The enterprise financial risk under the new era economy exists in all aspects of enterprise operation, thus this paper screens the enterprise financial risk early warning indicators from the four aspects of enterprise profitability, operation ability, development ability, and debt repayment ability. The logistic Steele model is introduced to optimize the population size function of differential evolution algorithm to achieve the dynamic adaptive population size. Then the adaptive differential evolution algorithm is used to optimize the threshold value of BP neural network, and the neural network prediction model based on the improved differential evolution algorithm is derived. Analyze the operation steps of the improved differential neural network algorithm model in enterprise financial security detection to realize the optimal solution of the enterprise financial risk warning model. Compare and analyze the predicted value of the improved differential neural network algorithm model with the real value of the enterprise financial development, and analyze the use of differential evolutionary algorithm in the prediction of enterprise financial risk.The prediction error of the net asset growth rate of enterprise Q in the 1st quarter and the 3rd quarter of the year 2024 is 0.0119 and -0.05309, respectively, with a smaller absolute value of the error, and the improved differential neural network algorithm is able to effectively predict the corporate financial risk.