基于RS和GA的动态模糊神经网络在短期电力负荷预测中应用

Application of dynamic fuzzy neural network based on rough set theoryand GA in power system short-term load forecast

分析和探讨了粗糙集(RS)理论、遗传算法(GA)、模糊神经网络相结合的短期负荷预测方法。首先,对采集到的信息进行特征提取,然后利用模糊粗糙集理论中的信息熵进行属性简化、去掉冗余信息,最后用得到的属性作为模糊神经网络的输入进行训练预测。在模糊神经网络内部引入递归环节,构成了动态模糊神经网络DFNN(DynamicFuzzyNeuralNetwork),并采用具有全局寻优能力的遗传算法训练网络,克服了单纯BP算法易陷入局部最优解的缺点。用该方法与常用BP神经网络及Fuzzy法分别对某电网进行一周的日负荷预测,实例的对比分析表明了该方法收敛速度、预测精度和网络规模等方面都有较大改善。

An approach to power system short-term load forecast combining rough set theory, GA （Genetic Algorithm） and fuzzy neural network is discussed. The information features are extracted. The information entropy of fuzzy-rough set theory is used to throw away the redundant information and simplify the attributes,which are put into the fuzzy neural network for training. A DFNN（Dynamic Fuzzy Neural Network） is constructed by introducing recursion segment into the fuzzy neural network,which is trained using genetic algorithm and BP algorithm to avoid being trapped in local convergence. The daily loads of a week are forecasted for a provincial power system with the presented method, the BP neural network and the fuzzy method respectively. Results show that the presented method is better in convergence speed,forecast precision and network scale.