储能科学与技术 ›› 2024, Vol. 13 ›› Issue (5): 1731-1740.doi: 10.19799/j.cnki.2095-4239.2023.0828

• 储能技术经济性分析 • 上一篇    下一篇

基于GRU算法的弃电量预测及电-氢混合储能系统的运行优化

何婷1,2(), 乔俊强1,2(), 吴国栋3   

  1. 1.甘肃自然能源研究所
    2.甘肃太阳能利用重点实验室
    3.国网甘肃省电力公司,甘肃 兰州 730046
  • 收稿日期:2023-11-16 修回日期:2024-01-07 出版日期:2024-05-28 发布日期:2024-05-28
  • 通讯作者: 乔俊强 E-mail:heting@gneri.com.cn;qiao@gsas.ac.cn
  • 作者简介:何婷(1987—),女,硕士研究生,助理研究员,主要从事光伏发电及并网技术研究,E-mail:heting@gneri.com.cn
  • 基金资助:
    兰州市人才创新创业项目(2021-RC-78);甘肃省科技计划项目(22YF7GA064)

Curtailed power forecasting based on GRU and operation optimization of electric-hydrogen hybrid energy storage system

Ting HE1,2(), Junqiang QIAO1,2(), Guodong WU3   

  1. 1.Gansu Natural Energy Research Institute
    2.Gansu Key Laboratory of Solar Energy Utilization
    3.State Grid Gansu Electric Power Company, Lanzhou 730046, Gansu, China
  • Received:2023-11-16 Revised:2024-01-07 Online:2024-05-28 Published:2024-05-28
  • Contact: Junqiang QIAO E-mail:heting@gneri.com.cn;qiao@gsas.ac.cn

摘要:

风/光等可再生能源的高渗透接入将增强电源侧输出随机性与波动性,导致弃风、弃光现象时常存在,将弃风/光量制氢是解决可再生能源深度消纳的有效途径。本文提出基于Adam优化的门控循环单元神经网络算法(GRU)对风/光弃电量进行预测,利用拉丁超立方和同步回代削减算法生成典型风/光发电不确定性场景。以系统建设成本最小化和运营成本最小化为目标,构建双层目标函数,第1层优化碱性电解槽(AWE)和电池储能系统(BESS)的容量配置,第2层保证系统在生成场景中运行最佳,同时引入惩罚项,以最大限度利用风/光弃电。以西北某地区风/光弃电量数据为例,采用不同结构的神经网络算法对弃风/光量进行预测,通过比较不同算法的均方根误差,验证了所提算法的准确性。最后对3种不同储能方案进行优化运行分析,研究结果表明,仅采用BESS系统年利润为负数且年投资成本最高,采用BESS-AWE混合储能系统比仅采用AWE系统年投资成本增加15.66%,但年利润增加了255%,弃电利用率为92%,有效提高系统经济性。

关键词: 可再生能源弃电量, 门控循环单元神经网络, 电-氢混合储能, 双层目标函数

Abstract:

The high penetration of renewable energy sources such as wind and solar will enhance the randomness and volatility of power output, leading to frequent occurrences of wind/solar curtailment. Hydrogen production from wind/solar curtailment is an effective means to handle the deep consumption of renewable energy. In this study, we propose a gated recurrent unit algorithm based on Adam optimization for predicting wind/solar curtailment using Latin hypercube and synchronous backpropagation reduction algorithms to generate typical wind/light generation uncertainty scenarios. A double-layer objective function is constructed to minimize system construction and operating costs. The first layer determines the capacity configuration of the alkaline electrolysis cell (AWE) and battery energy storage system (BESS), and the second layer ensures the system's optimal operation in the generated scenario. In addition, a penalty term is introduced to maximize the abandoned power use. Considering the wind/solar curtailment data from a certain region in northwest China as an example, different structures of neural network algorithms were used to predict the amount of wind/light curtailment. The accuracy of the proposed algorithm was verified by comparing the root mean square errors of the various algorithms. Finally, an optimization configuration analysis was performed on three energy storage schemes. The results showed that using only the BESS system resulted in negative annual profits and the highest annual investment cost. Using the BESS-AWE hybrid energy storage system increased the annual investment cost by 15.66% compared with using only the AWE system, but the annual profit increased by 255%, and the abandoned power usage rate was 92%, effectively improving the system's economy.

Key words: renewable energy curtailment, gated recurrent unit, electric hydrogen hybrid energy storage, two-stage objective function

中图分类号: