| 引用本文: | 李 虹,韩雨萌.基于LSTM-CGAN的多微网数据驱动分布鲁棒协同优化运行策略[J].电力系统保护与控制,2024,52(18):133-148.[点击复制] |
| LI Hong,HAN Yumeng.Data-driven distributionally robust collaborative optimization operation strategy for a multi-microgrid based on LSTM-CGAN[J].Power System Protection and Control,2024,52(18):133-148[点击复制] |
|
| 摘要: |
| 新能源的强烈不确定性给多微网协同运行带来了可靠性和安全性的巨大挑战。为此，提出一种基于长短期记忆(long short-term memory, LSTM)网络和条件生成对抗网络(conditional generative adversarial networks, CGAN)的多微网数据驱动两阶段分布鲁棒协同优化调度模型。首先，为更准确地描述新能源的不确定性，该模型以LSTM-CGAN生成和K-means++聚类算法削减得到的场景集作为分布鲁棒优化集合的初始新能源场景。其中CGAN网络模型使用Wasserstein距离作为判别器损失函数，以新能源日前预测值作为生成对抗网络的条件变量，并采用LSTM构建生成器和判别器。其次，提出一种基于多能点对点交易贡献率的利益分配方法，以实现合作收益的公平分配。然后，为保护各主体隐私并提高求解效率，提出一种耦合可并行计算列与约束生成(column and constraint generation, C&CG)的交替方向乘子法(alternating direction multiplier method, ADMM)进行求解能量交易问题。算例结果表明，所提场景驱动方法生成的场景集能更准确、更有效地描述新能源的不确定性，能兼顾系统的鲁棒性、经济性和隐私性，并实现每个主体公平合理的利益分配。 |
| 关键词: 多微网 分布鲁棒优化 合作收益 长短期记忆网络 条件生成对抗网络 |
| DOI：10.19783/j.cnki.pspc.231538 |
| 投稿时间：2023-12-04修订日期：2024-02-19 |
| 基金项目:国家自然科学基金项目资助(51607068) |
|
| Data-driven distributionally robust collaborative optimization operation strategy for a multi-microgrid based on LSTM-CGAN |
| LI Hong,HAN Yumeng |
| (State Key Laboratory of Alternate Electrical Power System with Renewable Energy Sources, North China
Electric Power University, Baoding 071003, China) |
| Abstract: |
| The high uncertainty of renewable energy poses significant challenges to the reliable and secure operation of multi-microgrids. To address this, a two-stage distributed robust coordination optimization model driven by data is proposed based on long short-term memory (LSTM) networks and conditional generative adversarial networks (CGAN). First, to accurately characterize the uncertainty of renewable energy, the model generates an initial set of renewable energy scenarios for the distributionally robust optimization (DRO) set using LSTM-CGAN and the K-means++ clustering algorithm. The CGAN model uses Wasserstein distance as the discriminator loss function, with the day-ahead renewable energy forecast as the conditional variable for the generative adversarial network, and employs LSTM to construct the generator and discriminator. Secondly, a benefit allocation method based on the contribution rate of multi-node energy trading is proposed to achieve fair distribution of cooperative benefits. In addition, to protect the privacy of individual entities and improve solution efficiency, an alternating direction multiplier method (ADMM) is proposed. This couples parallel computation of columns and constraint generation (C&CG) to solve the energy trading problem. Case study results demonstrate that the proposed approach, with the scenario-driven method for generating scenario sets, can accurately and effectively describe the uncertainty of renewable energy, while considering system robustness, economic efficiency, and privacy, and achieving fair and reasonable benefit allocation for each entity. |
| Key words: multi-microgrid distributionally robust optimization cooperative benefits long short-term memory network conditional generative adversarial networks |
