TOWARDS MORE COST-EFFECTIVE PV CONNECTION REQUEST ASSESSMENTS VIA TIME-SERIES BASED GRID SIMULATION AND ANALYSIS

In this paper, we present methodological improvements to one of the most often performed distribution grid analysis tasks in Europe today – the assessment of PV connection requests and the analysis of whether or not a most-often costly local grid reinforcement becomes necessary. For this we model a typical European lowvoltage distribution grid with the request for a sizeable PV unit at the fringes of the grid topology as it is often the case for PV units on farm buildings in rural areas. Such a setup is a demanding constellation for grid operators, typically requiring costly grid upgrades.


INTRODUCTION
The state-of-the-art methods for assessing PV connection request rules are static analysis approaches such as the DACHCZ guidelines [1].These guidelines define that the relative voltage rise (∆V) incurred by the totality of all generating plants at their nominal or installed power rating (P installed ) compared to zero feed-in (P=0) must not exceed 3% voltage rise at any point of common coupling in low-voltage (LV) networks, and only up to 2% in medium-voltage (MV) networks.This voltage rise evaluation scheme is illustrated by Figure 1 and is typically calculated under the assumption of having no load demand in the given distribution grid section.Obviously, such an approach will be rather conservative.The implicit conservativeness of this grid planning approach is particularly obvious when considering that the European Standard EN 50160 [2] defines the variation of voltages in distribution grids can under normal operating conditions vary up to ±10% of the nominal operating voltage in 95% of the 10-minuteintervals of the measured root mean square value within the interval of one week (for 5% of 10-minute-inverals, voltage can drop down to -15%).The typical result is thus that larger PV units cannot be connected to the distribution grid without prior grid reinforcement and its associated costs.In an exemplary case study, we show that a time-series based photovoltaics (PV) connection request assessment leads to a more detailed analysis, by calculating the bus voltage as well as line loading time-series for the whole distribution grid topology based either on the inclusion of actual grid measurements, i.e. load demand and other local power generation profiles, and/or synthetic but plausible load demand and generation profiles.This approach provides better decision-support for distribution grid planning by analysing how close grid operation is to actual voltage and line loading limits in the worst case (i.e.maximum voltage and line loading) as well as providing the necessary data for a statistical analysis over longer time-frames, i.e. one full year.Such previously unavailable insights can avoid costly grid upgrades that are otherwise deemed necessary.

SIMULATION PLATFORM DPG.SIM
In order to overcome these limitations, the SmartGrid time-series based simulation platform DPG.sim (Distributed Prosumer and Grid Simulation) is developed by ETH Zurich spin-off Adaptricity [3].The detailed simulations and subsequent grid analytics provide valuable qualitative as well as quantitative decisionsupport for all aspects of distribution grid operation and planning.On the grid operation side this includes notably the design and performance analysis of active network management operation strategies.On the grid planning side this includes the integration of SmartGrid elements into distribution grid planning procedures.The main advantages of the DPG.sim simulation platform are threefold: First, its possibility to realistically model and simulate the operation of active distribution grids as well as the temporal evolution of generation, load, storage states, and their operational control algorithms down to the level of individual households and household units.Its unique feature in this respect is its versatile Prosumer modelling approach [5], which allows capturing all relevant modelling details and operational constraints of controllable loads, distributed generation, and storage as well as SmartMeter communication infrastructure (Fig. 3).Second, the ability to perform large-scale time-series simulations and operational (big) data analytics based on heterogeneous sets of grid data as well as end-consumer data sets by tapping into scalable cloud-based computation and data storage resources.
Third, data visualization and statistics functionality that provides detailed insights into electricity grid operation and enable robust decision support.

CASE STUDY
In the following, a case study of a typical PV connection request for an exemplary European low-voltage distribution grid is presented in order to illustrate the potential advantages of time-series based grid simulation schemes for distribution grid planning tasks.

Benchmark Low-Voltage Grid
All grid simulation and analysis are based on a modified version of the urban low-voltage distribution grid defined by the CIGRE Task Force C6.04.02 [5].
The distribution grid (Figure 32) consists of a 20 kV medium-voltage grid topology that supplies low-voltage grid feeders via three separate 400 V transformer stations.
The resulting 42-bus distribution grid topology consists of typical residential, commercial and industrial load types [5].Typical full-year time-series of load demand with 15 min.sampling time have been modelled and linked to the grid topology.A realistic PV power production profile has also been created.
The grid topology and the associated power consumption and load demand profiles are used in the following both for static grid simulations (snapshots) as well as for timeseries based grid simulations and statistical analyses.

Voltage Rise Calculation & DG Hosting Capacity
The classical PV connection request assessment according to the DACHCZ guidelines [1], allows for static voltage rise of up to 3.0%, or 0.03 per-unit (pu) at low-voltage grid connection points.Depending on the specific bus, i.e. its location inside the low-voltage grid topology, a minimum of 30 kW (bus 17) and a maximum of up to 360 kW (bus 2) of Distributed Generation (DG) can be hosted by the low-voltage grid (Figure 4).Significantly more DG power can be hosted on the medium-voltage side of the distribution grid, e.g. more than 6.0 MW on each MV bus (buses 1, 19 and 40).
The buses with the lowest DG hosting capacities are typically those buses with the largest (electrical) distance from the substation, respectively the LV transformers, since those buses (buses 15, 16, 17, 18 and buses 37-39) are weaker than others.Please note that the following assumptions are being used for the calculations: Bus 0 is the slack bus and is disregarded, a typical delta voltage rise constraint of 2% (MV) and 3% (LV) is defined as the constraint for limiting DG hosting capacity.In order to make the static voltage rise assessment less conservative, a minimum load situation, i.e. 1 st of May (3am), is considered instead of a more conservative but rather unrealistic no-load assumption.

Assessment of PV Connection Requests
We consider a connection request for a sizable PV unit of about 70 kW p , corresponding to a roof-top area of about 400m 2 for instance on top of a warehouse, at the weak bus 18 (Figure 32).The classical PV connection request assessment would allow a maximum of only 33 kW p (no-load assumption).
The planned PV unit with 70 kW p would produce a higher static voltage rise of up to 6.45%.A grid connection according to DACHCZ guidelines would require a conventional grid upgrade, i.e. a low-voltage cable with larger cross-section, potentially costing up to EUR 50'000.
Since the typical installation costs for a 70 kW p PV unit are currently estimated on average at about EUR 1350 per kW p (early 2017) in Germany [6], the total PV unit investment costs are about EUR 94'500.Thus, from a national economic perspective, the additional costs for proper distribution grid reinforcement are rather significant, i.e. more than 50% on top of the PV plant investment cost.
There is thus a strong motivation for applying less conservative yet robust PV connection request assessment methodologies that are on average more costeffective than the currently used DACHCZ guidelines.

Time-Series based Grid Simulation and Analysis
In the following we analyse how a time-series based PV connection request assessment according to the European norm EN50160 [2] compares to the previous static voltage rise calculation.Contrary to the previous analysis, we now incorporate the available load timeseries data for all LV and MV buses of the urban LV benchmark distribution grid.Such data could simply be obtained via existing SmartMeter measurements or via various available load profile generators for residential and commercial customer profiles or a mix of the above.Indeed would a time-series based grid simulation clearly indicate that a large PV unit with 70 kW p would induce a significant maximum voltage rise of 1.07 pu (bus 18) as well as a maximum line loading of up to 82% (line 10-18) for a typical summer week (Figure 5).The simulated voltage bands would thus respect the requirements of the EN50160 norm, which allows for a maximum overvoltage of up to 1.10 pu (10% over-voltage).
In a full-year simulation the maximum voltage rise may go up to 1.076 pu and the maximum line loading up to 93.5% (Figure 6, top).Throughout the full-year the loading of the critical line 10-18 is only very rarely above 80% of the maximum allowable load, i.e. 47 hours or 0.5% of the time (Figure 6, bottom).With these quantitative results based on time-series grid simulations, the responsible distribution grid operator can thus approve the pending PV connection request.This cost-saving finding is robustly based on the clear indication that voltage band norms as well as line loading limits would be respected according to the full-year simulation of the reference year, i.e. consisting of altogether 35'040 steps with 15 min.sampling time.

CONCLUSION
The results of this typical PV connection request assessment, one of the most commonly performed grid analysis task nowadays, illustrates that time-series based grid simulations are able to provide valuable insights for distribution grid operation and planning.By running detailed full-year grid simulations that incorporate all available grid measurements, as obtained by SmartMetering schemes or other sources, a larger DG hosting capacity is possible.Time-series-based grid planning can thus avoid otherwise costly conventional grid upgrades.

Figure 4 :
Figure 4: Hosting capacity of individual buses of the urban low-voltage benchmark distribution grid [5].

Figure 5 :
Figure 5: PV Connection Request -Time-Series based Grid Analysis allows much larger 70kW p PV unit.

Figure 6 :
Figure 6: PV Connection Request -Time-Series based Grid Analysis allows much larger 70kW p PV unit.