Model parameters are the basis of power system simulation calculations. With the continuous maturity of photovoltaic (PV) power generation technology and its promotion and application in the world, the proportion of PV power generation in the installed capacity of power system continues to increase. The model parameters of PV power generation also become an indispensable condition for the simulation calculation of power systems with a high proportion of PV power generation. Many scholars have carried out extensive and in-depth research on PV power generation modeling requirements applicable to power system simulation calculation. In reference [
1], according to the four parameters provided by the manufacturer under the standard test conditions, two detailed methods for calculating the model parameters are given. In reference [
2], a disturbance observation method based on uncertainty reasoning was proposed to realize maximum power point tracking (MPPT). In reference [
3], a MPPT technique combining a prediction model with a disturbance observation algorithm is proposed. In reference [
4], an effective strategy is presented to realize IGBT open-circuit fault diagnosis for closed-loop cascaded PV grid-connected inverters. In reference [
5], an overview of MPPT methods for PV systems used in the Micro Grids of PV systems is presented. In reference [
6], a method to improve the performance of a distribution system by optimizing volt–var function of a smart inverter to alleviate the voltage deviation problem due to distributed generation connection is proposed. In reference [
7], a coordinated voltage and reactive power control architecture for large PV power plants is proposed. In reference [
8], an improved simplified nonlinear engineering mathematical model of silicon solar cells is proposed. The output current of PV cells under any light intensity and temperature can be calculated by using the four electrical parameters under the standard test conditions provided by the PV cell manufacturer. In reference [
9], the output characteristic curve of PV cells is proposed by using the trajectory of horizontal throwing. The test results show that the replacement model can meet the engineering application accuracy under the standard test conditions. Reference [
10] is aimed at the problem that the transcendental model contained in the general model is difficult to solve, and a PV cell model based on the Bezier function is proposed. This method simplifies the modeling of PV cells based on the test data. The advantage of this method is that it does not need to estimate the internal parameters of the cell and is easy to implement.
With the improvement of accuracy requirements, many parameter identification methods based on intelligent algorithms [
11,
12,
13,
14] have been applied to photovoltaic single diode and dual diode equivalent circuit models. However, the PV model in the parameter identification method is usually limited to polynomial or transcendental functions in mathematics, which makes it difficult to optimize or solve the parameters, and also limits the adaptability of the model. Small external changes may lead to large parameter deviation, which cannot meet the flexibility requirements in many cases. Therefore, the data-driven modeling method based on machine learning and artificial intelligence technology has been gradually applied in PV modeling due to its few constraints, flexible structure, and strong adaptability. In reference [
15], a white adaptive neural network MPPT control model is proposed to optimize the multi peak output characteristic curve. In reference [
16], the relationship between temperature, light intensity, and maximum power point is established based on the back propagation (BP) network of fuzzy control. In reference [
17], binary ant colony algorithm is used to optimize the parameters of the fuzzy neural network MPPT strategy. In reference [
18], it is proposed to improve the convergence speed of the MPPT white adaptive neural network by load voltage ergodic feedback. In reference [
19], it is proposed to use the convolutional neural network (CNN) to predict the probability of PV power for medium and long-term planning of power grids. In reference [
20], a PV power probability prediction method based on Improved Bayesian neural network, taking into account the influence of persistence and suddenness, is proposed to improve the adaptability of short-term power prediction to accidental factors. Since the inverter, the core component of PV power generation, is a power electronic device and its interface characteristic response time scale is microseconds, the existing models mainly belong to the category of electromagnetic transient. The electromagnetic transient model of the inverter is very close to the actual physical process, considering the pulse width modulation and the conduction state of the inverter switch. It can simulate the high-frequency switching characteristics of the inverter, obtain accurate voltage and current transient waveform, and conduct harmonic analysis. Therefore, the electromagnetic transient model is very suitable for the design and analysis of inverter operation characteristics and control strategy. The simulation step size of the electromagnetic transient simulation is generally less than 100 μs, while the step size of the electromechanical transient simulation is generally 5–20 ms. After the inverter is filtered, most of the high frequency harmonics are filtered out, which has little impact on the power system. Therefore, it is unnecessary to consider such a detailed switching process in electromechanical transient simulation. However, the large-scale power system has a complex network and a large amount of equipment. A single PV power station contains hundreds of inverters. The direct use of the microsecond-level model will reduce the simulation efficiency of the electromechanical transient of the power system, and the measured parameters are difficult to obtain. Therefore, this paper aims to establish a PV power generation model suitable for the electromechanical transient simulation of power system and proposes a test and identification method for its model parameters. Through a disturbance test and parameter identification based on the test system, the model and parameters of the electromechanical transient model of PV power generation can be successfully fitted.