Applying deep reinforcement learning to active flow control in weakly turbulent conditions