Information brokerage in wireless sensor networks involves producers (such as sensor nodes) storing in storage positions a large amount of data that they have collected and consumers (e.g. base stations, users, and nodes) retrieving that information. In this paper, first, the data storage problem is formalized into a one-to-one (one producer and one consumer) model, a many-to-one (m producers and one consumer) model, and a many-to-many (m producers and n consumers) model with the goal of minimizing the total energy consumption. Second, based on the above models, two algorithms are proposed to determine the storage positions based on data rates of producers, query rates of consumers, and transmission scheme of information brokerage. The optimal data storage (ODS) scheme, a greedy algorithm, produces the global optimal data storage positions and the near-optimal data storage (NDS) scheme, an approximate algorithm, can greatly reduce the computational overhead while achieving local optimal positions. Both ODS and NDS are able to adjust the storage positions adaptively to minimize energy consumption that includes the costs of storing and querying the data. Simulation results show that NDS not only provides substantial cost benefits but also performs as effective and efficient as ODS in over 70% of the tested cases.