---
title: "PV simulators comparison: Electrical modeling of the AC side"
slug: "comparison-electrical-modeling-ac-side"
description: "Explore the modeling of AC side losses in PV simulators, including auxiliary loads, transformer losses, and grid connection impacts on energy delivery."
updated: 2026-08-10T14:42:56Z
published: 2026-08-10T14:42:56Z
canonical: "kb.solargis.com/comparison-electrical-modeling-ac-side"
---
> ## Documentation Index
> Fetch the complete documentation index at: https://kb.solargis.com/llms.txt
> Use this file to discover all available pages before exploring further.
# PV simulators comparison: Electrical modeling of the AC side
**In this document**
This article compares how the analyzed PV simulators model the AC side of the electrical chain: from the inverter AC output through transformers and AC cabling to the grid connection point, including auxiliary loads on the AC bus, and losses due to unavailability.
### Overview
This article picks up from [Electrical modeling of the DC side](/v1/docs/pv-simulators-comparison-electrical-modeling-of-the-dc-side), which covered the simulation from PV cell through DC cabling and inverter conversion to AC output at the inverter terminals. The AC side begins at the inverter output and ends at the grid connection point.
Five modeling stages are covered:
- [Auxiliary losses](/v1/docs/comparison-electrical-modeling-ac-side#auxiliary-losses1) – energy consumed by non-PV equipment on the plant (monitoring, trackers, lighting, HVAC)
- [AC cable losses](/v1/docs/comparison-electrical-modeling-ac-side#ac-cable-losses) – ohmic losses on conductors between inverter, transformer stages, and the grid
- [Transformer losses](/v1/docs/comparison-electrical-modeling-ac-side#transformer1) – iron and copper losses on the step-up transformers from low to medium voltage (LV to MV) and from medium to high voltage (MV to HV)
- [Grid connection](/v1/docs/comparison-electrical-modeling-ac-side#grid-connection) – active power limits, reactive power, and curtailment at the point of interconnection with the electric grid
- [System unavailability](/v1/docs/comparison-electrical-modeling-ac-side#system-unavailability1) – limitation of power export to the grid caused by internal or external factors
Each of the six simulators implements these stages with different models, default values, and application points in the simulation chain. For the comparison setup (sites, system configurations, version numbers, and bias/RMSE definitions), see [Setup and test methodology](/v1/docs/comparison-setup-and-test-methodology).
### Auxiliary losses
Auxiliary losses are the energy consumed by non-PV equipment on the plant: monitoring systems, tracker drives, security lighting, ventilation and HVAC for inverter rooms, and any other electrical loads on the AC bus. They split into daytime loads (drawn while the plant is generating) and nighttime loads (drawn from the grid when no generation is available).
#### Auxiliary losses modeling
There are three approaches:
- **Explicit day and night modeling** (Solargis Evaluate, PVsyst, SolarFarmer) distinguish daytime consumption (constant or proportional to inverter output, with an inverter power threshold below which the load does not apply) from nighttime consumption (typically constant). SolarFarmer adds plant-level or inverter-level definition.
- **Inverter own-consumption only** (SAM) captures inverter standby and night losses within the inverter model itself, but does not model auxiliary equipment separately.
- **Not modeled** (Solargis Prospect, pvlib) leave auxiliary losses to be folded into availability losses, pvlib also allows implementation in custom user code.
Auxiliary losses are typically well under 1% of annual PVOUT for well-designed plants, but rise with tracker count, plant geographic spread, and climates requiring active cooling. Nighttime consumption is significant for plants in cold climates: the grid imports are not generation losses but reduce net energy delivered. Solargis Prospect and pvlib have no built-in auxiliary loss model, and SAM captures only inverter own-consumption. In all three cases the remaining auxiliary consumption must be absorbed into the availability percentage or quantified separately.
#### Methodology comparison
| Aspect | Solargis Evaluate | Solargis Prospect | pvlib | PVsyst | SAM | SolarFarmer |
| --- | --- | --- | --- | --- | --- | --- |
| **Day model** | Constant and proportional, with inverter power threshold | Not modeled | Not modeled Can be implemented in custom code | Constant and proportional, with an inverter power threshold | Not modeled separately (inverter own-consumption in inverter model) | Plant-level (% or W) or inverter-level (fixed or variable above threshold) |
| **Night model** | Constant | Not modeled | Not modeled | Constant | Inverter own-consumption only | Constant |
***Table 1****: Auxiliary losses modeling approach per simulator.*
Auxiliary losses - How does Solargis Evaluate compare
Solargis Evaluate models auxiliary consumption explicitly, splitting day and night loads, a treatment it shares with PVsyst and SolarFarmer.
Solargis Prospect and pvlib do not model auxiliary loads at all, and SAM only captures inverter self-consumption, missing everything else running on the AC bus.
For plants with tracker drives, monitoring systems, or active cooling, folding this into a single availability number — as three of the six simulators effectively do — understates a real, quantifiable loss.
#### Numerical results
Results are computed for the six test sites and four system configurations defined in the [Setup and test methodology](/v1/docs/comparison-setup-and-test-methodology) article.
Auxiliary losses in pvlib were implemented in custom code, since pvlib has no built-in model (Table 1). A separate auxiliary loss series is available only for Solargis Evaluate, PVsyst, and pvlib, compared in Figures 1 and 2. Bias stays within ±0.04% and RMSE below 0.10% at every site and configuration. Given identical input values, the differing implementations produce negligible differences in simulated yield. These differences are more than an order of magnitude smaller than the grouped AC side loss differences reported below.
.png)
***Figure 1****: Auxiliary loss bias for pvlib and PVsyst against Solargis Evaluate, as a percentage of the AC power at the inverter output entering the step, by mounting configuration across the six test sites.*
.png)
***Figure 2****: Auxiliary loss RMSE for pvlib and PVsyst against Solargis Evaluate, as a percentage of the AC power at the inverter output entering the step, by mounting configuration across the six test sites.*
**Per-site breakdowns of the auxiliary losses bias and RMSE**
.png)
.png)
### AC cable losses
AC cable losses are ohmic losses on conductors between the inverter outputs and the grid connection point, through any transformer stages. The basic loss mechanism is the same heat dissipation in conductor resistance covered for DC cabling in [Electrical modeling of the DC side](/v1/docs/pv-simulators-comparison-electrical-modeling-of-the-dc-side#modeling-approach): $P = R \cdot I^2$. The AC side differs in that the chain typically spans multiple voltage levels, and some simulators bundle AC cable losses with transformer losses into a single factor.
AC cable losses are typically 1 to 3% of generated energy. The choice of single-value versus multi-stage modeling matters most for utility-scale plants with long MV or HV cable runs, where neglecting these stages can understate total AC losses by half a percentage point or more. Plants with short cable runs and a single transformer stage are well-served by a single-value approximation.
#### AC cable losses modeling
There are several approaches:
- **Physical resistance with simulated current** (Solargis Evaluate, Solargis Prospect, PVsyst) computes $P = R \cdot I^2$ across the AC wiring at each time step.
- **STC (Standard Test Conditions) percentage proxy** is the common input form, with the percentage back-calculated to a resistance.
- **Aggregated loss factor combined with transformers** is SAM's approach: AC cable losses are bundled with transformer losses into a single AC losses factor applied to the inverter output.
- **Partial coverage** is SolarFarmer's approach: low voltage cabling between the inverter and its transformer is modeled, but medium and high voltage AC cabling are neglected. pvlib does not document a specific approach, users implement their own.
#### Voltage stages
Real-world PV power plants include several AC voltage stages – low voltage (LV) between the inverter and the first transformer, medium voltage (MV) between transformer stages, and high voltage (HV) between the power transformer and the grid connection. The compared simulators model some or all of these stages, and do so differently, as described in Table 2 below.
#### Detailed design support
Solargis Evaluate's cable sizing tool accepts explicit cable dimensions (length, diameter, layout, resistivity) and custom routes to the IEC standards listed in Table 2. This matters at the detailed design stage, when site-specific cable runs are known and a percentage approximation is too coarse for the plant's layout.
#### Methodology comparison
| Aspect | Solargis Evaluate | Solargis Prospect | pvlib | PVsyst | SAM | SolarFarmer |
| --- | --- | --- | --- | --- | --- | --- |
| **Modeling** | $P = R \cdot I^2$ across LV, MV, and HV stages with all connectors and combiner boxes | $P = R \cdot I^2$ across all AC components, single value | Not specifically documented, user implementation | $P = R \cdot I^2$ across LV, MV, and HV stages from resistivity LV losses definable per inverter or per system | Aggregated AC losses factor combined with transformer losses | LV only (inverter to transformer) MV and HV neglected |
| **Input** | Percentage at STC, defaults 1% LV, 0.5% MV, 0.05% HV | Percentage at STC | User-implemented | Relative ohmic loss against PV array or inverter nominal power | User-entered percentage of AC output | Percentage at inverter maximum AC output |
| **Sizing tool** | Custom cable routes per IEC 60287, 60502, 60038, 60228 | Not available | Not available | Not available | Not available | Not available |
***Table 2****: AC cable losses modeling, input, and sizing tool comparison per simulator.*
AC cable losses - How does Solargis Evaluate compare
Solargis Evaluate and PVsyst are the only two simulators that model AC cable losses across all three voltage stages (LV, MV, HV) explicitly.
Solargis Evaluate is the only one offering a dedicated cable-sizing tool for custom routes to IEC standards.
Solargis Prospect's single aggregate AC value is a reasonable simplification for its fast prefeasibility role, but not a substitute for multi-stage modeling at later design stages.
#### Numerical results
The simulators expose the AC side losses at different granularities, which prevents a fair comparison of this stage on its own. The numerical results for the AC side losses are therefore presented in the [grid connection](/v1/docs/comparison-electrical-modeling-ac-side#grid-connection) section, grouping AC cable losses, [transformer losses](/v1/docs/comparison-electrical-modeling-ac-side#transformer1), and [grid connection losses](/v1/docs/comparison-electrical-modeling-ac-side#grid-connection).
### Transformer losses
Transformers step up the inverter AC output to higher voltages for transmission. Transformer losses are typically 1% to 2% of annual generation and split into two components that behave differently over time:
- **Iron losses** (no-load losses) come from magnetization of the core. They are constant whenever the transformer is energized: the same absolute kW is drawn at night as at peak generation.
- **Copper losses** (load losses) are proportional to the square of the winding current. They fall sharply at low output and rise at high output.
The timing difference matters most at low capacity factors, for example high latitudes with short winter days or heavily shaded sites. Across the year, iron losses then make up a larger share of total transformer losses than the nameplate full-load percentage suggests. A single efficiency percentage cannot represent this time dependence, and can misstate total transformer losses by a fraction of a percentage point.
#### Modeling approach
There are three families of models:
- **Two-loss physical models** (Solargis Evaluate's proprietary model, PVsyst's explicit iron + copper formulation, pvlib's simple efficiency model) separate the two loss components, so iron losses are constant whenever the transformer is energized and copper losses scale with the square of the load current. SolarFarmer separates no-load and full-load ohmic losses (conceptually similar).
- **Simple efficiency model** (Solargis Prospect) collapses the losses into a single efficiency factor. Solargis Evaluate offers this as an override option (standard 1% loss, high-efficiency 0.9% loss, or custom).
- **Combined with AC losses** (SAM) bundles transformer iron and copper losses into the aggregated AC losses factor that also includes AC cable losses.
#### Multi-stage support
Many utility-scale plants have two transformer stages:
1. inverter transformers stepping LV to MV at the inverter pad, and
2. a single plant transformer stepping MV to HV at the grid connection point.
Table 3 shows which simulators model the stages separately.
#### Night disconnect
PVsyst supports an optional night disconnect feature: the transformer is de-energized at night, eliminating iron losses during non-generation hours. Solargis Evaluate keeps the transformer energized continuously by default. Where the option is not available, nighttime iron losses are absorbed into the annual loss figure.
#### Transformer sizing
Iron and copper losses are functions of the transformer's power rating, so the selected transformer size affects the loss calculation. Solargis Evaluate exposes three selection modes (Table 3); the other simulators handle this through the transformer specification.
#### Methodology comparison
| Aspect | Solargis Evaluate | Solargis Prospect | pvlib | PVsyst | SAM | SolarFarmer |
| --- | --- | --- | --- | --- | --- | --- |
| **Loss model** | Iron and copper losses (proprietary) Optional simple efficiency override (1% / 0.9% / custom) | Simple efficiency (1% / 0.9% / custom) | Simple efficiency with no-load and load losses | Iron and copper losses | Iron and copper losses combined into AC losses factor | No-load and full-load ohmic losses |
| **Multi-stage** | LV/MV and MV/HV stages | Single value | Single value | LV/MV and MV/HV stages Cascade supported | Combined with AC losses | Single value |
| **Sizing options** | Adaptive, manual, or apparent rated power | - | Via specification | Via specification | Via specification | Via specification |
***Table 3****: Transformer losses modeling approach per simulator.*
Transformer losses - How does Solargis Evaluate compare
Solargis Evaluate and PVsyst are the only two simulators that model transformer losses across multiple voltage stages. The others apply one factor across the whole AC chain.
Solargis Evaluate, PVsyst, pvlib, and SolarFarmer all separate load-independent (iron) from load-dependent (copper) losses. Solargis Prospect uses a single efficiency factor, and SAM folds transformer losses into its combined AC losses figure, losing the timing distinction in both cases.
Solargis Evaluate is the only simulator offering three transformer-sizing modes (adaptive, manual, or custom apparent power), letting the loss calculation follow the actual selected equipment.
#### Numerical results
The simulators expose the AC side losses at different granularities, which prevents a fair comparison of this stage on its own. The numerical results for the AC side losses are therefore presented in the [grid connection](/v1/docs/comparison-electrical-modeling-ac-side#grid-connection) section, grouping [AC cable losses](/v1/docs/comparison-electrical-modeling-ac-side#ac-cable-losses), transformer losses, and [grid connection losses](/v1/docs/comparison-electrical-modeling-ac-side#grid-connection).
### Grid connection
The grid connection point is where the plant interfaces with the grid operator's network. The simulator's grid connection model determines what limits and signals the plant must respect during simulation: active power limits (curtailment), reactive power or power factor requirements, and operational constraints. Where these are applied in the simulation chain affects how curtailment interacts with upstream losses.
For plants in transmission-constrained or high-renewable-penetration grids, where frequent curtailment is expected, the interaction between curtailment and clipping changes the modeled total energy delivered. Where curtailment signals vary within the simulated period, a static limit reflects average curtailment behavior but not intra-period variability. Table 4 shows which simulators represent time variation directly and which model reactive support.
#### Curtailment modeling
There are three approaches:
- **Per-timestep at inverter level** (Solargis Evaluate, PVsyst) applies the limit to each simulation step, distributing the constraint across inverters by reducing their maximum power point (MPP) target. Solargis Evaluate applies a single static limit, constant across the full simulation. PVsyst supports both a static limit (at the inverter rated value or at the injection point, with the inverter producing more to compensate for downstream losses) and time-varying curtailment via an hourly grid limit file.
- **End-of-simulation deduction** (SAM, SolarFarmer) applies curtailment as a post-simulation reduction to aggregated outputs, either as a constant percentage, hourly losses, or hourly losses with custom periods.
- **Not modeled** (Solargis Prospect, pvlib) treats the plant as unlimited at the grid connection.
#### Reactive power and cos phi
PV power plants can export reactive power by adjusting their output power factor (cos phi). This feature impacts the active (useful) power that is exported to the grid. The details of how this feature is modeled in the compared software is shown in Table 4 below.
#### Application timing
When curtailment is applied per time step at the inverter level, it correctly interacts with clipping: an inverter curtailed below AC nameplate has less DC power to clip, so clipping losses fall and curtailment losses rise. The total energy deducted from PVOUT can differ measurably from a simple percentage cut. When curtailment is applied as an end-of-simulation deduction, the modeled clipping losses upstream are unaffected, and the deducted energy is calculated against the already-simulated PVOUT – the loss attribution between clipping and curtailment then no longer reflects what would happen in reality.
#### Methodology comparison
| Aspect | Solargis Evaluate | Solargis Prospect | pvlib | PVsyst | SAM | SolarFarmer |
| --- | --- | --- | --- | --- | --- | --- |
| **Active power limit** | Static limit at inverter level | Not modeled | User implementation | Inverter rated value or injection point limit, hourly grid limit file | Constant percentage, hourly losses, or hourly losses with custom periods | Power cut applied at end of simulation |
| **Reactive power, cos phi** | Active power, reactive power, and cos phi at grid connection point | Not modeled | User implementation | Active or apparent power limit with fixed or monthly cos phi | Not modeled | Not modeled |
***Table 4****: Grid connection modeling approach per simulator.*
Grid connection - How does Solargis Evaluate compare
Solargis Evaluate and PVsyst are the only two simulators that apply curtailment per timestep at the inverter level, which lets it interact correctly with clipping: an inverter curtailed below nameplate has less DC power to clip in the first place.
SAM and SolarFarmer apply curtailment as an end-of-simulation deduction which leaves the upstream clipping loss unchanged.
Solargis Evaluate and PVsyst are also the only two that model reactive power and cos phi at the grid connection point at all. Solargis Prospect, SAM, and SolarFarmer do not model it, and pvlib leaves it to user implementation.
#### Numerical results
The simulators expose the AC side losses at different granularities, which prevents a fair comparison of this stage on its own. The numerical results for the AC side losses are therefore presented in this section, grouping [AC cable losses](/v1/docs/comparison-electrical-modeling-ac-side#ac-cable-losses), [transformer losses](/v1/docs/comparison-electrical-modeling-ac-side#transformer1), and grid connection losses.
Although these results group three AC side loss stages, agreement is close: bias stays within ±0.9% and RMSE below 1.2% across all 72 cases. The differences are systematic rather than random. PVsyst computes a smaller AC side loss than Solargis Evaluate at every site and configuration. pvlib bias sits near zero on average but changes sign between sites. SAM computes a larger loss in 21 of 24 cases, the exceptions being the bifacial tracker at Dharan and Yaren, and fixed-tilt monofacial at Yaren.
For all three simulators, the design choices that raise plant output also lower the bias: bifacial bias is below monofacial and tracker bias is below fixed tilt at nearly every site. The bifacial tracker, the highest-output configuration, carries the lowest bias at every site for every simulator. Three independently developed simulators moving together points to the reference: Solargis Evaluate's AC side loss grows faster with load than the alternatives. The pattern is consistent with a resistive treatment, where cable and copper losses scale with the square of the current, against a percentage-of-output factor that keeps the loss share flat.
.png)
***Figure 3****: AC side loss bias for pvlib, PVsyst, and SAM against Solargis Evaluate, as a percentage of the AC power at the inverter output entering the step, by mounting configuration across the six test sites.*
.png)
***Figure 4****: AC side loss RMSE for pvlib, PVsyst, and SAM against Solargis Evaluate, as a percentage of the AC power at the inverter output entering the step, by mounting configuration across the six test sites.*
**Per-site breakdowns of the AC side losses bias and RMSE**
.png)
.png)
### System unavailability
Unavailability covers periods when the plant cannot deliver energy despite available irradiance. It is conceptually distinct from the per-timestep losses discussed earlier in this article: those represent physical energy losses on conductors and components, while unavailability represents outages where the plant could have generated but did not.
Two physical sources are distinguished:
- **Internal unavailability** covers plant-side outages: equipment failures, scheduled maintenance, fault response time
- **External unavailability** covers grid-side outages: transmission faults, curtailment instructions from the grid operator, force majeure events affecting the grid. The two have different operational and contractual implications – internal is the responsibility of plant operations, external is typically borne by the grid operator or the offtaker.
The internal and external split lets the operator distinguish plant reliability from grid availability. That separation is meaningful in bankability studies, where the two are negotiated separately with the lender and the offtaker. A single combined figure loses the distinction, and defining unavailability per period allows scheduled maintenance to be planned against known seasonal generation.
#### Modeling approach
Five different treatments exist across the simulators:
- **Internal/external split** (Solargis Evaluate) separates the two types and applies each as a percentage to aggregated yearly long-term PVOUT.
- **Defined periods or percentage** (PVsyst) supports either an overall percentage or specific time periods within the simulated year, allowing scheduled maintenance windows to be simulated.
- **Single aggregate value** (Solargis Prospect, SolarFarmer) applies one percentage to aggregated outputs.
- **Combined with grid curtailment** (SAM) uses one input parameter to drive both grid curtailment and unavailability, as a constant percentage, hourly losses, or hourly losses with custom periods.
- **Not modeled** (pvlib) leaves the implementation to the user.
#### Application point
Unavailability losses are modeled at different points in the simulation chain in the different simulators, as Table 5 details.
#### Methodology comparison
| Aspect | Solargis Evaluate | Solargis Prospect | pvlib | PVsyst | SAM | SolarFarmer |
| --- | --- | --- | --- | --- | --- | --- |
| **Modeling** | Internal and external types, each as percentage to aggregated yearly long-term PVOUT | Single percentage on yearly PVOUT (entered as availability) | User implementation | Overall percentage or defined unavailability periods | Combined with grid curtailment: constant, hourly, or hourly with custom periods | Percentage to time-aggregated (monthly or annual) output |
| **Application** | End of simulation, to aggregated PVOUT | End of simulation, to aggregated PVOUT | User-defined | Per simulation period or overall | Per-timestep multiplier to simulation outputs | End of simulation, to aggregated PVOUT |
***Table 5****: System unavailability losses modeling approach per simulator.*
System unavailability losses - How does Solargis Evaluate compare
Solargis Evaluate is the only simulator that separates internal (plant-side) from external (grid-side) unavailability — a distinction that matters in bankability studies, where the two are typically negotiated separately with the lender and offtaker.
PVsyst's ability to define specific unavailability periods, rather than only an overall percentage, is a complementary strength for modeling scheduled maintenance windows against known seasonal generation.
Solargis Prospect and SolarFarmer apply a single aggregate figure, and pvlib leaves unavailability to user implementation.
SAM combines unavailability with grid curtailment into one input, which means a change in modeled curtailment implicitly changes modeled unavailability, unless the inputs are deliberately recalibrated.
#### Numerical results
Because the simulators each model system unavailability differently and apply the loss at different stages in the simulation chain, a fair numerical comparison is not possible.
### Further reading
#### Solargis knowledge base
- "[Argus electrical simulation overview](/v1/docs/argus-electrical-simulation-overview)": Solargis
- "[Argus simulation post-processing overview](/v1/docs/argus-simulation-post-processing-overview)": Solargis
- "[Argus PV simulation chain](/v1/docs/argus-pv-simulation-chain)": Solargis
- "[Comparison setup and test methodology](/v1/docs/comparison-setup-and-test-methodology)": Solargis
#### Cable losses
- ["IEC 60287 - Electric cables: calculation of the current rating"](https://webstore.iec.ch/en/publication/68118): by International Electrotechnical Commission (IEC)
- ["IEC 60502 - Power cables with extruded insulation and their accessories"](https://webstore.iec.ch/en/publication/2274): by International Electrotechnical Commission (IEC)
- ["IEC 60038 - IEC standard voltages"](https://webstore.iec.ch/en/publication/72877): by International Electrotechnical Commission (IEC)
- ["IEC 60228 - Conductors of insulated cables"](https://webstore.iec.ch/en/publication/1065): by International Electrotechnical Commission (IEC)
#### General simulator comparison
- ["Cross-validation of PV system simulation software"](https://www.researchgate.net/publication/335842590_Cross-validation_of_PV_System_Simulation_Software): by Driesse, A., Patel, N.