Global power systems are increasingly reliant on wind energy as a mitigation strategy for climate change. However, the variability of wind energy causes system reliability to erode, resulting in the wind being curtailed and, ultimately, leading to substantial economic losses for wind farm owners. Wind curtailment can be reduced using battery energy storage systems (BESS) that serve as onsite backup sources. Yet, this auxiliary role may significantly hamper the BESS's capacity to generate revenues from the electricity market, particularly in conducting energy arbitrage in the Spot market and providing frequency control ancillary services (FCAS) in the FCAS markets. Ideal BESS scheduling should effectively balance the BESS's role in absorbing onsite wind curtailment and trading in the electricity market, but it is difficult in practice because of the underlying coordination complexity and the stochastic nature of energy prices and wind generation. In this study, we investigate the bidding strategy of a wind-battery system co-located and participating simultaneously in both the Spot and Regulation FCAS markets. We propose a deep reinforcement learning (DRL)-based approach that decouples the market participation of the wind-battery system into two related Markov decision processes for each facility, enabling the BESS to absorb onsite wind curtailment while simultaneously bidding in the wholesale Spot and FCAS markets to maximize overall operational revenues. Using realistic wind farm data, we validated the coordinated bidding strategy for the wind-battery system and find that our strategy generates significantly higher revenue and responds better to wind curtailment compared to an optimization-based benchmark. Our results show that joint-market bidding can significantly improve the financial performance of wind-battery systems compared to individual market participation.