Research on Multi-Scale Modeling of Grid-Connected Distributed Photovoltaic Power Generation

Chen Lv, Wanxing Sheng, Keyan Liu, Xinzhou DongEnergy 2017 · DOI

Abstract

The complexity of distribution network models largely depends on the scale of grid-connected distributed photovoltaic (PV) power generation. This paper proposes a multi-scale modeling framework based on mathematical models of PV systems. We analyze simulation performance differences—including step-size adaptability, output accuracy, and effects on voltage profiles—across three model scales using the IEEE 33-node benchmark. Results demonstrate that multi-scale modeling effectively balances simulation accuracy and efficiency for diverse distribution network operating conditions.

Introduction

As distributed photovoltaic (PV) systems become integral to modern power grids, engineers face a significant challenge in simulation: balancing precision with computational speed. A model that is too simple may fail to capture critical dynamics like weather-driven transients or voltage dips. Conversely, a model that is too complex—simulating every high-frequency switch in an inverter—can become computationally prohibitive for large-scale network studies or real-time applications. This research addresses this "Goldilocks" problem by defining and evaluating three distinct scales of PV models.

System Architecture

The foundation of the study is a comprehensive mathematical model of a grid-connected PV system. This system is divided into two primary sections: the DC side and the AC side. The DC side includes PV arrays, Boost converters for voltage regulation, and Maximum Power Point Tracking (MPPT) controllers. The AC side consists of three-phase inverters, LC filters, and the connection to the utility grid.

Block diagram of the grid-connected PV system architecture including DC and AC sides.

Multi-Scale Modeling Framework

The researchers categorized PV systems into three levels of complexity to suit different simulation needs:

  1. Generalized Load Model: This is the simplest representation, where the PV system is treated as a "negative load." It consumes negative active and reactive power (-P, -Q), ignoring all internal switching and DC-side physics. It is ideal for steady-state power flow analysis.
  2. Model with DC Voltage Source: A mid-tier model that assumes the DC side is a constant voltage source. While it simplifies the solar input, it maintains a detailed model of the AC inverter and its control strategy, making it suitable for certain dynamic studies.
  3. Detailed Model: The most complex version, incorporating the full physics of the PV array and the DC-DC converter dynamics. This model uses the Perturbation and Observation (P&O) method for MPPT to track the maximum power point under varying irradiance.
The equivalent circuit model of a PV cell, the building block of the Detailed Model.

Simulation Performance and Step-Size Sensitivity

A critical finding of the research is the sensitivity of the Detailed Model to the simulation time step ($T_s$). While the model is highly accurate at a fine resolution (e.g., 5 $\mu$s), it becomes unstable and produces massive, unrealistic oscillations when the step size is increased to 50 $\mu$s.

In contrast, the Generalized Load and DC Voltage Source models remain stable even at much larger step sizes (up to 200 $\mu$s). This has significant implications for real-time hybrid simulations, which often require step sizes between 20 $\mu$s and 50 $\mu$s. In such environments, the DC Voltage Source model often provides the best balance of stability and detail.

Network Impact and Results

The models were tested using the IEEE 33-node distribution network, with the PV system connected at Node 17.

  • Power Tracking: Only the Detailed Model accurately tracked the non-linear power output during rapid changes in solar intensity (e.g., a step change from 100% to 30%). Simpler models introduced errors because they lack the DC-side mathematical "memory."
  • Voltage Profile: For steady-state analysis, all three models produced nearly identical voltage rise effects at the point of connection. This suggests that for basic voltage regulation studies, the simplest Generalized Load Model is sufficient and computationally efficient.

Conclusion

This research establishes a clear hierarchy for PV modeling. While detailed models are essential for capturing high-speed transients and weather-driven fluctuations, simpler models are often superior for large-scale network stability and real-time simulations. Future work may involve "adaptive multi-scale modeling," where a simulation dynamically switches between scales based on the specific event being analyzed, such as a fault or a rapid change in irradiance.

Reference: Lv, C., Sheng, W.X., Liu, K.Y. and Dong, X.Z. (2017) Research on Multi-Scale Modeling of Grid-Connected Distributed Photovoltaic Power Generation. Energy and Power Engineering, 9, 127-140. DOI: 10.4236/epe.2017.94B016

Summary updated

Abstract

The complexity of distribution network model mainly depends on the model scale of grid-connected distributed photovoltaic (PV) power generation. Therefore, the simulation performance of multi-scale PV model is the key factor of the simulation accuracy in the specific operating scenarios of distribution network. In this paper, a multi-scale model of grid connected PV distributed generation system is proposed based on the mathematical model of grid-connected distributed PV power generation. It is analyzed that differences of simulation performance, such as adaptability of simulation step size, accuracy of output and the effect on voltage profile of distribution network, between PV models with different scales in IEEE 33 node example. Simulation results indicate that the multi-scale model is effective in improving the accuracy and efficiency of simulation under different operating conditions of distribution network.
BibTeX
@inproceedings{lv2017research,
  author = {Chen Lv and Wanxing Sheng and Keyan Liu and Xinzhou Dong},
  title = {{Research on Multi-Scale Modeling of Grid-Connected Distributed Photovoltaic Power Generation}},
  booktitle = {Energy},
  year = {2017},
  doi = {10.4236/epe.2017.94b016},
}
Figure