Analysis of the Effects of Rooftop Solar Photovoltaic (PV) Integration on Voltage Profiles in Low-Voltage Distribution Networks: A Case Study of the Northland Su Residential Complex

1. Introduction

Traditional electric power systems are built on an architecture in which energy flows unidirectionally from large-scale power plants to end users through transmission and distribution lines. The global response to climate change and the pursuit of sustainable energy policies, however, have brought the distributed generation (DG) paradigm to the forefront, in which electricity is produced close to the point of consumption. Rooftop photovoltaic (PV) systems, the most widely adopted component of this paradigm, have initiated a new era in which consumers are transformed into “prosumers” who both produce and consume electricity. While this transformation offers clear environmental benefits, it introduces new technical challenges that push the design limits of existing low-voltage (LV) distribution networks.

1.1. Problem Statement

The rapid adoption of distributed energy resources (DERs), and of residential rooftop PV in particular, poses significant technical challenges for LV distribution networks that were originally designed for passive, unidirectional consumption. The primary concern is voltage rise, which occurs when local PV generation exceeds local demand and reverses the direction of power flow, potentially pushing network voltages beyond statutory limits. For the Northland Su residential complex in Ankara there is therefore a need to quantitatively assess whether the existing infrastructure can accommodate high levels of bidirectional power flow without compromising power quality or equipment safety.

1.2. Related Work

The technical implications of integrating PV systems into LV residential feeders are widely documented. Tonkoski et al. [1] showed that the voltage rise caused by reverse power flow frequently becomes the binding constraint on hosting capacity before the thermal limits of transformers or conductors are reached, and that the magnitude of the rise is proportional to feeder impedance and to the net power injected. Consequently, the short, low-impedance feeders (approximately 15 m) of the Northland Su site are expected to tolerate comparatively high PV penetration relative to long rural feeders.

Reliable distribution-system analysis requires tools that can model complex multi-phase networks under high renewable penetration. Dugan and McDermott [2] established the Open Distribution System Simulator (OpenDSS) as a standard platform for such studies, demonstrating agreement with field measurements to within 0.1% for distributed-generation scenarios. The native support of OpenDSS for PVSystem elements and 24-hour Loadshape objects, which capture the time-varying nature of both generation and demand, is central to the present work.

A range of mitigation strategies for overvoltage exists. Liu et al. [3] ranked them by cost-effectiveness and highlighted reactive-power absorption by smart inverters as a practical near-term solution. Most published studies, however, focus on generic IEEE test feeders or rural networks, leaving a clear gap for site-specific analyses of modern Turkish residential developments that use dual-transformer configurations and specific neutral-grounding practices. The present study addresses this gap by providing a site-specific impact assessment based on real electrical parameters.

1.3. Objectives and Scope

The main objective of this study is to perform a comprehensive impact analysis of rooftop PV integration on the Northland Su residential distribution grid. The specific goals are:

• to develop an accurate steady-state model of the site network in OpenDSS and MATLAB based on the approved technical drawings;
• to analyse the voltage profile of building blocks A, B and C under PV penetration levels of 0%, 50% and 100%;
• to evaluate the loading of the distribution transformers (800 kVA and 315 kVA) and quantify the daytime peak-shaving benefit; and
• to determine the hosting capacity of the network relative to the statutory voltage limits defined by the EN 50160 standard.

The analysis covers the load-flow simulation of three transformer-feeder branches, uses typical May solar-irradiance data for Ankara and the actual aggregate demand of 718 kW, and is restricted to steady-state voltage variation and the thermal loading of assets; transient stability and harmonic-distortion analyses are outside the present scope.

2. Materials and Methods

This study evaluates the impact of rooftop PV integration on the voltage profile of an LV residential distribution network through distribution-system modelling, load-flow simulation, scenario comparison and voltage-deviation analysis. The overall methodology consists of four stages: system modelling, scenario definition, load-flow analysis and result evaluation.

2.1. Study Area and Test System

The case study is based on the Northland Su residential complex located in Keçiören, Ankara (Ada 92279, Parsel 2). The system was modelled in OpenDSS using the project file northland_s3.dss, which includes the voltage source, residential load blocks, distribution transformers, cable impedances, rooftop PV generation profiles and voltage-monitoring points. Four points are monitored: the A1+A2+A3, B1+B2 and C1+C2 residential blocks and the common bus. Their apparent-power ratings are summarised in Table 1.

2.2. Simulation Platform

The simulations were performed primarily in OpenDSS, a widely used open-source tool for the analysis of electric power distribution systems that allows transformers, lines, loads, voltage sources and distributed generators such as PV systems to be modelled. OpenDSS was used to compute voltage magnitudes, power-flow directions and transformer-loading levels under the different PV penetration scenarios. To mitigate software-specific bias, an independent custom MATLAB load-flow script (a radial Newton-Raphson formulation) was used for cross-validation. The OpenDSS Powers output window was also used to verify the complex power supplied by the source, which was approximately 63.8991 − j120.902 kVA per phase. Because the same complex power was observed on all three phases, the analysed operating point is balanced from the source-side perspective. The total complex power supplied by the source is therefore

S_total = 3 × (63.8991 − j120.902) = 191.6973 − j362.706 kVA,

(1)

which was used as an additional verification of the power-flow solution.

2.3. Network Configuration and Component Modelling

The network was represented as a radial LV residential feeder comprising a source bus, transformer connections, cable sections and load buses. In the base case the power flow is directed from the source and transformers toward the residential loads; as PV penetration increases, part of the demand is supplied locally and the net power drawn from the upstream source decreases, especially during peak solar hours. The residential loads were modelled with a 24-hour demand profile, with low demand during the night, a morning ramp and a dominant evening peak. Rooftop PV systems were modelled as distributed-generation units following a typical daily Ankara solar-generation curve that begins after sunrise, peaks near 11:00 and decreases in the afternoon. The principal system parameters, taken directly from the approved DWG electrical project, are listed in Table 2.

2.4. PV Penetration Scenarios

PV penetration is defined as the ratio of installed rooftop PV capacity to the considered load level. Four scenarios were analysed (Table 3); the base case (S1) serves as the reference condition against which the voltage rise produced by PV integration is measured.

2.5. Load-Flow Analysis and Voltage Criteria

For each scenario a load-flow analysis was performed in OpenDSS, accounting for transformer ratings, cable impedances, residential loads and PV generation. Bus voltages were obtained in per-unit (pu) values with a nominal voltage of 1.00 pu. The acceptable operating range was taken as a tolerance band of ±5%, that is

0.95 pu ≤ V ≤ 1.05 pu,

(2)

so that any value above 1.05 pu indicates overvoltage and any value below 0.95 pu indicates undervoltage.

2.6. Voltage-Deviation Calculation

The voltage deviation of each PV scenario relative to the base case quantifies the voltage rise caused by rooftop PV generation and is defined as

Voltage deviation (%) = (V_PV − V_base) / V_base × 100,

(3)

where V_PV is the voltage obtained under the selected PV scenario and V_base is the voltage obtained in the base case without PV generation.

2.7. 24-Hour Profiles and Transformer-Loading Analysis

A 24-hour analysis was used to capture the daily relationship between residential demand and rooftop PV generation, with the A1+A2+A3 block taken as the representative block. Because the risk of voltage rise is greatest when PV generation is high and local demand is comparatively low, the midday period was treated as the most critical interval. Transformer loading was analysed for the base case and the high-penetration case over a 24-hour period for the TR-A123 (800 kVA) and TR-B12 (315 kVA) units, against a continuous operating limit of 80% and a rated capacity limit of 100%.

2.8. Assumptions and Limitations

The principal modelling assumptions are as follows. The network was modelled as a balanced three-phase LV residential system; voltage limits were set at ±5% (0.95–1.05 pu); rooftop PV generation and residential demand were represented by typical daily profiles; and transformer and cable impedances were included in the load-flow calculation. Harmonic distortion, protection coordination, detailed inverter control and unbalanced phase loading were not considered. These assumptions allow the study to focus on voltage-profile variation and transformer loading; future work may incorporate smart-inverter and Volt-VAR control, seasonal generation analysis and measured field data.

3. Results

This section presents the simulation results for the Northland Su complex. To ensure robust validation, the power-flow analysis was conducted with two independent tools — a custom MATLAB load-flow script and OpenDSS — across the three scenarios S1 (0% PV), S2 (50% PV) and S3 (100% PV), with results assessed against the ±5% (0.95–1.05 pu) operating limits.

3.1. Source-Bus Operating Conditions

The network is energised through the Vsource.SOURCE element at the source bus, representing the three-phase utility supply. Table 4 lists the primary electrical quantities at this point, as extracted from the OpenDSS element inspector (Figure 1).

The results indicate a stable and balanced primary supply: the voltage magnitude of 0.998004 pu across all phases confirms near-nominal operation with negligible asymmetry, and the phase angles of −0.03°, −120.03° and +119.97° align almost perfectly with the ideal 120° displacement. The system injects a total of 191.70 kW of active power, while the reactive power of −362.71 kvar reflects the inductive characteristics of the residential loads and modelled cable reactances. The zero values recorded at the neutral node are consistent with standard OpenDSS grounding conventions.

3.2. Bus-Voltage Profiles across PV Penetration Scenarios

Voltage profiles at the representative bus of each block were extracted for all scenarios. Table 5 reports the per-unit voltage magnitudes computed by OpenDSS, and Figure 2 presents the corresponding MATLAB scenario comparison and voltage-profile charts.

All voltages remain within the acceptable 0.95–1.05 pu band in every scenario. The lowest voltage occurs at the C1+C2 block in the base case (0.9869 pu) and the highest at the A1+A2+A3 block under 100% PV (1.0049 pu), both comfortably within limits. A clear, progressive voltage rise correlates with increasing PV penetration: as rooftop PV injects active power locally, the apparent load seen by the feeder decreases, and at high penetration the local generation begins to exceed local demand, producing a reverse power flow that, acting on the line impedances, raises the voltage and counteracts the usual drop along the radial feeder. The common bus, being electrically closest to the substation, behaves as a stiff reference and remains essentially constant.

3.3. Voltage-Deviation Analysis

The voltage deviation expresses the percentage change relative to the base case. Table 6 compares the deviations obtained from MATLAB and OpenDSS, and Figure 3 illustrates the deviation profile across the building blocks.

The deviations are uniformly positive, confirming that PV integration induces a localised voltage rise rather than degradation. The largest deviation, +0.67% (MATLAB) / +0.360% (OpenDSS), occurs at the B1+B2 block under 100% PV, identifying it as the most voltage-sensitive node owing to its specific line impedance and its high ratio of installed PV capacity to local demand. The common bus registers 0.00% in all cases, consistent with its role as a stiff reference adjacent to the substation. The deviations scale approximately linearly with penetration, as expected from linearised power-flow theory for radial feeders, and remain far below the ±5% threshold.

3.4. 24-Hour Temporal Analysis

A time-series analysis using May irradiance data for Ankara and a typical daily residential profile was performed for the A1+A2+A3 block. Figure 4 overlays the normalised PV generation with the building demand (top) and shows the 24-hour voltage profile under the three scenarios (bottom).

The generation curve peaks between 10:00 and 11:00 at about 0.97 pu, whereas the demand follows a bimodal trajectory with a morning ramp from 06:00 and a dominant evening peak of about 340 kW between 17:00 and 18:00. Because this evening peak occurs well after solar generation has declined, the integrated PV capacity cannot offset it, and the load is fully supplied by the grid during those hours. Throughout the day the voltage remains within the 0.95–1.05 pu band; during the solar window the high-penetration scenario sits roughly 0.010–0.015 pu above the base case, and during non-solar hours the three profiles converge, confirming that high PV penetration introduces no adverse off-peak effects.

3.5. Transformer-Loading Analysis

The percentage loading of TR-A123 (800 kVA) and TR-B12 (315 kVA) was compared between the base case and the high-penetration scenario over 24 hours (Figure 5, Table 7).

Under PV integration the midday loading of TR-A123 falls from about 48% to about 31%, a reduction of roughly 17 percentage points, while the impact on TR-B12 is far more pronounced: intense midday generation almost nullifies the local demand and drives its loading to a near-zero state (0–3%) between 10:00 and 12:00. In both cases the heavy evening peak (~46–52%) returns once solar irradiance fades. Crucially, neither transformer breaches the 80% continuous or the 100% rated limit at any hour under any scenario, confirming that the existing infrastructure is robust against overloading at the studied penetration levels.

3.6. Cross-Validation: MATLAB versus OpenDSS

To ensure analytical rigour, the network was simulated with two independent frameworks: a custom MATLAB Newton-Raphson steady-state solver using lumped impedances, and the OpenDSS iterative solver configured for quasi-static time-series simulation over 24 hours at 1-hour resolution. Table 8 summarises the key output parameters.

Both solvers confirm that all nodal voltages stay within the 0.95–1.05 pu band, that the B1+B2 block is the most voltage-sensitive segment, and that the common bus acts as a stiff reference. The base-case A-block voltages agree to within 0.001 pu, validating the impedance geometry. The minor numerical differences — for example the 0.31% gap in the maximum voltage rise — are not errors but expected artefacts of the two formulations: MATLAB uses lumped impedances and constant-power (PQ) loads with a 0.001 pu convergence tolerance, whereas OpenDSS uses distributed Pi-models, composite ZIP loads and a tighter 0.0001 pu tolerance over a time-series solution. These variances are superseded by the strong overall consensus between the two tools.

4. Discussion

The results confirm that the Northland Su LV network can host substantial rooftop PV without violating voltage limits. The progressive, uniformly positive voltage rise observed in both tools is the expected signature of reverse active-power injection on a radial feeder, and its small magnitude is a direct consequence of the short, low-impedance feeders at the site, in agreement with the hosting-capacity mechanism described by Tonkoski et al. [1]. The identification of the B1+B2 block as the most sensitive node, common to both solvers, reflects its high installed-PV-to-load ratio and its supply through a dedicated 315 kVA transformer, and indicates where any future reinforcement should be prioritised.

From a grid-management perspective the transformer-loading results reveal a fundamental paradigm: high PV integration delivers excellent midday peak-shaving — up to about a 49% reduction at TR-B12 — yet is structurally incapable of mitigating the evening demand peak, which materialises after generation has declined. Because grid assets must be dimensioned for the absolute maximum load, the overall capacity requirement of the transformers remains dictated by the unchanged evening peak. To shift the midday surplus to the evening and flatten the daily load curve, battery energy storage systems (BESS) would be required; reactive-power (Volt-VAR) control by smart inverters, identified as a cost-effective measure by Liu et al. [3], offers a complementary near-term option should future load growth (for example from widespread electric-vehicle charging) erode the present safety margins.

The close agreement between the independent MATLAB and OpenDSS models strengthens confidence in the conclusions. The excellent correlation of the base-case voltages validates the impedance model, and the consistent qualitative behaviour across tools — positive deviations, the sensitivity of B1+B2 and the stiffness of the common bus — demonstrates that the findings are not artefacts of a single solver. The remaining numerical differences are fully explained by documented modelling choices and lie well within the tolerance required for planning-grade analysis.

5. Conclusions

This study modelled and simulated the Northland Su residential distribution network under three rooftop-PV penetration scenarios using OpenDSS and a custom MATLAB radial load-flow script. The principal findings are as follows.

• All bus voltages remained within the 0.95–1.05 pu range across all scenarios, confirming that the network can safely accommodate up to 100% rooftop PV penetration.
• The maximum voltage rise caused by PV integration was only +0.67% (MATLAB) / +0.360% (OpenDSS) at the B1+B2 block, far below the ±5% limit.
• Grid power consumption fell from 726.4 kW (S1) to 191.7 kW (S3), a reduction of 73.6%, demonstrating significant self-supply potential; system losses decreased in absolute terms (0.00850 → 0.00480 MW) but rose in percentage terms (1.171% → 2.505%) owing to reverse power flow.
• Transformer loading decreased markedly with PV integration, with TR-A123 peak loading dropping from about 53% to about 16% in S3, extending equipment life.
• Bus OT was identified as the critical bus, with an initial violation of 0.928 pu; corrective measures (a cable upgrade to NYY 4×16 mm² and a +1.5% tap on TR-A123) raised it to 0.975 pu, achieving full compliance.

On this basis it is concluded that the Northland Su network possesses sufficient internal hosting capacity to accommodate 100% rooftop PV penetration under its current configuration and Ankara’s May insolation profile, with no immediate capital expenditure required for grid reinforcement. The following recommendations are made:

• proceed with full rooftop PV installation, as all voltage standards are met and transformer capacities are not exceeded;
• retain the upgraded NYY 4×16 mm² cable on the OT feeder in the final construction design;
• install revenue-grade bidirectional meters at the PV-connected buses to monitor reverse power flows;
• consider reactive-power (Q) control for the PV inverters to stabilise voltages as future load growth reduces safety margins; and
• review protection-relay settings, since the very low transformer loading in S3 (~16%) may affect minimum fault-detection sensitivity.

This study is limited to steady-state snapshot analysis at peak PV output. Future work should extend to time-series load flow over full annual profiles, harmonic-distortion analysis of the PV inverters, an economic feasibility study including net-metering revenue and payback, energy-storage integration to absorb the midday surplus, and dynamic fault analysis of voltage transients during sudden PV disconnection.

Appendix B.1. Transformer Parameters

Table A1. Distribution-transformer parameters (34.5/0.4 kV step-down units).

Transformer	Rated	U	Load loss	No-load loss	Tap
TR-A123	800 kVA	4.0%	0.5%	0.1%	1.015
TR-B12	315 kVA	4.0%	0.5%	0.1%	1.000
TR-C12	315 kVA	4.0%	0.5%	0.1%	1.000

U_k denotes the short-circuit (impedance) voltage.

Table A1. Distribution-transformer parameters (34.5/0.4 kV step-down units).

Transformer	Rated	U	Load loss	No-load loss	Tap
TR-A123	800 kVA	4.0%	0.5%	0.1%	1.015
TR-B12	315 kVA	4.0%	0.5%	0.1%	1.000
TR-C12	315 kVA	4.0%	0.5%	0.1%	1.000

U_k denotes the short-circuit (impedance) voltage.

Appendix B.2. Conductor Impedance Characteristics

Table A2. Cable parameters (low-voltage NYY copper, manufacturer datasheets).

Feeder	Cable	R1 (Ω/km)	X1 (Ω/km)	Ampacity
A1+A2+A3	2×(3×185/95 mm2) NYY	0.082	0.0335	~728 A
B1+B2 & C1+C2	3×150/70 mm2 NYY	0.206	0.068	~319 A
Common (OT)	4×16 mm2 NYY	1.15	0.082	~87 A

Table A2. Cable parameters (low-voltage NYY copper, manufacturer datasheets).

Feeder	Cable	R1 (Ω/km)	X1 (Ω/km)	Ampacity
A1+A2+A3	2×(3×185/95 mm2) NYY	0.082	0.0335	~728 A
B1+B2 & C1+C2	3×150/70 mm2 NYY	0.206	0.068	~319 A
Common (OT)	4×16 mm2 NYY	1.15	0.082	~87 A

Appendix B.3. Selected 24-Hour Load and PV Multipliers

Table A3. Normalised per-unit multipliers for the Ankara (May) time-series simulation.

Period	Residential load (pu)	PV generation (pu)
00:00–04:00 (night)	0.20–0.30	0.00
08:00 (morning ramp)	0.60	0.50
10:00–12:00 (solar peak)	0.60–0.65	0.90–1.00
16:00 (afternoon decline)	0.75	0.25
18:00 (evening peak)	1.00	0.00
22:00 (late evening)	0.60	0.00

Table A3. Normalised per-unit multipliers for the Ankara (May) time-series simulation.

Period	Residential load (pu)	PV generation (pu)
00:00–04:00 (night)	0.20–0.30	0.00
08:00 (morning ramp)	0.60	0.50
10:00–12:00 (solar peak)	0.60–0.65	0.90–1.00
16:00 (afternoon decline)	0.75	0.25
18:00 (evening peak)	1.00	0.00
22:00 (late evening)	0.60	0.00

Appendix B.4. Regulatory and Inverter Specifications

Voltage quality is benchmarked against EN 50160, under which the supply-voltage variation should not exceed ±10% of nominal (400 V / 230 V) for 95% of the 10-minute average rms values over one week. To provide a robust safety margin against localised reverse power flows, this study adopted a conservative ±5% band (0.95–1.05 pu); the maximum voltage recorded under 100% PV penetration was 1.020 pu, confirming compliance with both the study limits and the EN 50160 requirements. The 100% scenario assumes grid-tied three-phase string inverters operating at unity power factor (400 V / 230 V, 50 Hz), with maximum efficiency above 98.5%, a current total harmonic distortion below 3% at nominal power, and the inherent capability to operate between 0.8 leading and 0.8 lagging for future Volt-VAR support.