Addressing the issue that the error in locating the source of microseisms during underground excavation is significantly influenced by the geometry of the monitoring network, current research mainly focuses on optimizing individual error metrics with a lack of consideration for both accuracy and cost, resulting in low practical application of theoretical findings in engineering applications. To this end, a Geiger-ABC double-coupling inversion framework is proposed. At the algorithmic level, the honey-bee-follower mechanism of the artificial bee colony (ABC) algorithm is incorporated into the residual target function used in Geiger positioning. This approach enables the simultaneous optimization of the initial value of the seismic source and the spatial arrangement of the monitoring stations, overcoming the limitations of traditional least squares methods, such as their strong dependence on initial values and tendency to reach local optima. The numerical test results show that, under 5% wave velocity disturbances and the same computational conditions, the average positioning error of this approach is 42% lower than that of the traditional empirical placement method; the maximum error is kept within 0.89 m. It is also determined that the minimum number of stations required to ensure an average positioning error of no more than 0.45 m within a 200 m monitoring range is at least 40 stations. Field tests show that by deploying 40 microseismometers along the side walls of the tunnel and on the ground surface according to this scheme, an area of 200 m³ can be effectively monitored. The research results presented in this paper provide a standardized monitoring system that is both highly accurate and cost-effective for early warning of rock bursts (impact pressure) in deep tunnels. This approach is of great significance for the rapid deployment of such systems on-site, as well as for the precise prevention and control of dynamic hazards.