As internet services newly emerge with diversity and complexity, great challenges and demands are presented to the Open Flow controlled software defined optical networks(SDON) to achieve better match between services ...As internet services newly emerge with diversity and complexity, great challenges and demands are presented to the Open Flow controlled software defined optical networks(SDON) to achieve better match between services and SDON. With this aim, this paper proposes a naive Echo-State-Network(Naive-ESN) based services awareness algorithm of the software defined optical network, where the naive ESN model adopts the ring topology structure and generates the probability output result to determine the Qo S policy of SDON. Moreover, the Naive-ESN engine is also designed in controller node of SDON to perform services awareness by obtaining service traffic features from data plan, together with some necessary extension of the Open Flow protocol. Test results show that the proposed approach is able to improved services-oriented supporting ability of SDON.展开更多
火电机组在现代电力系统中承担着大量的调峰调频任务,通过运行参数建立出力预测模型有助于快速稳定地调整功率。提出一种改进的深度回声状态网络(Deep Echo State Networks,DESN)用于建立机组出力预测模型。该改进型具备可变的记忆能力...火电机组在现代电力系统中承担着大量的调峰调频任务,通过运行参数建立出力预测模型有助于快速稳定地调整功率。提出一种改进的深度回声状态网络(Deep Echo State Networks,DESN)用于建立机组出力预测模型。该改进型具备可变的记忆能力以应对调整部分运行参数作用于机组出力变化存在的延时性,并根据运行参数聚类生成输入权重进一步挖掘运行参数与出力之间的映射信息。利用华北地区某火电机组不同工作状况下的两种数据集验证了模型效果。结果表明,改进得到的KM-VML-DESN相较于深度回声状态网络、多层感知机、长短期记忆网络等具备更强的预测性能。展开更多
光伏发电系统受到气象因素和环境变化的影响,功率波动性较大,短期预测存在较高的不确定性。目前,传统的预测方法面临精度不足和实时性差的挑战。研究基于动态回声状态网络(Dynamic Echo State Network,DESN)的光伏发电短期功率预测模型...光伏发电系统受到气象因素和环境变化的影响,功率波动性较大,短期预测存在较高的不确定性。目前,传统的预测方法面临精度不足和实时性差的挑战。研究基于动态回声状态网络(Dynamic Echo State Network,DESN)的光伏发电短期功率预测模型,采集气象数据与历史发电数据,构建适应光伏发电特点的DESN模型框架,并对模型进行了优化与实验验证,为光伏发电的智能调度与优化运行提供有力支持。展开更多
标准灰狼优化(grey wolf optimizer,GWO)算法存在局部探索和全局开发难以平衡等问题。针对此类问题,提出基于多策略结合的灰狼优化算法(multi-strategy grey wolf optimization,MSGWO)。首先,灰狼算法引入非线性收敛因子和Tent映射;然后...标准灰狼优化(grey wolf optimizer,GWO)算法存在局部探索和全局开发难以平衡等问题。针对此类问题,提出基于多策略结合的灰狼优化算法(multi-strategy grey wolf optimization,MSGWO)。首先,灰狼算法引入非线性收敛因子和Tent映射;然后,利用广泛学习、精英学习和协调学习三种策略,在GWO优化过程中协调工作;最后,利用轮盘赌进行策略选择,以获得更具多样性灰狼位置和更具全局代表性的个体。通过标准基准函数测试,采用算法变体进行对比。结果显示,MSGWO算法拥有较好的全局搜索、局部开发的平衡能力以及更快的收敛速度。在此基础上,利用MSGWO算法优化回声状态网络(echo state networks,ESN)超参数进行回归预测。实验表明平均绝对百分比误差为0.38%,拟合程度达到0.98,验证了MSGWO算法的优化性能。展开更多
基金supported by the Science and Technology Project of State Grid Corporation of China:“Research on the Power-Grid Services Oriented “IP+Optical” Coordination Choreography Technology”.
文摘As internet services newly emerge with diversity and complexity, great challenges and demands are presented to the Open Flow controlled software defined optical networks(SDON) to achieve better match between services and SDON. With this aim, this paper proposes a naive Echo-State-Network(Naive-ESN) based services awareness algorithm of the software defined optical network, where the naive ESN model adopts the ring topology structure and generates the probability output result to determine the Qo S policy of SDON. Moreover, the Naive-ESN engine is also designed in controller node of SDON to perform services awareness by obtaining service traffic features from data plan, together with some necessary extension of the Open Flow protocol. Test results show that the proposed approach is able to improved services-oriented supporting ability of SDON.
文摘火电机组在现代电力系统中承担着大量的调峰调频任务,通过运行参数建立出力预测模型有助于快速稳定地调整功率。提出一种改进的深度回声状态网络(Deep Echo State Networks,DESN)用于建立机组出力预测模型。该改进型具备可变的记忆能力以应对调整部分运行参数作用于机组出力变化存在的延时性,并根据运行参数聚类生成输入权重进一步挖掘运行参数与出力之间的映射信息。利用华北地区某火电机组不同工作状况下的两种数据集验证了模型效果。结果表明,改进得到的KM-VML-DESN相较于深度回声状态网络、多层感知机、长短期记忆网络等具备更强的预测性能。
文摘光伏发电系统受到气象因素和环境变化的影响,功率波动性较大,短期预测存在较高的不确定性。目前,传统的预测方法面临精度不足和实时性差的挑战。研究基于动态回声状态网络(Dynamic Echo State Network,DESN)的光伏发电短期功率预测模型,采集气象数据与历史发电数据,构建适应光伏发电特点的DESN模型框架,并对模型进行了优化与实验验证,为光伏发电的智能调度与优化运行提供有力支持。
文摘标准灰狼优化(grey wolf optimizer,GWO)算法存在局部探索和全局开发难以平衡等问题。针对此类问题,提出基于多策略结合的灰狼优化算法(multi-strategy grey wolf optimization,MSGWO)。首先,灰狼算法引入非线性收敛因子和Tent映射;然后,利用广泛学习、精英学习和协调学习三种策略,在GWO优化过程中协调工作;最后,利用轮盘赌进行策略选择,以获得更具多样性灰狼位置和更具全局代表性的个体。通过标准基准函数测试,采用算法变体进行对比。结果显示,MSGWO算法拥有较好的全局搜索、局部开发的平衡能力以及更快的收敛速度。在此基础上,利用MSGWO算法优化回声状态网络(echo state networks,ESN)超参数进行回归预测。实验表明平均绝对百分比误差为0.38%,拟合程度达到0.98,验证了MSGWO算法的优化性能。
