Multi-stage power supply balancing planning based on an improved multi-objective gray wolf algorithm under dual carbon goals

Abstract

Under the dual-carbon goals, the large-scale deployment of power electronic equipment in the power system has led to multiple uncertainties on both the power generation and load sides, posing two urgent problems for multi-stage power supply planning: one is how to collaboratively reduce generation costs and carbon emissions during periods of surplus power; the other is how to achieve temporal and spatial balance and low-carbon coordination of power shortages during periods of shortage. Traditional multi-objective optimization methods such as NSGA-II (Nondominated Sorting Genetic Algorithm II) and MOPSO (Multi-Objective Particle Swarm Optimization) encounter difficulties when dealing with such high-dimensional non-convex optimization problems. Due to the existence of numerous local optimal regions in the solution space and the discontinuity of the Pareto front, the crossover mutation or particle update mechanisms relied upon by these methods are prone to lose population diversity in high-dimensional conditions, causing the algorithm to prematurely converge to the local Pareto front and making it difficult to obtain a uniformly distributed global optimal solution set. As a result, the temporal and spatial balance and low-carbon coordination goals of multi-stage power supply are difficult to achieve. This article proposes a multi-stage power supply balancing planning method based on an improved multi-objective grey wolf algorithm, which predicts the load of the power storage system using a scenario tree, integrates user side demand response to predict consumer load, and constructs national and provincial power grid planning decision models. When there is an excess of electricity, the model reduces the comprehensive cost of power generation and carbon emissions, when there is a power shortage, an improved algorithm optimization model combining Kent chaotic mapping, Levier flight strategy, and greedy selection mechanism is applied to set multidimensional constraints to ensure spatiotemporal balance and low-carbon transformation, while greedy selection ensures convergence. The experiment shows that this method can accurately predict the power load on the user side. When there is an excess of electricity, the average peak valley load difference on the user side decreases by more than 50%. When there is a shortage of electricity, the power deficit decreases and stabilizes, verifying its effectiveness in strengthening the integration of renewable energy and alleviating load fluctuations. Clinical trial: Not applicable