This paper presents a strategy and associated algorithms for the management of energy in a microgrid. The focus of the paper is on hourly energy transactions with the main grid. These transactions take into account consumer load, solar power generation, and dynamic electricity pricing. The algorithm employed is the deep reinforcement learning (DRL) algorithm, which aims to optimise the operation of the energy storage system (ESS) to maximise the financial benefits of the microgrid, while considering variations in consumer load, solar power generation, and electricity prices. In addition to optimising monetary outcomes, the algorithm also ensures a minimal energy reserve (SOCr), thereby guaranteeing the continuity of mission-critical operations within the microgrid. The proposed DRL model incorporates a risk factor, enabling it to address critical events involving price fluctuations. This model will be compared with a standard DRL model that does not incorporate a risk factor. The comparison will be based on the probability of discharging under the minimal energy reserve and the average monetary daily benefit.