广东电力
廣東電力
엄동전력
GUANGDONG ELECTRIC POWER
2015年
2期
64-69,92
,共7页
短期负荷预测%核主成分分析%最小二乘支持向量回归机%粒子群算法%协同优化算法%单纯形法
短期負荷預測%覈主成分分析%最小二乘支持嚮量迴歸機%粒子群算法%協同優化算法%單純形法
단기부하예측%핵주성분분석%최소이승지지향량회귀궤%입자군산법%협동우화산법%단순형법
short-term load forecasting%kernel principle component analysis(KPCA)%least square support vector regression (LSSVR)%particle swarm optimization%coordinate optimization%simplex method
分析了一种基于核主成分分析(kernel principle component analysis,KPCA)和混沌单纯形混合粒子群协同(chaos and simplex method-particle swarm coordinate optimization,CSM-PSCO)算法优化最小二乘支持向量回归机(least square support vector regression,LSSVR)的短期负荷预测模型。首先,采用 KPCA对训练样本的输入个数进行降维优选,以较少输入代替原始大量输入,同时信息大部分得以保留；然后,采用 LSSVR 对训练样本进行回归训练,训练过程中采用CSM-PSCO对LSSVR的相关参数进行优化,得到满足要求的模型；最后,采用训练好的模型对未知负荷进行预测。算例表明该模型的预测精度和速度均能满足实际的预测需求。
分析瞭一種基于覈主成分分析(kernel principle component analysis,KPCA)和混沌單純形混閤粒子群協同(chaos and simplex method-particle swarm coordinate optimization,CSM-PSCO)算法優化最小二乘支持嚮量迴歸機(least square support vector regression,LSSVR)的短期負荷預測模型。首先,採用 KPCA對訓練樣本的輸入箇數進行降維優選,以較少輸入代替原始大量輸入,同時信息大部分得以保留；然後,採用 LSSVR 對訓練樣本進行迴歸訓練,訓練過程中採用CSM-PSCO對LSSVR的相關參數進行優化,得到滿足要求的模型；最後,採用訓練好的模型對未知負荷進行預測。算例錶明該模型的預測精度和速度均能滿足實際的預測需求。
분석료일충기우핵주성분분석(kernel principle component analysis,KPCA)화혼돈단순형혼합입자군협동(chaos and simplex method-particle swarm coordinate optimization,CSM-PSCO)산법우화최소이승지지향량회귀궤(least square support vector regression,LSSVR)적단기부하예측모형。수선,채용 KPCA대훈련양본적수입개수진행강유우선,이교소수입대체원시대량수입,동시신식대부분득이보류；연후,채용 LSSVR 대훈련양본진행회귀훈련,훈련과정중채용CSM-PSCO대LSSVR적상관삼수진행우화,득도만족요구적모형；최후,채용훈련호적모형대미지부하진행예측。산례표명해모형적예측정도화속도균능만족실제적예측수구。
This paper analyzes a kind of short-term load forecasting model based on kernel principle component analysis(KP-CA)and least square support vector regression(LSSVR)due to chaos and simplex method coordinating with particle swarm optimization. Firstly,KPCA was applied for dimension reduction optimization on input numbers of training samples so as to replace original great deal of input information by less input and retain most information. Then,LSSVR was used for regres-sion training on samples and CSM-PSCO was used for optimizing relevant parameters of LSSVR in process of training in or-der to obtain satisfied model. Finally,the well-trained model was used to forecast unknown load. Examples indicate that forecasting precision and velocity of this model could meet practical forecasting requirements.
