The growing use of photovoltaic energy necessitates accurate modeling of modules, particularly in non-linear current–voltage characteristics, despite challenges in finding ideal circuit model parameters. Several meta-heuristic techniques have been developed to determine these parameters. This work presents a new method that combines the Crayfish Optimization Algorithm (COA) with fractional-order chaos maps (FC-maps), a recent development in the area. The goal of this addition is to adaptively adjust the COA settings. Moreover, the COA algorithm has been improved with the dimension learning-hunting (DLH) search scheme, inspired by crayfish hunting. This method establishes neighborhoods for each crayfish, enhancing local and global search efforts while preserving diversity. First, benchmark functions are used to conduct a thorough mathematical assessment of the improved COA (ICOA). Subsequently, the ICOA is combined with the Newton-Raphson numerical method to estimate PV parameters. Six cell and module types, including RTC France, Photowatt-PWP201, and STP6-120/36, are investigated using various PV models within the single-diode and double-diode models. The root mean squared error (RMSE) for RTC France, PWP201, and STP6-120/36 that our system employing (SDM, DDM) obtained is as follows, based on the experimental findings: (7.844017E-04, 7.520345E-04), (2.084019E-03, 2.067626E-03), and (1.441938E-02, 1.432568E-02), respectively. A statistical analysis between several well-established metaheuristic methods was conducted to showcase the superior performance of the ICOA. The results unequivocally exhibit that the ICOA excels in accurately estimating the best PV parameters when compared to the other techniques. In conclusion, ICOA surpasses alternative algorithms in terms of precision, consistency, and convergence.