that will enhance the capability of the DSO of understanding the load profile of the grid by predicting its behaviour, identifing any upcoming constraints, and validating which are the available flexibilities to solve the predicted constraints, without impacting customer's comfort standards. As far as flexibility estimation is concerned, a new philosophy is introduced considering data recordings and transmitting rates that smart meters can provide, bundled with a Home Energy Management System (HEMS) that offers native interoperability with home appliances and systems. For InteGrid project, a HEMS system has been developed with specific functionalities that enable end-customers to have full control of their assets and with minimum effort. Added value is obtained from the provision of flexibility since it enables new business models not only for themselves but also for all the energy value chain.

Background
To address these challenges, the InteGrid Portuguese demonstrator was implemented over an already existing architecture, updating several elements and adding others to reach the objectives. The existing Outage Management System (OMS) was replaced by an Advanced Distribution Management Systems (PowerOn Advantage from GE) integrated with third party systems of EDP Distribuição (GIS, AMI, DMS, SGL, EI-Server), and the grid management tools developed by INESC TEC were incorporated. The operation of the ADMS is ensured by near-real-time and historical measurements as well as by the available flexibilities gathered by the tools, allowing producing forecasts and setting control strategies to solve foreseen network problems. A Grid-Market Hub (GM-hub) provides, on this architecture, the interface between the different actors. Figure 1 shows the general architecture of this solution.

2-Distributed Monitoring and Control of LV Networks
As previously mentioned, the proposed control approach encompasses three core tools (LVSE, LVC and Forecast), and a gateway HEMS supported by two platforms, the ADMS and the GM-hub.

Low Voltage State Estimator (LVSE) Tool
The LVSE [1] is a data-driven state estimator that uses historical data and a small subset of smart meters with near real-time communication of active power and voltage magnitude measurements, bringing the following innovations: a) Uses a very limited number of real-time metering points (if smart meter measurements are sent every 24 hours or every week).
b) The output takes the form of quantiles representing uncertainty in the state estimation process. This enables the creation of probabilistic alarms for voltage problems, e.g. 95% probability of overvoltage in a specific node. c) Neither topological information, nor electrical characteristics of the elements of the grid are necessary. d) Exogenous variables can be integrated in the state estimation process. The LVSE provides a full snapshot of the LV network operating conditions (including a probabilistic characterization of the electrical measurements) and it is used to generate voltage alarms that trigger the LVC module. For instance, the DSO can define which kind of alarms generated by the LVSE, or control actions determined by the LVC, require its notification and/or validation. The central principle behind the proposed LVSE is to search for analogue events in the historical dataset, using a set of explanatory variables as depicted in Figure 2. The method combines measurements collected by the subset of smart meters with real-time communication and the MV/LV substation meter, together with voltage observations from the previous day and from all the meters installed in the LV grid.

Low Voltage Control (LVC) Tool
The LVC tool [2] has the following innovations: a) Exploits forecast data to run a predictive management of the available flexible resources, including customer flexibility, public lighting flexibility, and DSO owned resources such as Energy Storage (ES) devices and transformers with On Load Tap Changing (OLTC) capabilities. b) Uses pre-booked flexibility from domestic clients to manage voltage deviations.
c) Re-evaluates the conditions of the LV grid in real-time using the most recent data available. d) Uses a visual interface for the dispatch centre operator in which flexibility activation orders are validated and after updated on the HEMS [3].

Preventive Active Network Management
The tool enables an active management of the voltage profiles of LV networks in a preventive manner, using forecasts for the n-hours ahead of the energy consumption, renewable generation and flexibility provided by different types of DERs present in LV networks. These DERs can be property of the DSOsuch as OLTC transformers and ES devicesor owned by consumers willing to participate in the operation and management of the LV network. The latter are connected to the GM-hub via a HEMS to establish their participation in the grid management actions. The HEMS is responsible for the optimization of the customer's consumption and for being the interface with the GM-hub, communicating the available flexibility that the customer has for the next hours, which can be used by the DSO to support grid operation and to receive the flexibility activation set-points according to the operation scenario. The LVC tool has two main operation modes, or submodules: the preventive control submodule and the real-time control module. The preventive control submodule initially performs an analysis of the voltage profiles in the LV network, using forecasts of the energy consumption and production. With an unbalanced three-phase power flow algorithm, it estimates the voltage values in the LV network's nodes for each instant of the n-hours ahead. A control action plan is then defined for the available flexibilities at each instant of the n-hours ahead. The resources to activate are selected according to a merit order, which prioritizes DSO-owned resources relatively to private consumers' flexibility, with the objective of reducing the operational costs. Once a control action plan is defined, the flexibility for each period is pre-booked, thus reserving the amount of flexibility required for each node and for each period of the n-hours ahead. The objective of the real-time control submodule is to assess the adequacy of the control action (previously defined by the preventive control module) to the current grid state. It uses the LVSE presented before to obtain the voltage magnitude values in the network nodes. If the control action is no longer valid, which means the grid conditions are significantly different from the forecasts, the real-time control submodule updates the previously established control action plan to reflect the current grid conditions, using the flexibilities available in the network at that instant.

Forecast
Forecast mainly works together with the preventive control mode from the LVC. The periodicity of this scheduled action is aligned with the availability of new load and RES forecast data (e.g. every 6 hours). Once the preventive control mode is triggered, it acquires the required data to build the network snapshots for the hours ahead, i.e. load and RES forecasts, from the forecasting systems, and private consumers' flexibility, from the GM-hub.

Home Energy Management System (HEMS)
The HEMS are gateways installed on customer premises composed of several functional modules that provide overall the necessary functionalities that support its operation for energy management in domestic buildings. The available modules have different purposes: • Interface Manager: defines the interconnections between HEMS, and the external apps and supporting services.

3-Demonstration Site
For this paper, the results we analysed are based on the rollout of the project on one of the two LV grid demonstrations in Valverde. [4] Valverde is a small rural village in the countryside of Évora, the first Portuguese smart city in the National Project InovGrid, with a population of 450 inhabitants, all connected to the LV grid. Valverde was also the chosen location for the Horizon 2020 SENSIBLE project, because it already has many flexibility sources in place (e.g. DSO's LV grid storage) which makes this village a natural choice for the InteGrid project. The demo presented in Figure 4 covers 2 secondary substations, 241 customers, 39 of which joined the demonstration and are equipped with smart meters, photovoltaic panels (1.5kWp), storage (3.4 kW) and electric water heaters.

4-Demonstration Rollout
The demonstration was divided in 4 different steps as shown in Figure 5. The first step focuses on the enablement of flexibility by the end customer added to the DSO directly controlled resources [5]. The second step performs the detection of constraints by the tools. These constraints will be presented on the ADMS as an alarm.
Step three compiles a flexibility activation plan, so the constraint can be mitigated. Once a control action plan is defined, the flexibility for each period is pre-booked, thus reserving the amount of flexibility required for each node and for each period of the n-hours ahead.
Step four makes use of the real-time control module to assess the adequacy of the control action (previously defined by the preventive control module) to the current grid state. It uses the LVSE presented before to obtain the voltage magnitude values in the network nodes. If the control action is valid, the constraints will be solved or mitigated. If the defined plan is no longer valid (which means the grid conditions are significantly different from the forecasts), the real-time control module updates the previously established control action plan to reflect the current grid conditions, using the flexibilities available in the network at that instant.

5-Results
Several tests have been running after the infrastructure was deployed within the real environment. This means that operation actions are being performed on several flexibility assets, not only on grid/DSO controlled assets, but also inside customer premises through the interaction with the HEMS. The three results presented cover a limited time for one secondary substation in Valverde. For this instant, the LVC algorithm forecasted 13 grid nodes with voltage violations (violations were defined as Vn+5% or Vn-4%).
The first result addresses the Violation Frequency Reduction (VFR) after flexibility plan was applied: The next result represents the percentage of total flexibility available from private consumers that was used to address the detected constraints.
Where FA% is the percentage of the total flexibility from private consumers available in the network that was activated for grid management purposes, for the period flex,LVCt is the total flexibility from private consumers available in the network at each time instant used for grid management purposes and flext is the total flexibility from private consumers available in the network at each time instant .

6-Conclusions
Early results achived on InteGrid project are clear and show that an effective operational grid management strategy supported on flexibility can improve the quality and continuity of service mainly when applied to grids with high penetration of RES and EV, it may not address all the nodes with violations, mainly due to the position in the feeder of flexible controlled assets regarding the node where the violation exists.
Nevertheless, the global output of the activation plan will definitely improve both quality and continuity of service.
Two key aspects are being identified along the project, the importance of a simplified architecture ( Figure 1) with data management capabilities that can compute heavy data sets, and the most relevant aspect, keep clients engaged making them the most relevant actor on the developed ecosystem.

7-Acknowledgements
The research leading to this work is being carried out as a part of the InteGrid project, which received funding from the European Union's Horizon 2020 Framework Programme for