In recent years, the application of the Internet of Things (IoT) in agriculture has gained traction, including wearable devices in Precision Livestock Farming (PLF). This paper explores the use of reinforcement learning algorithms for energy management in wearable IoT devices with energy harvesting applied in PLF scenarios. These devices monitor animal health, welfare, and location. Ensuring energy efficiency is critical due to the autonomous operation expected from these devices, which are often powered by limited-capacity batteries, and sometimes integrate energy harvesting systems. This study implements Proximal Policy Optimization (PPO), Soft Actor-Critic (SAC), and Twin Delayed Deep Deterministic Policy Gradient (TD3) algorithms to optimize energy consumption, creating an energy manager that adapts the duty cycle, frequency of data transmissions, and frequency of GPS executions based on weather conditions, orientation, and battery charge. Experiments demonstrate that reinforcement learning-based energy managers can effectively adapt consumption to different conditions. Using 10 000 episodes for training, the TD3 algorithm achieved the best overall performance. However, the PPO algorithm performed better during the autumn and winter seasons. This study contributes to future real-world implementation of reinforcement learning techniques for energy management in IoT devices for PLF.