Full-duplex communication has garnered significant attention due to its spectral efficiency advantages, yet it faces the challenge of nonlinear self-interference stemming from simultaneous transmission and reception on the same frequency. Existing nonlinear self-interference cancellation methods typically rely on nonlinear modeling to reconstruct interference, but suffer from strong correlation among basis functions—leading to slow convergence and limited performance—as well as additional modeling inaccuracies caused by In-phase/Quadrature (IQ) imbalance. Conventional orthogonalization techniques such as singular value decomposition (SVD) decomposition are computationally complex and require signal stationarity, while most IQ compensation schemes do not incorporate orthogonalization, limiting their practical utility. To tackle these challenges, this paper proposes an expanded basis function-based orthogonalized nonlinear model that enables online orthogonalization in non-stationary scenarios. The proposed method improves interference suppression performance by over 3.5 dB, effectively accommodates IQ imbalance, adapts to non-stationary systems, and achieves performance comparable to the state-of-the-art upper bound..