<p>The fluctuation of distributed photovoltaic grid-connected output leads to a high peak–valley difference rate, which compromises the stability of the power system. To address this issue, an optimization method for peak–valley time-of-use electricity pricing on the generation side is proposed, taking into account the fluctuation of distributed photovoltaic grid-connected output. This method involves constructing an output model of the photovoltaic power station to capture the uncertainty of photovoltaic output. After characterizing the fluctuation state of the distributed photovoltaic grid-connected output, peak, flat, and valley periods are identified using fuzzy clustering. A time-of-use electricity price optimization model on the generation side is established, considering variations in distributed photovoltaic grid-connected output. The objective function aims to maximize the electricity consumption during valley periods and minimize the difference in electricity consumption between peak and valley periods. An optimization strategy for time-of-use electricity pricing is designed to achieve effective pricing adjustments on the generation side in response to distributed photovoltaic output fluctuations. Experimental results demonstrate that the proposed method effectively reduces the peak–valley difference rate of the power grid. Compared with genetic algorithm and particle swarm optimization, this approach obtains high-quality solutions more rapidly, reduces computational cost, and enhances the stability of the power system.</p>