[Objectives] In large-scale wind farms

the horizontal spacing between upstream and downstream wind turbines is generally 5 to 10 times the impeller diameter. The wake generated by upstream turbines will cause wind speed deficit and increased turbulence intensity in the downstream area

significantly reducing the wind energy capture efficiency of downstream turbines and resulting in a 10%～20% decrease in the overall power generation of the wind farm. Accurate prediction of wind turbine wake characteristics is a core prerequisite for improving wind farm power generation efficiency

optimizing turbine layout

and implementing effective wake control. Current mainstream Gaussian wake models mostly set the wake expansion coefficient as a fixed value or only correct parameters through a single empirical formula

failing to fully combine measured data to achieve multi-factor coupling optimization

which leads to obvious limitations in prediction accuracy and engineering applicability under complex wind farm environments. This paper aims to address the above problems. [Methods] The authors derive and construct the Jensen-Gaussian wake model by introducing the Gaussian probability density function to describe the wake velocity deficit distribution based on the linear wake expansion assumption of the classical Jensen wake model. Combined with the Frandsen wake turbulence intensity model

a quantitative correlation between the wake expansion coefficient and turbulence intensity is established

breaking through the limitation of fixed parameters in conventional models. Using the gradient descent algorithm

the key parameters km and kn of the model are iteratively optimized based on the measured wind speed data from the coastal wind farm in Wenling

Zhejiang Province

and the Nibe wind farm in Denmark

so as to determine the optimal parameter combination. The optimized model is compared and verified with the classical two-dimensional Gaussian model

two-dimensional entrainment(2 DCLE) model

yawed wake model

and three-dimensional coupling(3 DCOU) model. [Results] The results show that the optimized Jensen-Gaussian wake model can more accurately characterize the velocity distribution characteristics in the near and far wake regions of wind turbines. In the Zhejiang wind farm case

the average error is only 2.06%

and the prediction accuracy is improved by 4.42% compared with that before optimization. In the Nibe wind farm case

the average error is 1.47%

and the prediction accuracy is increased by 4.10% compared with the pre-optimization model. Moreover

the prediction accuracy in the far wake region is significantly higher than that in the near wake region

and the overall performance is superior to all the comparison models. [Conclusion] This model takes into account both theoretical rationality and engineering practicability

and can accurately predict the wake wind speed under different turbine specifications and environmental conditions. It provides reliable theoretical support and technical reference for micro-site selection of wind farms

unit layout optimization

and the formulation of active wake control strategies.