In view of the current problems of only single operation mode, vague profit method, and low utilization rate of the energy storage business for power grid, a business operation mode combining self-operated electricity and shared energy storage is proposed. First, a multi-agent collaborative scheduling model of the power distribution network considering shared energy storage is established. This model takes into account the interests of the power distribution network, load aggregators, and energy storage providers, while effectively reducing the peak-to-valley difference of the net load and hence relieving the peak shaving pressure of the power grid. Then, the model implements a two-stage optimization. The first stage is the leasing optimization of energy storage capacity, that is, the power distribution network leases energy storage on demand for peak shaving and valley filling, minimizing leasing costs and net load variance. The second stage is multi-agent collaborative optimization, where energy storage providers arbitrage by using remaining capacity to "store at low prices and discharge at high prices" based on time-of-use electricity prices, and respond to peak shaving together with load aggregators. This minimizes the cost of the power distribution network, maximizes the benefits of load aggregators, and maximizes the arbitrage of energy storage providers, achieving mutual benefit and win-win results for all parties. Finally, the effectiveness of the proposed method is verified through example analysis.