Coordinated Voltage Control in Distribution Grids Leveraging Local Flexibility and Direct Control of a Large Battery Storage

Coordinated Voltage Control in Distribution Grids Leveraging Local Flexibility and Direct Control of a Large Battery Storage Edoardo De Din ICE–1 Energy Systems Engineering Forschungszentrum J¨ ulich 52428, J¨ ulich, Germany [email protected] Daniele Carta ICE–1 Energy Systems Engineering Forschungszentrum J¨ ulich 52428, J¨ ulich, Germany [email protected] Andrea Benigni ICE–1 Energy Systems Engineering Forschungszentrum J¨ ulich 52428, J¨ ulich, Germany [email protected] Abstract Interaction between electrical and thermal systems can amplify the impact of load variations on voltage profiles. Control algorithms can leverage the flexibility of the Distributed Energy Resources to compensate for undesired voltage violations. This paper therefore presents a coordinated voltage control system for a medium voltage distribution grid, which was tested using the HIL setup of the Forschungszentrum J¨ ulich campus. This setup is enhanced by integrating real-time measurements of electrical and thermal demands collected from the field. The voltage control is then tested in an undervoltage scenario, with the control set-point applied to a large battery energy storage system installed on the campus for one of the grid nodes. Index Terms Voltage control, Hardware-in-the-Loop simulation, Battery energy storage system, Networked control systems I. INTRODUCTION The increased installation of Distributed Energy Resources (DER)s at the distribution grid level could potentially result in voltage operating limits being exceeded [1]. Moreover, if the electrical system is integrated with the heating system, e.g. through the use of Heat Pump (HP)s, a corresponding increase in the amount of power requested at the distribution grid level is to be expected, leading to possible congestion or voltage violations [2]. Therefore, in order to mitigate these undesirable conditions, the development of advanced control strategies using real-time measurements is required [1]. These control strategies can exploit the aggregated flexibility of the DERs located at Low Voltage (LV) level for the real-time voltage control [3], [4]. The effectiveness of the control algorithms can be evaluated using a Hardware-in-the-Loop (HIL) set-up that combines electrical and thermal systems, as presented in [2], [5]. However, none of these studies used real-time measurements from the fields. Furthermore, while there are some examples of the implementation of the control signals on real Battery Energy Storage System (BESS), only a few have considered large systems [6], [7]. However, these works have focused on providing fast ancillary services [6] and on solving instability problems at the transmissions level [7], thus excluding the integration of the BESS in the Information and Communications Technology (ICT) platform and the use of the large BESS in voltage control of the distribution grid. Hence, this work presents the HIL testing of a Distributed Voltage Control (DVC) for the Medium Voltage (MV) level, which utilizes the flexibility calculated for the LV level while taking into account the impact of the Low Temperature District Heating (LTDH) network. The developed control strategy is applied to a portion of the distribution grid and of LTDH network of the Forschungszentrum J¨ ulich (FZJ) campus, evaluated in a HIL set-up. This set-up integrates the real-time data measured from the field, collected in a ICT platform [8], to obtain a realistic simulation. Finally, the control set-point calculated for one of the MV nodes is applied to a high-power BESS from Riello (1.5 MW / 0.5 MWh) installed on the campus. The implementation of control signals is enabled by the utilization of a Legacy Protocol Converter (LPC), which facilitates the conversion of signals into the protocol employed by the BESS. This test verifies the applicability of the DVC to large BESS as well as validates the ICT platform, which connects the HIL laboratory with the field for the sending and receiving of messages. In summary, the novel contributions of this paper are: (1) the integration of real-time measurements of the electrical and thermal demands in the HIL set-up, by means of the developed ICT platform; (2) the implementation of the DVC control signal to the Riello BESS installed on the campus. This work has been partially supported by funding from the European Commission Horizon Europe Programme under Grant Agreement No. 101096511. Views and opinions expressed are however those of the author(s) only and do not necessarily reflect those of the European Union or CINEA. Neither the European Union nor the granting authority can be held responsible for them. 10 kV 10 kV 35 kV 400 V Power Grid BESS LTDH Base PV HP BESS Load LTDH Grid Gas Free Central JSC Boiler Cooler Cooling Fig. 1. Schematic overview of the coupling between the electrical grid and the LTDH network. II. ELECTRICAL GRID AND LOW TEMPERATURE DISTRICT HEATING NETWORK OF THE CAMPUS This work considers the electrical distribution grid and the LTDH network of the FZJ campus located in J¨ ulich, Germany. The electrical grid is composed of several MV distribution feeders at 10 kV to which office and laboratory buildings are connected via MV/LV transformers (10 kV to 0.4 kV). As shown in the green box in Fig. 1, each building is assumed to have a non-controllable base load and could be equipped with BESS, Photo-Voltaic (PV) and HP. In addition to the buildings, the electrical grid of the campus is comprised of large-scale DERs, e.g. a large free field PV plant and a high-power BESS from Riello, connected to the MV grid by means of the MV/LV transformer. Moreover, as described in Fig. 1, a LTDH network, which utilises the waste heat generated by the J¨ ulich Supercomputing Centre (JSC), connects seven buildings surrounding the JSC. Each of these buildings is equipped with its own HP for heat supply to increase the temperature of the LTDH network to the required level of the building. A schematic representation of the coupling between electrical and LTDH network is described in Fig. 1, which shows how, by means of the HPs, the two networks are connected. III. CALCULATION OF THE LV FLEXIBILITY This section describes the algorithm employed to obtain the aggregated flexibility, which, in this work, is considered exclusively in terms of active power flexibility. The calculation of the flexibility for all buildings connected to the distribution grid of the campus is achieved by applying a power aggregation methodology, based on the inner-box approximation method proposed in [9], which represents a simple approximation of the real aggregated feasibility region with a box-shape feasible region. Thus, the maximal upper and lower operational flexible margins over time T, defined as flexibility trajectories, are obtained by solving (1)–(8), which represents the Maximal-Flexibility Power Aggregation (MPA). 1) Cost Function: max T X t=1 (P∧ i(t)−P∨ i(t)) ·∆t(1) where P∧ i(t)and P∨ i(t)represent the maximal and minimal active flexibility power at transformer i, defined as: P∧/∨ i(t)=Pi,load(t)+P∧/∨ i,BESS(t)+P∧/∨ i,PV (t)(2) where the sum of the active power consumption of the base load and HP is denoted by Pi,load(t)∈R, whereas P∧/∨ i,PV (t)∈R and P∧/∨ i,BESS(t)∈Rdefines the maximal and minimal flexible power for the PV and BESS respectively. In this work, only the maximal and minimal flexibility trajectories of BESS and PV are considered, given that the base load of the buildings and the HPs are regarded as non-flexible components. 2) Model Constraints: The constraints for the aggregated model at the transformers are described as follows: Pmin i,PV (t)≤P∧/∨ i,PV (t)≤Pmax i,PV (t)(3) Pmin i,BESS ≤P∧/∨ i,BESS(t)≤Pmax i,BESS (4) (P∧/∨ i,BESS(t))2≤(Prated i,BESS )2(5) SOCi,BESS(t+ 1) = SOCi,BESS(t)+ηBESS Pi,BESS(t) Ei,BESS ∆t(6) where Pmax i,PV (t)and Pmin i,PV (t)is the maximal and minimal active power of PV. The variable Pmax i,BESS,Pmin i,BESS and Prated i,BESS are the BESS maximal, minimal and rated active power. The State of Charge (SOC) of BESS is defined as SOCi,BESS, with Ei,BESS the BESS capacity, and ηBESS the efficiency of BESS, considered identical for both charging and discharging modes. For each transformer, the apparent power balance is defined as: (P∧/∨ i(t))2+ (Qi,load(t))2≤(Srated i,trafo )2(7) where Qi,load is the reactive power consumption of the base load and HP and Srated i,trafo is the rated power of the MV/LV transformer. 3) Joint Constraints: To guarantee the disaggregation flexibility the maximal trajectory should be always larger than the minimal one, as follows: P∨ i,BESS(t)≤P∧ i,BESS(t); P∨ i,PV(t)≤P∧ i,PV(t)(8) The result of the MPA as the maximal and minimal flexibility for the aggregated power at the MV/LV transformer are then provided to the each control node of the DVC to perform the voltage control at the MV level. As detailed in Section I, the FZJ campus features large DERs with individual connections to the MV grid (through a MV/LV transformer), e.g. the Riello high-power BESS, which are also regarded in this study as flexible resources for voltage control. In this case, the cost function and constraints still apply; however, they are adjusted to the specific DER without the need for aggregation. IV. MV DISTRIBUTED VOLTAGE CONTROL The control of the MV distribution grid is accomplished through the implementation of the DVC algorithm. This approach enhances reliability and flexibility due to its plug-and-play functionality, which facilitates seamless integration and ensures operational efficiency [10]. This work refers to the DVC algorithm for the control of active and reactive power presented in [1]. However, in addition to these preceding studies, the flexibility margins provided for each MV node by the MPA are considered as constraints for the calculation of the control set-points. For a node i∈ N in the communication graph, with N={1,2, ..., N}the set of nodes, and k∈Nthe iteration of the algorithm, the dual ascent steps for the active power control result in: λmax i(k+ 1) = [λmax i(k) + αi(Vi(k)−Vmax)]≥0(9) λmin i(k+ 1) = λmin i(k) + αi(Vmin −Vi(k))≥0(10) µmax i(k+ 1) = µmax i(k) + γP(¯ Pi(k)−P∧ i(t))≥0(11) µmin i(k+ 1) = µmin i(k)+γP(P∨ i(t)−¯ Pi(k))≥0(12) where Vi(k)is the measured voltage and ¯ Pi(k)the active power control set-point calculated at iteration k. Variables λmax i, λmin iare the Lagrangian multipliers associated to the voltage (Vmax, V min ∈R) constraints, whereas µmax i,µmin iare the Lagrangian multipliers associated to the power limits (P∧ i(t), P ∨ i(t)∈R) defined by the MPA stage, which are considered constant for the iterations k∈ {t−1, t}. Furthermore, αi, γP∈Rare the step sizes for the Lagrangian multipliers update, calculated following [1], whereas the [·]≥0operator represents the projection into the positive orthant. The second step, that is the minimisation of the primal variable, follows: ¯ Pi(k+ 1) = −(λmax i(k+1)−λmin i(k+ 1)) −Bi,i(µmax i(k+1)−µmin i(k+ 1)) −X j∈Ni Bi,j(µmax j(k+1)−µmin j(k+ 1),(13) where Bis the real part of the admittance matrix and Niis the set of neighbouring nodes of i. As described in (13), only the Lagrangian multipliers are exchanged among the nodes in the communication graph. According to (11) (12), these multipliers attain values different from zero only when either the maximal or minimal flexibility is reached. Thus, when a node ireaches its maximum flexibility threshold, it communicates the non-zero Lagrangian multipliers to the adjacent nodes, which adjust their control set-points accordingly, even if they are not directly measuring any voltage violations. Once the set-points are calculated for the MV level, they are disaggregated for the flexible resources at the building level while ensuring the feasibility of the decomposed decision [9]. For each transformer i(i.e., we assume one transformer at each node), the auxiliary coefficient ξi(k+ 1) ∈[0,1] is defined as: ξi(k+ 1) = P∧ i(t)−¯ Pi(k+ 1) P∧ i(t)−P∨ i(t)(14) hence ¯ Pi(k+ 1) = ξi(k+ 1)P∨ i(t) + (1 −ξi(k+ 1))P∧ i(t). Consequently, the disaggregated set-points for the flexible DERs result in: ¯ Pi,BESS(k+ 1) = ξi(k+ 1)P∨ i,BESS(t) +(1−ξi(k+ 1))P∧ i,BESS(t)(15) ¯ Pi,PV(k+ 1) = ξi(k+ 1)P∨ i,PV(t) +(1−ξi(k+ 1))P∧ i,PV(t)(16) which satisfy the MPA constraints. In case of large DERs that are individually connected to the MV grid, the control set-points calculated by the DVC are applied directly, thereby eliminating the necessity for disaggregation. V. INTEGRATION OF THE MEASUREMENTS FROM THE CAMPUS The electrical and thermal measurement data from the campus are collected from the sensors that have been installed to monitor, among others, electrical power injections/absorption and heating demand at the building level. Electrical measurements are collected from more than 80 PQI-DA smart power quality (PQ) meters manufactured by the company a-eberle [11]. These devices are in compliance with IEC 61000-4-30 Ed.3 (2015) standard for an A-Class device [12] and are able to provide synchronized measurements of voltage and current with an accuracy of 0.1 %, at the fundamental frequency. Real-time data can be collected from the PQI-DA smart using the Modbus TCP protocol. In line with [12], different Modbus registers can be used to collect data with different time aggregation intervals (i.e., from 10 ms to 2 h). By leveraging this functionality, the most relevant quantities are collected every second from these devices, and stored on the campus FIWARE-based ICT platform [8]. Such data are then available to researchers as historical data. Authorized users can access them via the Message Queues Telemetry Transport (MQTT) broker of the platform. Nevertheless, for real-time and latency-sensitive applications, monitoring data of interest can be collected directly from the PQI-DA smart via Modbus. To monitor the thermal quantities of the LTDH, pressure and temperature sensors from various manufacturers are installed at the inlet and outlet of the LTDH components (e.g., collectors, heating distributors, heat pumps, thermal storages, and more). As a general reference, the base components inside each building are: heat-pump (with evaporator, condenser and compressor), thermal storage, and radiators. Excluding the radiators, this leads to at least 8 different temperature measurement points for each building of the LTDH. Another important aspect to underline is that these measurements are not provided with a fixed reporting rate, as for the electrical quantities, but on the change of value with an accuracy of 0.5 K and resolution of 0.1 K. Due to the heterogeneous nature of such components, the corresponding monitored values are collected using dedicated BACnet-to-MQTT adapters. In fact, as mentioned for the electrical measurements, also thermal measurements are stored in the FIWARE-based ICT platform of the campus. Nevertheless, contrarily to the previous case, these quantities cannot be gathered directly from the devices on the field. Thus, the MQTT broker of the ICT platform is the only collection point for such data. VI. LEGACY PROTOCOL CONVERTER FOR BESS CONTROL The BESS considered in this work is a 1.5 MW/ 0.5 MWh system from Riello. It is composed of 3 inverters, model SPS HE 500, with a nominal power of 500 kW each. The system, permanently connected to the campus network, operates 24/7 as a Uninterruptible power supply (UPS) and it is also used for scientific purposes. The communication with each inverter of the Riello BESS is established using two different cards. The NetMan 204, and the EnergyManager for monitoring and controlling purposes, respectively. In both cases, the communication protocol used at FZJ is Modbus TCP. In this framework, to monitor and control the Riello BESS, the LPC developed in the context of the European Project InterSTORE [13] has been deployed. The LPC is an open source tool that acts as a middleware, allowing devices using different communication protocols to exchange data using the IEEE2030.5 standard [14]. It can handle IEEE2030.5 messages in both JSON and XML formats, uses the Neural Autonomic Transport System (NATS) messaging protocol, and can translate MQTT and Modbus messages into IEEE2030.5 format. Thus, the LPC facilitates the conversion of the MQTT message transmitted by the DVC into Modbus [15], by firstly converting it into an IEEE2030.5 format. Due to its flexibility and efficient way of translation, it has been deployed on a Raspberry Pi (RPi) using a Docker container. The mapping between MQTT, IEEE2030.5, and Modbus messages is defined in a configuration file. Here, after defining the connections (e.g., MQTT broker, NATS server, and Modbus slaves), the transformations can be used to describe from which topics to listen, to map incoming/outgoing messages to a specified format and structure. VII. IMPLEMENTATION OF THE HIL SET-UP The HIL set-up proposed in this work consists of an electrical and thermal co-simulation, a set of controllers implemented in the laboratory and the measurements and Riello BESS of the FZJ campus. The co-simulation, which is used to replicate the real-time behavior of the electrical grid and LTDH network, has been presented in [2] and comprehends a real-time simulator from OPAL-RT for the electrical system and the python-based simulator MPA k t t k+1 k+1 DVC Riello BESS Heating Measurements Electrical Measurements LPC LPC Fig. 2. HIL co-simulation and controllers set-up with field communication. HeatNetSim for the thermal system [16]. The co-simulation infrastructure belongs to the HIL laboratory developed at the Institute of Climate and Energy Research: Energy Systems Engineering (ICE-1) of FZJ. In this work, HeatNetSim is used to calculate the electrical active power consumption of the HPs, with this calculation being based on the heating demands and the outcome of the LTDH network simulation. The resulting HPs electrical consumption is then provided via communication to OPAL-RT, where it is added to the total electrical load. The measurements obtained from the FZJ campus and collected with the ICT platform described in Section V are integrated into the simulators by means of the HIL laboratory communication infrastructure. The heating demand real-time data collected from the heating sensors are provided to HeatNetSim, which performs the thermal simulation of the LTDH network, thus providing the electrical consumption of the simulated HPs. The electrical load measurements are supplied in real-time to simulate both the LV and MV part of the electrical grid. At each MPA update iteration, t= 1 minute, the electrical grid simulator provides the simulated data of the LV grid to the MPA, whereas at each control iteration, k= 5 seconds, the simulated voltage measurements of the MV are provided to the DVC. As regards the implementation of controllers, the MPA, running in a virtual machine hosted within the HIL laboratory’s computing infrastructure, receives the updated simulated measurements, depicted with light blue lines in Fig. 2, from the co-simulation every t. Every time new measurements are received, the MPA performs the optimization using CVXPY, an open-source Python package for convex optimization [17]. The DVC is implemented using Docker containers running on RPi units, following the approach described in [18]. At every control iteration k, each controller ireceives from the electrical grid simulator the voltage simulated measurements (i.e., we consider one RPi/controller at each MV node), light-blue lines in Fig. 2. Moreover, at every MPA update iteration teach controller ireceives from the MPA the aggregated flexibility margins for each MV node i, green line in Fig. 2. Once the DVC has calculated the set points, the disaggregated control set-points are sent to the electrical grid simulator. In the case of the controller linked to the node where the Riello BESS is installed, the set-point calculated by the RPi is sent to the Riello BESS via the LPC. All the DVC control signals are depicted in orange lines in Fig. 2. VIII. GRID UNDER TEST AND RESULTS In this work, a subset of the FZJ electrical and LTDH network has been examined. The electrical grid consists of two MV distribution feeders, each starting from a dedicated 35 kV to 10 kV transformer, as illustrated in Fig. 3. As mentioned above, each node of the feeder is connected to the campus buildings and to the Riello BESS by means of a MV/LV transformer (10 kV to 0.4kV). The rated power of BESSs and PVs are described in Table I, with all the BESSs efficiency set to 0.9. All the BESSs and PVs shown in Fig. 3 are simulated in OPAL-RT, whereas the Riello BESS located at node 16 is integrated in the simulation environment by means of the LPC. In this test, the maximum available power for the Riello BESS is reduced to 100 kW. This choice reflects two main motivations: (i) to impose a conservative limit on the amount of power injected into the actual Riello a b a b 567 1234 8910 11 12 13 14 15 16 17 Fig. 3. Distribution network and LTDH network under test. TABLE I NOMINAL PARAMETERS OF BESS AND PV SYSTEM. Node 1 2a 2b 3a 5 7 10 11 14 15 16 Pmax i,PV [kWp] 100 0 0 60 100 10 15 80 0 0 0 Prated i,BESS [kW] 200 100 100 200 200 0 80 200 150 200 1500 Erated i,BESS [kWh] 320 150 150 320 320 0 100 320 250 320 500 campus grid; (ii) to demonstrate the activation of the nearby simulated BESS when the Riello BESS power limit is reached, thus demonstrating the operation of the DVC control via communication. In particular, point (i) also reflects the fact that the Riello BESS is connected to the actual campus grid. This implies that, as will be shown subsequently, when the load is modified in the simulation to generate a voltage problem, the battery will act to resolve a problem that is unobserved in the actual grid. Therefore, limiting the power of the Riello BESS allows the mitigation of potential influences on the actual grid, which is in a stable condition, unlike the simulated one. As described in Section V, the load profiles are derived from the real-time measurements obtained by aggregating the power for each phase. The LTDH network, depicted in red in Fig. 3, is composed of seven HPs, six connected to the upper feeder, at node 1, 2b, 3a, 3b, 5, 6, and one connected to the lower feeder, at node 17. Each node iof the DVC, implemented in a RPi, controls one node of the MV distribution grid, including those where no flexible DERs are installed. In this particular case, e.g. for node 17, the associated DVC controller is configured with flexibility margins set to zero. This indicates that these controllers do not perform any control calculations; however, they are capable of exchanging Lagrange multipliers. Consequently, this enables the adjacent controllable nodes to become aware of any voltage violations occurring within the set of neighboring nodes. Given the location of the Riello BESS, the results focus on the bottom feeder, with a particular emphasis on the nodes 14 to 17, as these represent the more sensitive points of the grid. In particular, node 17 combines the electrical demands of the building and of the HP, thereby resulting in a larger overall load at the end of the feeder. The result of the test is presented in Fig. 4, where on the top plot the load profiles of electrical demands for nodes 14 to 17 (with blue lines) and the electrical demand of the HP at node 17 (with red line) are shown. As described in Section VII, the real-time measurements obtained from the field are incorporated into the co-simulation as input data. The plot underscores the variation in load consumption behavior across different buildings, with some exhibiting relatively static patterns and others exhibiting significant fluctuations. However, to induce a condition of voltage violation and thus activating the DVC, the measurement data for node 17 have been modified in the simulation environment during the test. Therefore, after 6minutes and 18 seconds, the measurement of the thermal demand obtained from the field for the building at node 17 is increased by 160 kW, which consequently increases the electrical demand of the HP calculated with HeatNetSim. Consequently, the phase-to-ground voltage of the selected nodes, depicted in the centre plot of Fig. 4, decreases in accordance with the dynamic behavior of the HP load, ultimately attaining the threshold voltage, depicted with a yellow line, of 5.48 kV at 8minutes and 13 seconds. Thus, the DVC controller at node 17, measuring the undervoltage, begins to send the Lagrangian multiplier to the node 16 Fig. 4. Top: active power load consumption. Centre: voltage profiles. Bottom: active power of the BESSs. and, as a result, the controller at node 16 begins to calculate control set-points to solve the voltage violation. The resulting set-point trajectory for node 16, described with orange line in the bottom plot of Fig. 4, is then applied to the Riello BESS that injects the requested power to the FZJ electrical grid as depicted with the purple line. At 12 minutes and 42 seconds, the 163 kW increase of the electrical load demand at node 17 produces an additional voltage drop at node 17 that, as a consequence, requires an additional power injection at node 16.At13 minutes and 40 seconds the Riello BESS reaches its limit, thus activating the transmission of the Lagrangian multiplier to node 15. Hence, the simulated battery at node 15 begins to inject active power to the grid, as depicted with the green line in Fig. 4. Consequently, the injection of active power at node 15 and 16 causes the voltage at node 17 to increase until it returns to the acceptable range at 17 minutes and 38 seconds. The centre plot also shows the comparison of the controlled voltage at node 17 with the voltage, depicted with the dashed line, that would result if not controlled would be applied, thus highlighting the positive impact of the DVC. As it can be noticed in the zoomed-in plot, the Riello BESS is able to track the DVC control set-point with a good accuracy throughout the test. The delay between the set-point and the actual power injection results from the transmission time of the data from the HIL laboratory to the field. IX. CONCLUSIONS This paper presents the testing of a coordinated voltage control for distribution grid in a HIL set-up, with the integration of real-time measurements collected from the FZJ campus and provided the electrical and thermal co-simulation. The developed control strategy employs the output of the co-simulation to calculate the aggregated flexibility trajectory at the LV level and the control set-points for the MV level, which are then applied to the simulated DERs, with the exception of one that is applied to a large BESS installed in the field. The obtained results show the realistic behavior of the proposed set-up, highlighting the variation of the load profiles and the consequent variation of the simulated voltage. The results clearly show the ability of the coordinated voltage control to solve the voltage violation even at the node where no available flexibility, by leveraging the communication between the neighboring controllers. The control set-point is also applied to a large BESS installed in the campus, proving the effectiveness of the LPC in converting and transmitting the signal from the controllers to the field. The results obtained in this work indicate possible future investigations: (i) the comparison of different control strategies using the same HIL set-up; (ii) the accurate analysis of the communication network and the integration of solutions for reducing the time delay; (iii) increase the number of DERs of the campus to which apply the control set-points. REFERENCES [1] E. De Din, M. Pau, F. Ponci, and A. Monti, “A coordinated voltage control for overvoltage mitigation in lv distribution grids,” Energies, vol. 13, no. 8, 2020. [2] D. Liu, D. Carta, A. Xhonneux, D. M¨ uller, and A. Benigni, “Short-term control of heat pumps to support power grid operation,” IEEE Open Journal of the Industrial Electronics Society, vol. 5, pp. 1221–1238, 2024. [3] Y. Wang, T. Zhao, C. Ju, Y. Xu, and P. Wang, “Two-level distributed volt/var control using aggregated pv inverters in distribution networks,” IEEE Transactions on Power Delivery, vol. 35, no. 4, pp. 1844–1855, 2020. [4] B. Wei, F. Zobiri, and G. Deconinck, “Peer to peer flexibility trading in the voltage control of low voltage distribution network,” IEEE Transactions on Power Systems, vol. 37, no. 4, pp. 2821–2832, 2022. [5] E. Widl, A. Sporr, R.-R. Schmidt, T. Natiesta, and N. Marx, “Digitalenergytestbed: An open testbed prototype for integrated energy systems,” in 2023 Open Source Modelling and Simulation of Energy Systems (OSMSES), 2023, pp. 1–7. [6] Z. Yuan, A. Zecchino, R. Cherkaoui, and M. Paolone, “Real-time control of battery energy storage systems to provide ancillary services considering voltage-dependent capability of dc-ac converters,” IEEE Transactions on Smart Grid, vol. 12, no. 5, pp. 4164–4175, 2021. [7] F. Arra˜ no-Vargas, S. Jiang, B. Bennett, and G. Konstantinou, “Mitigation of power system oscillations in weak grids with battery energy storage systems: A real-world case study,” Energy, vol. 283, p. 128648, 2023. [8] L. Westphal, M. Schr¨ oder, D. Carta, A. Xhonneux, A. Benigni, and D. M¨ uller, “Development and Application of a FIWARE-based ICT-Platform for Multi-Energy Systems on Building and District Level,” in 2024 Open Source Modelling and Simulation of Energy Systems (OSMSES), 2024, pp. 1–6. [9] X. Chen, E. Dall’Anese, C. Zhao, and N. Li, “Aggregate Power Flexibility in Unbalanced Distribution Systems,” IEEE Transactions on Smart Grid, vol. 11, no. 1, pp. 258–269, 2020. [10] S. Bolognani, R. Carli, G. Cavraro, and S. Zampieri, “On the Need for Communication for Voltage Regulation of Power Distribution Grids,” IEEE Transactions on Control of Network Systems, vol. 6, no. 3, pp. 1111–1123, 2019. [11] PQI-DA smart. [Online]. Available: https://www.a-eberle.de/en/products/power-quality-analyzer-pqi-da-smart/ [12] IEC International Electrotechnical Comission, “Electromagnetic compatibility (EMC) - Part 4-30: Testing and measurement techniques - Power quality measurement methods.” 2015. [13] InterSTORE. [Online]. Available: https://interstore-project.eu/ [14] Legacy Systems Protocol Converter for IEEE2030.5. [Online]. Available: https://https://github.com/Horizont-Europe-Interstore/Legacy-Protocol-Converter [15] Modbus. [Online]. Available: https://modbus.org [16] S. Dibos, T. Pesch, and A. Benigni, “Heatnetsim: An open-source simulation tool for heating and cooling networks suitable for future energy systems,” Energy, vol. 312, p. 133588, 2024. [17] CVXPY. [Online]. Available: https://www.cvxpy.org [18] E. De Din, M. Ginocchi, F. Ponci, A. Monti, and A. Benigni, “Global sensitivity analysis based method for distributed voltage control performance study,” in 2024 International Conference on Smart Energy Systems and Technologies (SEST), 2024, pp. 1–6.