-
摘要: 對光熱電站的出力進行短期預測，可以有效應對太陽能隨機性和波動性帶來的影響，為電網調度做好準備. 該文以青海某光熱電站為例，首先使用模糊C均值聚類算法對預處理后的實驗數據進行分類，然后通過分析不同聚類類型下出力和氣象數據中各因子間的關聯程度，充分挖掘出數據間的關系，確定不同類型預測模型的輸入變量，進而構建出不同類別下的長短期記憶神經網絡預測模型. 結果表明，與傳統長短期記憶神經網絡模型、BP神經網絡模型、支持向量機模型和隨機森林模型的預測結果相比，基于模糊C均值聚類的長短期記憶預測模型效果良好，大幅減少了預測誤差，驗證了該預測模型的有效性.Abstract: In China, the development of concentrated solar power has gained momentum to harness the country’s abundant solar energy resources. Predicting the short-term power generation capacity of concentrated solar power stations is crucial for mitigating the impact of the randomness and volatility of solar energy and facilitating effective grid dispatching. To solve this problem, this study presents a short-term concentrated solar power prediction combination model based on fuzzy C-means clustering. Fuzzy C-means clustering is an objective function–based fuzzy clustering algorithm that yields more flexible clustering results by incorporating fuzzy theory. Using a concentrated solar power station in Qinghai as an example, this study employs cubic spline interpolation to preprocess experimental data and divide the data into training and testing sets. Subsequently, a fuzzy c-means clustering algorithm is used to classify the preprocessed data. Different forecast scenarios are established, enhancing the precision of the prediction model. The relationship between the data is fully explored by calculating the Pearson correlation coefficient between meteorological factors and each factor in the output data under different types. Based on the degree of correlation between the factors, the input variables of different prediction submodels are determined. The influence of various meteorological factors on the prediction model under different scenarios was fully considered. Additionally, the neural network prediction model of long short-term memory in different scenarios is constructed. The test set is used to evaluate the accuracy of the combined model, and the membership degree of each sample group is determined by calculating their distance from different cluster centers to divide the test data and classify them into different scenarios. Consequently, the combined prediction model is tested. To fully confirm the feasibility and accuracy of the combined model, the test results are compared with the prediction results of the traditional long short-term memory neural network model, BP neural network model, support vector machines, and random forest. Results demonstrate that the long short-term memory neural network prediction model based on fuzzy C-means clustering has a good effect, which considerably reduces prediction error and closely aligns with actual output compared to the other two prediction models. Therefore, this model can provide a reference for power grid dispatching, effectively capturing the influence between weather factors and concentrated solar power and proving the applicability and effectiveness of the combined prediction model in different scenarios.
-
表 1 預測模型參數信息
Table 1. Parameters of the prediction models
Type Input dimension Batch_size Epochs Optimizer Ⅰ 20×3 32 50 Adadelta Ⅱ 20×5 128 100 Adadelta Ⅲ 20×4 32 50 Adadelta 
下載: 導出CSV
表 2 對比模型參數信息
Table 2. Parameters of the compared models
Model Parameters Batch_size Epochs Optimizer LSTM 16 200 Adam BP 32 20 sgd Model Parameters Penalty coefficient Kernel SVM 1 radial basis function Model Parameters n_estimators max_depth max_features RF 300 5 sqrt
下載: 導出CSV
表 3 五種模型的評價指標
Table 3. Evaluation index of the five models
Model RMSE MAE FCM–LSTM 1.8538 1.1638 LSTM 2.6864 1.5539 BP 2.8978 1.7713 SVM 3.3045 1.7898 RF 3.1063 1.9091
下載: 導出CSV
啪啪啪视频<noscript id="y4y0w"><source id="y4y0w"></source></noscript> <table id="y4y0w"><option id="y4y0w"></option></table> <li id="y4y0w"></li> <noscript id="y4y0w"></noscript> <noscript id="y4y0w"><kbd id="y4y0w"></kbd></noscript> <noscript id="y4y0w"><source id="y4y0w"></source></noscript> <menu id="y4y0w"></menu> <table id="y4y0w"><rt id="y4y0w"></rt></table> -
參考文獻
[1] Li Y Y. Distribution of solar energy resources in China. Energy China, 1977(1): 47李益言. 我國太陽能資源的分布. 中國能源, 1977(1):47 [2] Wang K, Huang J. Domestic and abroad research status and prospects of solar energy resource evaluation methods [J/OL]. Clim Change Res (2022-12-01) [2023-02-24]. http://kns.cnki.net/kcms/detail/11.5368.P.20221130.0847.001.html王科, 黃晶. 國內外太陽能資源評估方法研究現狀和展望[J/OL]. 氣候變化研究進展 (2022-12-01) [2023-02-24]. http://kns.cnki.net/kcms/detail/11.5368.P.20221130.0847.001.html [3] Li H W. Research on the current situation of solar thermal power generation technology and the existing problems of key equipment. Electron Test, 2020(2): 131 doi: 10.3969/j.issn.1000-8519.2020.02.051李煥偉. 太陽能光熱發電技術現狀極其關鍵設備存在問題探究. 電子測試, 2020(2):131 doi: 10.3969/j.issn.1000-8519.2020.02.051 [4] Huang Y R, Hou Y Y, Gao Z H. The development status and prospect analysis of international concentrating solar power industry. Sci Technol Ind, 2014, 14(9): 54 doi: 10.3969/j.issn.1671-1807.2014.09.012黃裕榮, 侯元元, 高子涵. 國際太陽能光熱發電產業發展現狀及前景分析. 科技和產業, 2014, 14(9):54 doi: 10.3969/j.issn.1671-1807.2014.09.012 [5] Niu Z Y, Li L J. Status and analysis of CSP demonstration projects. Energy Energy Conserv, 2020(12): 76 doi: 10.16643/j.cnki.14-1360/td.2020.12.031牛志愿, 李麗君. 太陽能熱發電示范項目現狀及分析. 能源與節能, 2020(12):76 doi: 10.16643/j.cnki.14-1360/td.2020.12.031 [6] Li J J. Research on Output Prediction Model of Concentrating Solar Power Station Based on DBN–DNN Algorithm [Dissertation]. Lanzhou: Lanzhou University of Technology, 2020李錦鍵. 基于DBN–DNN算法的光熱儲能電站出力預測模型研究[學位論文]. 蘭州:蘭州理工大學, 2020 [7] Zhang H B, Yang M Y. Ultra-short-term forecasting for photovoltaic power output based on least square support vector machine. Mod Electr Power, 2015, 32(1): 70 doi: 10.3969/j.issn.1007-2322.2015.01.012張華彬, 楊明玉. 基于最小二乘支持向量機的光伏出力超短期預測. 現代電力, 2015, 32(1):70 doi: 10.3969/j.issn.1007-2322.2015.01.012 [8] Li G M, Zheng L N, Fan W, et al. Research on ultra-short term output forecasting of PV station based on BP neural network. Chin J Power Sources, 2021, 45(8): 1052 doi: 10.3969/j.issn.1002-087X.2021.08.025李光明, 鄭麗娜, 范威, 等. 基于神經網絡的光伏電站出力超短期預測研究. 電源技術, 2021, 45(8):1052 doi: 10.3969/j.issn.1002-087X.2021.08.025 [9] Jiang T L, Xu Z M, Wang G. Research on short-term power forecasting model of solar-thermal power generation based on mobile ad hoc network. Turbine Technol, 2019, 61(2): 151 doi: 10.3969/j.issn.1001-5884.2019.02.018姜鐵騮, 徐志明, 王剛. 基于移動Ad Hoc網絡的太陽能光熱發電短期功率預測模型研究. 汽輪機技術, 2019, 61(2):151 doi: 10.3969/j.issn.1001-5884.2019.02.018 [10] Wang Y, Li G J, Yue W H. Prediction model of photo-thermal power generation based on G (1, 1) optimization. MATEC Web Conf, 2018, 246: 02057 doi: 10.1051/matecconf/201824602057 [11] Song S J, Li B H. Short-term forecasting method of photovoltaic power based on LSTM. Renew Energy Resour, 2021, 39(5): 594 doi: 10.3969/j.issn.1671-5292.2021.05.005宋紹劍, 李博涵. 基于LSTM網絡的光伏發電功率短期預測方法的研究. 可再生能源, 2021, 39(5):594 doi: 10.3969/j.issn.1671-5292.2021.05.005 [12] Wang C Q, Chen X D, Cao H W, et al. Short term photovoltaic power prediction. J Liaoning Tech Univ (Nat Sci), 2023, 42(1): 99王琛淇, 陳曉東, 曹瀚文, 等. 短期光伏發電系統功率預測. 遼寧工程技術大學學報(自然科學版), 2023, 42(1):99 [13] Li S S, Ma Z C, Zhang X Y, et al. Research on thermal power prediction of concentrating solar power station based on deep learning. Ind Instrum Autom, 2021(6): 58 doi: 10.3969/j.issn.1000-0682.2021.06.013李嵩山, 馬志程, 張曉英, 等. 基于深度學習的聚光太陽能電站熱功率預測研究. 工業儀表與自動化裝置, 2021(6):58 doi: 10.3969/j.issn.1000-0682.2021.06.013 [14] Li Q, Gao C Y, Hu C X, et al. Short-term photovoltaic power generation prediction based onLong-short-term memory network and attention mechanism. Electr Autom, 2020, 42(5): 19 doi: 10.3969/j.issn.1000-3886.2020.05.006李清, 高春燕, 胡長驍, 等. 基于長短期記憶網絡與注意力機制的短期光伏發電預測. 電氣自動化, 2020, 42(5):19 doi: 10.3969/j.issn.1000-3886.2020.05.006 [15] Chen Z, Che S Y. Photovoltaic output forecast based on cloud models. Acta Energiae Solaris Sin, 2019, 40(11): 3054陳中, 車松陽. 基于云變換的光伏出力預測模型. 太陽能學報, 2019, 40(11):3054 [16] Huang D M, Zhuang X K, Hu A D, et al. Short-term load forecasting based on similar-day selection with GRA-K-means. Electr Power Constr, 2021, 42(7): 110 doi: 10.12204/j.issn.1000-7229.2021.07.013黃冬梅, 莊興科, 胡安鐸, 等. 基于灰色關聯分析和K均值聚類的短期負荷預測. 電力建設, 2021, 42(7):110 doi: 10.12204/j.issn.1000-7229.2021.07.013 [17] Chen G R, Wang B, Cao Z J, et al. Research on wind power prediction based on cluster analysis and optimized neural network. Electr Autom, 2020, 42(3): 24 doi: 10.3969/j.issn.1000-3886.2020.03.008陳桂儒, 王冰, 曹智杰, 等. 基于聚類分析和優化神經網絡的風電功率預測研究. 電氣自動化, 2020, 42(3):24 doi: 10.3969/j.issn.1000-3886.2020.03.008 [18] Yu Y, Chen G, Yu J L, et al. Grid-connected power forecasting of concentrating solar power plants based on clustering and long short-term memory neural network. Therm Power Gener, 2021, 50(9): 128 doi: 10.19666/j.rlfd.202101048余洋, 陳庚, 余佳磊, 等. 基于聚類和長短期記憶神經網絡的光熱電站并網電力預測. 熱力發電, 2021, 50(9):128 doi: 10.19666/j.rlfd.202101048 [19] Liu X L, Huang C, Wang L, et al. Clustering and LSTM-based robust day-ahead hourly forecasting of photovoltaic power. Comput Technol Dev, 2023, 33(3): 120 doi: 10.3969/j.issn.1673-629X.2023.03.018劉興霖, 黃超, 王龍, 等. 基于聚類和LSTM的光伏功率日前逐時魯棒預測. 計算機技術與發展, 2023, 33(3):120 doi: 10.3969/j.issn.1673-629X.2023.03.018 [20] Wen T T, Guo Y X, Dong S R, et al. Assessment of CRU, ERA5, CMFD grid precipitation data for the Tibetan Plateau from 1979 to 2017. Arid Zone Res, 2022, 39(3): 684 doi: 10.13866/j.azr.2022.03.03溫婷婷, 郭英香, 董少睿, 等. 1979—2017年CRU、ERA5、CMFD格點降水數據在青藏高原適用性評估. 干旱區研究, 2022, 39(3):684 doi: 10.13866/j.azr.2022.03.03 [21] Zhang J B, Shen R P, Shi C X, et al. Evaluation and comparison of downward solar radiation from new generation atmospheric reanalysis ERA5 across China’s mainland. J Geo Inf Sci, 2021, 23(12): 2261 doi: 10.12082/dqxxkx.2021.180357張俊兵, 沈潤平, 師春香, 等. 中國大陸地區ERA5下行短波輻射數據適用性評估與對比. 地球信息科學學報, 2021, 23(12):2261 doi: 10.12082/dqxxkx.2021.180357 [22] Liu T T, Zhu X F, Guo R, et al. Applicability of ERA5 reanalysis of precipitation data in China. Arid Land Geogr, 2022, 45(1): 66 doi: 10.12118/j.issn.10006060.2021.132劉婷婷, 朱秀芳, 郭銳, 等. ERA5再分析降水數據在中國的適用性分析. 干旱區地理, 2022, 45(1):66 doi: 10.12118/j.issn.10006060.2021.132 [23] Zhu J, Yuan H Z. Applicability of ERA reanalysis data of land surface temperature in Zhejiang Province. Meteorol Sci Technol, 2019, 47(2): 289 doi: 10.19517/j.1671-6345.20180171朱景, 袁慧珍. ERA再分析陸面溫度資料在浙江省的適用性. 氣象科技, 2019, 47(2):289 doi: 10.19517/j.1671-6345.20180171 [24] Meng X G, Guo J J, Han Y Q. Preliminarily assessment of ERA5 reanalysis data. J Mar Meteorol, 2018, 38(1): 91孟憲貴, 郭俊建, 韓永清. ERA5再分析數據適用性初步評估. 海洋氣象學報, 2018, 38(1):91 [25] Dunn J C. A fuzzy relative of the ISODATA process and its use in detecting compact well-separated clusters. J Cybern, 1973, 3(3): 32 doi: 10.1080/01969727308546046 [26] Lü W J, Fang Y F, Cheng Z. Prediction of day-ahead photovoltaic output based on FCM-WS-CNN. Power Syst Technol, 2022, 46(1): 231 doi: 10.13335/j.1000-3673.pst.2020.2246呂偉杰, 方一帆, 程澤. 基于模糊C均值聚類和樣本加權卷積神經網絡的日前光伏出力預測研究. 電網技術, 2022, 46(1):231 doi: 10.13335/j.1000-3673.pst.2020.2246 [27] Hochreiter S, Schmidhuber J. Long short-term memory. Neural Comput, 1997, 9(8): 1735 doi: 10.1162/neco.1997.9.8.1735 [28] Sahin S O, Kozat S S. Nonuniformly sampled data processing using LSTM networks. IEEE Trans Neural Netw Learn Syst, 2019, 30(5): 1452 doi: 10.1109/TNNLS.2018.2869822 -
點擊查看大圖
圖(7) / 表(3)
計量
- 文章訪問數: 159
- HTML全文瀏覽量: 38
- PDF下載量: 22
- 被引次數: 0
