基于粒子群算法的自回归加权马尔可夫链的负荷预测

Load forecasting based on particle swarm optimization autoregressive weighted Markov chain

针对城市居民人均年用电量序列具有波动性、随机性和相依性的特点,建立了基于粒子群优化算法的自回归加权马尔可夫链的负荷预测模型(PSO-AR-W-MC)。首先利用粒子群算法和AIC准则确定出自回归AR模型的系数和阶数,并对负荷变化趋势进行预测。利用平行曲线法划分该模型得到的残差序列,建立马尔可夫链的状态区间,以此求出状态转移概率矩阵,利用归一化后的自相关系数对其进行改进,确定出预测数据所属状态区间。根据状态区间对预测值进行第二次拟合。实例分析表明该算法具有较高的精确度和可靠性,应用前景广阔。

Volatility and randomness and dependencies of the characteristics of urban residents per capita electricity consumption sequence established based Particle Swarm Optimization algorithm self the regression weighted Markov chain load forecasting model （PSO-AR-W-MC）.First use of the PSO and AIC criterion to determine from the AR model coefficients and order , and load trend forecast .Divided the model residual series parallel curve method to establish the range of state of the Markov chain , in order to find the state tran-sition probability matrix , normalized autocorrelation coefficient to improve it , to determine the forecast data belongs state interval .According to state interval predictive value second fitting .The case study shows that the proposed algorithm has a high degree of accuracy and reliability , broad application prospects .