PDF(3532 KB)
Study on Photovoltaic Plant Site Selection Models Based on Geographic and Environmental Features
RAO Zhi, YANG Zaimin, YANG Xiongping, LI Jiaming, YANG Ping, WEI Zhichu
Electric Power Construction ›› 2025, Vol. 46 ›› Issue (7) : 163-174.
PDF(3532 KB)
PDF(3532 KB)
Study on Photovoltaic Plant Site Selection Models Based on Geographic and Environmental Features
[Objective] To improve the accuracy of global horizontal irradiance （GHI） prediction and completely explore its application value in solar energy resource assessments， such as photovoltaic site selection， this study proposes a GHI prediction model that integrates the temporal convolutional network （TCN） with the former architecture. [Methods] To address the presence of anomalies in the GHI data， raw data were first cleaned and preprocessed to eliminate outliers and ensure data quality. Then， the model leveraged the temporal feature extraction capability of the TCN to perform deep representation learning on preprocessed multisource input data， whereas the former network was employed to capture long-term dependencies. A high-precision prediction framework driven by multiple features was constructed by incorporating environmental and geographical parameters into the model input to enhance the overall performance. [Results] Comparative experiments conducted on real-world datasets from multiple regions demonstrated that the proposed TCN-Informer model outperformed mainstream prediction models in terms of mean absolute error， mean absolute percentage error， and root mean square error. Compared with the second-best performing informer model， the proposed model achieved reductions of 24.0%， 23.1%， and 28.5% in the mean absolute error， mean absolute percentage error， and root mean square error， respectively. [Conclusions] The TCN-Informer model exhibited significant advantages in terms of accuracy and robustness for GHI prediction， enabling a more effective capture of temporal variation patterns in solar irradiance. It has a strong engineering application potential and provides solid data support for solar resource evaluation and photovoltaic site planning.
global horizontal irradiance prediction / site selection of photovoltaic power stations / environmental features / geographic features / model of temporal convolutional network （TCN） / model of Informer
| [1] |
|
| [2] |
国家能源局. 国家能源局发布2022年全国电力工业统计数据[EB/OL].(2023-01-18)[2024-06-11]. https://www.nea.gov.cn/2023-01/18/c_1310691509.htm.
|
| [3] |
周满国, 黄艳国, 段锦锋. 基于GRU-RF模型的太阳辐照度短时预测[J]. 太阳能学报, 2022, 43(7): 166-173.
针对现有太阳辐照度短期预测方法的建模复杂、准确度低等问题,提出一种基于深度学习的GRU-RF动态权值组合预测方法。大气因素与太阳辐照度数据融合,将运算速度较快且模型复杂度较低的随机森林（RF）模型与带有时序记忆的门控循环单元（GRU）神经网络进行动态权值的加权集成,分别将地表接收到的太阳辐照度、近地层气温、相对湿度、近地层风速和相对气压等变化特征进行预测研究。通过几种模型对比分析,结果表明使用GRU-RF模型预测短时（9 h）太阳辐照度结果较好,运行速度较快,在不同时间间隔（5、10以及15 min）下能够很好地预测太阳辐照度数据。
Aiming at the problems of complex modeling and low accuracy of existing short-term solar irradiance prediction methods, a GRU-RF dynamic weight combination prediction method based on deep learning is proposed. The atmospheric factors and solar irradiance data are fused, and the random forest （RF） model with fast operation speed and low model complexity is integrated with the gated recurrent unit （GRU） neural network with time sequence memory for dynamic weight weighting. The variation characteristics of solar irradiance, surface air temperature, relative humidity, surface wind speed and relative pressure received by the surface are predicted respectively. Through the comparison and analysis of several models, the results show that the GRU-RF model can be used to predict the solar irradiance in short time （9 h） with faster running speed, and can be used to predict the solar irradiance data well in different time intervals （5,10 and 15 min）.
|
| [4] |
|
| [5] |
|
| [6] |
王飞, 李娜, 苏营, 等. 基于卫星遥感的辐照度时空关联映射与预测建模[J]. 太阳能学报, 2024, 45(3): 1-9.
|
| [7] |
|
| [8] |
贾东于, 李开明, 高晓清, 等. 基于FY-4A和机器学习的太阳辐照度超短期预测[J]. 太阳能学报, 2024, 45(4): 578-583.
|
| [9] |
陈龙, 张菁, 张昊立, 等. 基于VMD和射箭算法优化改进ELM的短期光伏发电预测[J]. 太阳能学报, 2023, 44(10): 135-141.
为了提高光伏发电预测的准确性,提出一种结合变分模态分解（VMD）、改进的射箭算法（AA）和改进的极限学习机（ELM）的短期光伏功率预测模型。首先,将光伏数据进行变分模态分解;然后,利用混合核函数改进极限学习机;之后,利用随机反向学习策略改进射箭算法;最后,通过改进的射箭算法对混合核极限学习机中的核参数寻优并建立预测模型。通过对澳大利亚DKA太阳能中心的数据进行验证,证明该文方法的准确性。
In order to improve the accuracy of photovoltaic power generation forecast, a short-term photovoltaic power generation forecast model based on variational mode decomposition（VMD）, improved archery algorithm（IAA） and improved extreme learning machine（ELM） was proposed. Firstly,decompose the photovoltaic data by variational modal decomposition algorithm Secondly,use hybrid kernel to improve extreme learning Machine,and then use the Random opposition-based learning to improve archery algorithm. Finally, the algorithm is used to optimize the kernel function parameters. The accuracy of this method is verified by the data of DKA solar energy center in Australia.
|
| [10] |
|
| [11] |
黄莉, 甘恒玉, 刘兴举, 等. 基于Transformer编码器的超短期光伏发电功率预测[J]. 智慧电力, 2024, 52(5): 16-22, 59.
|
| [12] |
董明, 李晓枫, 杨章, 等. 基于数据驱动的分布式光伏发电功率预测方法研究进展[J]. 电网与清洁能源, 2024, 40(1): 8-17, 28.
|
| [13] |
黄泽, 毕贵红, 谢旭, 等. 基于MBI-PBI-ResNet的超短期光伏功率预测[J]. 电力系统保护与控制, 2024, 52(2): 165-176.
|
| [14] |
黄杨珏, 张晓珂, 沈开程, 等. 基于智能混合预测策略的短期光伏功率预测[J]. 电网与清洁能源, 2023, 39(11): 111-119.
|
| [15] |
叶畅, 柳丹, 曹侃. 基于云图特征的超短期光伏发电功率预测模型[J]. 电网与清洁能源, 2023, 39(10): 70-79.
|
| [16] |
史加荣, 殷诏. 基于GRU-BLS的超短期光伏发电功率预测[J]. 智慧电力, 2023, 51(9): 38-45.
|
| [17] |
雷柯松, 吐松江·卡日, 伊力哈木·亚尔买买提, 等. 基于WGAN-GP和CNN-LSTM-Attention的短期光伏功率预测[J]. 电力系统保护与控制, 2023, 51(9): :108-118.
|
| [18] |
|
| [19] |
|
| [20] |
臧海祥, 张越, 程礼临, 等. 基于ICEEMDAN-LSTM和残差注意力的短期太阳辐照度预测[J]. 太阳能学报, 2023, 44(12): 175-181.
为提升短期太阳辐射预测的准确性,提出一种基于ICEEMDAN-LSTM和残差注意力的短期太阳辐照度预测方法。该方法利用改进的自适应噪声完备集合经验模态分解（ICEEMDAN）将原始辐射序列分解为多尺度模态分量,同时引入残差注意力机制对原始气象特征进行重构,然后利用长短期记忆网络分别提取两部分的时序特征,并融合所得特征输入至多层感知器,进行提前1小时的水平面总辐照度预测。实验结果表明,该方法能捕捉辐射序列的波动和突变,并考虑不同气象特征的重要程度,可有效提高短期太阳辐照度的预测精度。
In order to improve the accuracy of short-term solar radiation forecasting, a novel forecasting method based on ICEEMDAN-LSTM and residual attention is proposed. The original radiation sequence is first decomposed into multiple modal components by the improved complete ensemble empirical modal decomposition with adaptive noise. Residual attention is introduced to reconstruct the original meteorological features at the same time. Long short-term memory（LSTM） network is then utilized to extract temporal features of the two parts, respectively. After that, temporal features of each part are concatenated as inputs of a multi-layer perception, which can generate one-hour-ahead prediction results of global horizontal irradiance. Experimental results demonstrate that the proposed method can capture the fluctuations and abrupt changes of the irradiance series and is able to consider the importance of different meteorological features for the prediction task. The proposed method is proved to be effective in improving the prediction accuracy of short-term solar irradiance.
|
| [21] |
李庆生, 张裕, 龙家焕, 等. 基于IMFO-LSTM模型的光伏功率预测研究[J]. 电力科学与技术学报, 2024, 39(3): 199-206.
|
| [22] |
闫文杰, 徐志杰, 薛桂香, 等. 基于EMD-TCN的多能供热系统太阳辐照度预测模型[J]. 太阳能学报, 2023, 44(11): 182-188.
针对太阳辐照度的非平稳性和非线性影响多能供热系统运行效率和可靠性问题,该文提出一种基于经验模态分解 （EMD） 和时间卷积网络 （TCN） 的太阳辐照度混合预测模型EMD-TCN,更精准地从气象数据中提取太阳辐照度非线性和非平稳的隐含特征,获得更佳的预测精度。该研究利用逐时气象数据对所提出的EMD-TCN模型进行不同时间尺度的太阳辐照度预测实验,并与4种主流深度学习预测算法进行对比分析,结果表明该太阳辐照度预测模型具有更高的预测精度和泛化能力。
Aiming at the problem that the non-stationary and non-linear of solar irradiance affects the operation efficiency and reliability of multi-energy heating system, this paper proposes a hybrid forecasting model of solar irradiance based on empirical mode decomposition （EMD） and temporal convolutional network （TCN） named EMD-TCN, which can extract the hidden features of non-linear and non-stationary of solar irradiance from meteorological data more accurately, and obtain better prediction accuracy. The proposed EMD-TCN model is used to predict solar irradiance at different time scales using hourly meteorological data, and compared with four mainstream deep learning prediction algorithms. The results show that the proposed solar irradiance prediction model has higher prediction accuracy and generalization ability.
|
| [23] |
|
| [24] |
|
| [25] |
吴沂洋, 刘继春, 伍峻杭. 考虑不同区域气象与地理条件差异的广域光伏电站规划[J]. 电网技术, 2024, 48(9): 3653-3662.
|
| [26] |
王海金, 唐若笠, 周雨诗, 等. 基于ArcGIS与多因子模型的光伏电站选址评估[J]. 太阳能学报, 2023, 44(11): 120-130.
该文基于中国光伏资源的时空分布,研究并提出一种基于ArcGIS与多因子评价的光伏电站选址及发电潜力预测模型。利用ArcGIS软件与层次分析法,综合考虑多种影响大型光伏电站建设的因素,对不同地势和土地类型赋予不同的光伏利用系数,得到大型光伏电站建设适宜度、光伏潜力与光伏电站选址的多因子评价模型。基于该模型得出中国光伏发电潜力并给出光伏电站选址建议。评估结果表明,全国年光伏总发电量约为570.07×10<sup>6</sup> kWh,其中新疆为年光伏总发电量最高的省份。中国较为适宜建设光伏电站的地区主要为西北部地区,其中适宜区面积为1.08×10<sup>6 </sup>km<sup>2</sup>,最适宜区面积为2.10×10<sup>5</sup> km<sup>2</sup>,研究可结果将为中国光伏产业长期的规划与建设提供可靠的理论依据。
Based on the spatial and temporal distribution of photovoltaic resources in China, this paper studies and proposes a prediction model for the location of photovoltaic power stations and potential of photovoltaic power based on ArcGIS and multi-criteria evaluation. ArcGIS software and analytical hierarchy process were used to comprehensively consider a variety of factors affecting the construction of large-scale photovoltaic power stations, and different photovoltaic utilization coefficients were assigned to different terrain and land types, so as to obtain a multi-factor evaluation model of the suitability of large-scale photovoltaic power station construction, photovoltaic potential and photovoltaic power station location. Based on the model, the potential of photovoltaic power generation in various regions in China is obtained and the suggestions on the location of photovoltaic power stations are given. The evaluation results show that the total annual photovoltaic power generation in China is about 570.07×10<sup>6</sup> kWh, among which Xinjiang is the province with the highest total annual photovoltaic power generation. The most suitable areas for the construction of photovoltaic power stations in China are the northwest region, in which the suitable area is 1.08×10<sup>6</sup> km<sup>2</sup>, and the most suitable area is 2.10×10<sup>5</sup> km<sup>2</sup>. The conclusion will provide a reliable theoretical basis for the long-term planning and construction of China's photovoltaic industry.
|
| [27] |
尹德扬, 梅飞, 郑建勇, 等. 分布式光伏系统最优运维周期确定方法[J]. 电力自动化设备, 2022, 42(5): 135-141.
|
| [28] |
代义东, 陆之洋, 熊炜, 等. 基于CDAE和TCN/BLSTM模型的电能质量扰动分类方法[J]. 智慧电力, 2023, 51(12): 59-66.
|
| [29] |
陈海鹏, 李赫, 阚天洋, 等. 考虑风电时序特性的深度小波-时序卷积网短期风功率预测[J]. 电网技术, 2023, 47(4): 1653-1662.
|
| [30] |
杨迪, 王辉, 贺仁杰, 等. 基于改进经验模态分解和混合深度学习模型的风速预测[J]. 智慧电力, 2024, 52(1): 1-7.
|
| [31] |
赵洋, 王瀚墨, 康丽, 等. 基于时间卷积网络的短期电力负荷预测[J]. 电工技术学报, 2022, 37(5): 1242-1251.
|
| [32] |
|
| [33] |
|
| [34] |
|
| [35] |
|
| [36] |
Five statistical models for nowcasting solar irradiance are evaluated from different perspectives. The first four models are purely statistical ones: random walk, moving average, exponential smoothing and autoregressive integrated moving average. These models can be considered as benchmarks of different levels of complexity. The fifth model is a version of the two-state model, an applications suite for nowcasting solar irradiance developed by our team. The two-state model connects in an innovative manner an empirical estimator for clear-sky solar irradiance with a statistical predictor for the sunshine number, a binary indicator stating whether the sun shines or not. On the basis of different error metrics, the models' performances are analyzed from four perspectives: forecast accuracy, forecast precision, data series granularity and variability in data series. The study is conducted with high-quality radiometric data measured at a high frequency of four samples per minute on the Solar Platform of the West University of Timisoara, Romania. No model is ranked as the best, but the peculiarities that cause a model to perform better than others are discussed. By processing information about the atmospheric transmittance, the two-state model proves a slight advance in the forecast accuracy and a notable performance in the forecast precision. (C) 2019 Elsevier Ltd.
|
| [37] |
|
| [38] |
Solargis. 世界太阳能资源图[EB/OL]. [2024-05-21]. https://www.nea.gov.cn/2023-01/18/c_1310691509.htm.
|
| [39] |
|
| [40] |
|
| [41] |
|
/
| 〈 |
|
〉 |
