电站锅炉“四管”泄漏的精确定位对于火电机组的安全、经济运行具有极其重要的意义。该文建立了双曲面定位模型并给出炉膛内平面四元阵列的测点分布,泄漏定位实验表明通过极大似然(maximum likelihood,ML)广义互相关对能取得非常稳定和尖锐的时间估计峰值,频域信噪比函数能较好地抑制炉膛强背景噪声的影响。为了避免初值猜测与远场假设,采用遗传算法:十进制浮点编码,非均匀高斯变异算子,进行双曲面方程组定位解的全局性概率搜索。对比遗传算法和拟牛顿法的双曲面方程组优化,表明遗传算法的优化误差略小于拟牛顿法,具有稳健性和高精度,但其需要较长的收敛时间。时延估计误差对定位影响很大,时间迟延误差不大于0.1ms,平面四元阵列空间坐标定位误差保持在1m之内从而满足精确定位的要求。 The precise positioning of the boiler “four tubes ” leakage of power station is of great significance to the safety and economic operation of thermal power units. In this paper, a hyperboloid positioning model is established and the distribution of measuring points in a planar quaternary array in a furnace chamber is given. The leak localization experiment shows that a very stable and sharp time estimate can be obtained by means of generalized maximum likelihood (ML) cross-correlation The peak and frequency-domain SNR function can restrain the influence of strong background noise in the furnace. In order to avoid initial guess and far-field assumption, genetic algorithm: decimal floating-point coding and inhomogeneous Gaussian mutation operator are used to search the global probabilistic search of hyperboloid equations. The optimization of hyperbolic equations by genetic algorithm and quasi-Newton method shows that the optimization error of genetic algorithm is slightly less than quasi-Newton method, which is robust and accurate. However, it requires longer convergence time. The time delay estimation error has a great influence on the positioning, the time delay error is not more than 0.1ms, and the space quadrilateral array spatial coordinate positioning error is kept within 1m so as to meet the requirements of precise positioning.