基于自适应极限学习机的变压器故障识别方法

针对变压器状态数据累积规模和复杂程度均增大的情况，单一智能算法进行数据处理的能力有限、精度低，提出了基于自适应极限学习机的变压器故障识别方法。利用免疫算法(IA)的多样性调节机制和存储机制对粒子种群进行优、劣分类，对优、劣粒子分别采用不同的进化方式。经IA改进的粒子群优化(PSO)算法有效克服了种群容易早熟从而导致进化停滞的缺点，提高了全局寻优能力。在参数寻优的基础上，根据寻优输出结果建立变压器故障识别模型。实验计算结果表明所提方法比极限学习机(ELM)、粒子群优化极限学习机(PSO-ELM)、遗传算法优化极限学习机(GA-ELM)方法的故障识别精度高。

Transformer fault identification method based on self-adaptive extreme learning machine

In view of the problem of limited data processing ability and low accuracy of single intelligent algorithm when the cumulative scale and complexity of transformer state data increase, a transformer fault identification method based on self-adaptive extreme learning machine is proposed. The IA(Immune Algorithm) is used to classify the superior and inferior particle populations due to its diversity adjustment mechanism and storage mechanism, and the superior and inferior particles are evolved in different ways. The PSO(Particle Swarm Optimization) algorithm improved by IA effectively overcomes the shortcoming that the population is prone to premature development and thus leads to evolutionary stagnation, and improves the global optimization ability. On the basis of parameter optimization, the transformer fault identification model is established according to the optimization output results. The experimental results show that the fault identification accuracy of the proposed method is higher than the ELM(Extreme Learning Machine) method, the PSO-ELM(Particle Swarm Optimization-based Extreme Learning Machine) method and the GA-ELM(Genetic Algorithm-based Extreme Learning Machine) method.