Integrated Intelligent Energy

Study on dynamic surrogate model for MPPT of PV systems

ZHANG Xiaoshun, TAN Tian, MENG Die, ZHANG Guiyuan, FENG Yongkun

2021, 43(8): 1-10. doi:10.3969/j.issn.1674-1951.2021.08.001

Asbtract ( 151 )

HTML ( 5)

PDF (1896KB) ( 338 )

Figures and Tables | References | Related Articles | Metrics

Under partial shading condition（PSC）,there will be multiple peaks on the power-voltage output characteristic curves of photovoltaic（PV）systems.The traditional maximum power point tracking（MPPT） algorithm is no longer applicable for solving the problem since it is easy to fall into local optimal solution and of poor stability.In order to realize MPPT under PSC,a dynamic surrogate model-based optimization（DSMO） method for a PV system is designed.To avoid aimless search,by taking real-time data of the PV system,the radial basis function（RBF） network is adopted to construct the dynamic surrogate model of input/output features.Based on the dynamic surrogate model,greedy search is used to accelerate the convergence.The practicability and superiority of the method are evaluated by tests under three conditions,constant temperature and constant light start-up experiment,constant temperature and step-changed light test,and variable temperature and variable light test.Compared with ant colony algorithm（ASO）,gray wolf optimizer（GWO）,perturb and observation method（P&O） and particle swarm optimization algorithm（PSO）,the DSMO method proposed can facilitate PV systems to generate more energy and smaller power fluctuation quickly and stably under PSC.

Research on PV power grid connection stability based on decoupled active disturbance rejection control

LIU Huiqiang, MU Teng, XING Huadong, WU Haiyan, LIU Jianqiang, LEI Ke, GUO Qi

2021, 43(8): 11-19. doi:10.3969/j.issn.1674-1951.2021.08.002

Asbtract ( 132 )

HTML ( 3)

PDF (1231KB) ( 216 )

Figures and Tables | References | Related Articles | Metrics

To set up modern energy system in the context of carbon neutrality,power grid should consume more renewable energy,for example,solar energy.In order to improve the dynamic frequency support and transient voltage stability of the power grid connection system congregated distributed PV power stations,static synchronous compensators（STATCOM-BESS）have been widely deployed in PV stations due to their fast response.The damping controller with STATCOM-BESS can suppress the power oscillation of the system.In this context,to improve the stability of the grid connection system,STATCOM-BESS with particle swarm optimized active disturbance rejection controller（PSO-ADRC）for PV power stations is studied.Firstly,the ADRC is analyzed.Then,a STATCOM-BESS integrated with dynamic stability strategy is deduced,and a STATCOM-BESS with PSO-ADRC is designed,whose damping effect is verified by a simulation test.

Wind turbine blades icing detection with multi-parameter models based on AdaBoost algorithm

FAN Daqian, LIU Bosong, GUO Peng

2021, 43(8): 20-26. doi:10.3969/j.issn.1674-1951.2021.08.003

Asbtract ( 127 )

HTML ( 4)

PDF (4438KB) ( 245 )

Figures and Tables | References | Related Articles | Metrics

Blade ice accretion is a factor that affects the safe operation of wind turbines in high-altitude and high-humidity areas of China in winter. Early detection of ice accretion and making adjustment on operation mode timely can guarantee the safety of wind turbines. The effects of ice accretion on the operation performance and parameters were analyzed thoroughly, and power, rotor blade speed and ambient temperature were taken as variables to monitor blade icing. Models of power and rotor blade speed were constructed by AdaBoost algorithm, and the prediction residuals of the two models were made by exponentially weighted moving average （EWMA）, in order to detect the abnormalities of power and rotor speed. If both abnormal power and rotor speed are detected and ambient temperature drops below 0 ℃ simultaneously, the blade icing alarm will be triggered. The effectiveness of the method has been proved by the icing data of a wind farm in Kunming.

A path planning strategy based on ant colony algorithm for series-connected battery packs

CHEN Yang, CHENG Lefeng, ZOU Tao

2021, 43(8): 27-32. doi:10.3969/j.issn.1674-1951.2021.08.004

Asbtract ( 206 )

HTML ( 3)

PDF (1728KB) ( 238 )

Figures and Tables | References | Related Articles | Metrics

To pursue carbon neutrality and carbon peaking, the scale of electric vehicles and electric energy storage is expanding. With the ever-growing demand for efficient and fast equalization of Li-ion batteries, the design, especially the path planning, of equalization circuits for Li-ion batteries becomes crucial. An ant colony algorithm-based path planning strategy for series-connected battery packs is proposed. First, a graph model is used to represent the equalization paths between different battery units. Then, the optimal equalization efficiency and speed models are established,and are solved by an ant colony algorithm,a practical heuristic swarm intelligence algorithm. Finally, taking an equalization system with 13 series-connected batteries as an example,the effectiveness of the proposed path planning strategy is verified.

Study on smart distribution network operation and management methods in the context of big data

LI Wei, YAN Chuan, SHENG Qingbo, WANG Jianping, WANG Xiaodong, TAN Jia

2021, 43(8): 33-40. doi:10.3969/j.issn.1674-1951.2021.08.005

Asbtract ( 117 )

HTML ( 3)

PDF (976KB) ( 216 )

Figures and Tables | References | Related Articles | Metrics

Smart distribution network has developed new characteristics during its extension,such as a large quantity of distributed power supplies and supports for users demand interaction. Thus,the operation and management of smart distribution network is more complex.Big data technology is booming and provides technical support for improving the operation and management level.The application value of big data technology in smart distribution network is described. Then,four potential application scenarios of big data in smart distribution network,data preprocessing,transmission loss management,state estimation and health evaluation, are systematically analyzed.The network security problems and the corresponding protection measures of smart distribution network in the context of big data are expounded. Combining the problems in practical cases with feasible research areas provides an idea for building a smart distribution network operation management system based on big data technology.

Intelligent evaluation of cable harmonic loss based on CSO-BP neural network

CHEN De, MENG Anbo, CAI Yongfeng

2021, 43(8): 41-47. doi:10.3969/j.issn.1674-1951.2021.08.006

Asbtract ( 167 )

HTML ( 2)

PDF (1054KB) ( 181 )

Figures and Tables | References | Related Articles | Metrics

Researches on the harmonic loss of cable lines at home and abroad are mainly based on electromagnetic physical analysis. The correction of equivalent parameters mostly relies on empirical formulas, and the accuracy is inadequate. In order to accurately evaluate the cable harmonic loss, an intelligent loss evaluation model based on crisscross optimization algorithm optimized-back propagation （CSO-BP） neural network is proposed. Generally, cable lines under the influence of harmonics are of various harmonic orders, different proportions of varied orders and multiple influencing factors on training samples. In order to overcome the shortages of the traditional BP algorithm such as slow convergence and being easy to fall into local optimum, BP neural network is optimized by CSO algorithm which can search better. After the optimization, an intelligent evaluation model for cable harmonic loss based on CSO-BP neural network is obtained. The values calculated by this model, traditional BP model and physical formulas are compared. The simulation results show that the cable harmonic loss calculated by CSO-BP neural network based intelligent evaluation model is closer to the actual value. The model is of accuracy and stability.

Research on on-line fault diagnosis and treatment of power plant equipment based on KPCA

JIA Zhijun, BAI Delong, SONG Yanjie, WANG Jianfei, LI Chunxin

2021, 43(8): 48-53. doi:10.3969/j.issn.1674-1951.2021.08.007

Asbtract ( 159 )

HTML ( 7)

PDF (1260KB) ( 203 )

Figures and Tables | References | Related Articles | Metrics

Taking the condensate water system of a 600 MW subcritical thermal power unit as the research object,in order to realize fault diagnosis and automatic accident treatment based on machine learning algorithm and expert knowledge,the sample set suitable for data analysis and model training is established by selecting and cleaning the historical operation data,and the kernel principal component analysis algorithm is used to build the early warning model of condensate pump operation characteristics.The model is used to warn the deviation of the operating parameters of the condensate pump from the normal value,and the early warning results and related parameters are logically integrated as the criterion for equipment fault diagnosis.Finally,the criterion is used as the trigger condition for automatic accident treatment,and the whole process of automatic control of the early warning,fault diagnosis and automatic accident treatment of the condensate pump is realized.The results show that the reconstruction accuracy of the model for the key parameters of the condensate pump is greater than 95 %,which can accurately diagnose the abnormal intake and output of the condensate pump,improve the automatic control and intelligent level of the condensate system,and have practical engineering application value.

Power consumption prediction based on multi-parameter cost-sensitive coefficient learning and data-driven model

SHI Jie, ZHANG Anqin

2021, 43(8): 54-60. doi:10.3969/j.issn.1674-1951.2021.08.008

Asbtract ( 136 )

HTML ( 3)

PDF (867KB) ( 226 )

Figures and Tables | References | Related Articles | Metrics

Energy conservation and emission reduction in enterprises are the frontier issues and research hot-spots in the way of China´s development. Comprehensive analysis and evaluation on the current situation of power consumption in enterprises is the premise and foundation for energy-saving transformations or energy conservation designs. Having described the common forecasting methods for electricity consumption, the shortcomings of constructing a prediction model taking Classification and Regression Tree （CART）algorithm as the weak learner are analyzed. To deal with the deficiency of the traditional AdaBoost algorithm focusing on the minimum prediction error rate only, an improved AdaBoost algorithm based on multi-parameter cost-sensitive coefficient learning is studied and proposed based on the essence of the algorithm. A regression prediction model constructed based on the improved AdaBoost algorithm can make short-term power consumption prediction according to real data, which verifies the improvement of the model´s performance.

A positioning method for pressure plates of automatic relay protection devices based on Hough transform

HUANG Weihao, LU Yongyin, LUO Qifeng, FAN Dehe, YUAN Tuolai, YAN Chao

2021, 43(8): 61-66. doi:10.3969/j.issn.1674-1951.2021.08.009

Asbtract ( 98 )

HTML ( 2)

PDF (8748KB) ( 230 )

Figures and Tables | References | Related Articles | Metrics

The current intelligent image recognition method applied for pressure plates of automatic relay protection devices has problems such as large deviations in switching status recognition for pressure plates and inaccurate identification of pressure plates' switching statuses due to the interference of redundant information in images.We propose a positioning method for pressure plates of automatic relay protection devices based on image pre-processing technology and Hough transform.By processing the original image of the pressure plate with image denoising and edge detection,the method can make line detection on the processed pressure plate images based on improved Hough transform,realizing the positioning for pressure plates of automatic relay protection devices.The proposed method reduces the amount of data for the subsequent intelligent image recognition by removing the redundant and useless information away from the original image. The test results show that the method can locate the pressure plate in the image.

Data-driven based research on anomaly detection for high-pressure heaters in thermal power units

HAN Xudong

2021, 43(8): 67-73. doi:10.3969/j.issn.1674-1951.2021.08.010

Asbtract ( 148 )

HTML ( 6)

PDF (3311KB) ( 299 )

Figures and Tables | References | Related Articles | Metrics

To ensure the safe and economic operation of high-pressure heaters in thermal power units,data-driven research was made on anomaly detection for high-pressure heaters.The research took the performance and state of the high-pressure heater into consideration,and chose five parameters including upper end difference,lower end difference,temperature rise,water level and normal opening of drain valves as the characteristic parameters to establish a model for anomaly detection.Then,driven by the real historical operation data under steady-state condition,Principal Component Analysis was used to solve the "Curse of Dimensionality" in high-dimensional data analysis,and Mahalanobis distance and PauTa criterion were taken to determine the parameters' threshold.This method can realize the anomaly detection for high-pressure heating system.Finally,verified by a group of real historical operation data,the model can capture the early abnormalities of the high-pressure heating system effectively.By providing effective abnormal information for high-pressure heaters in thermal power units,this method can ensure the safe and economic operation of the units.

Optimal power flow calculation of power grid based on reinforcement learning and crisscross PSO algorithm particle swarm optimization

MENG Anbo, WANG Peng, DING Weifeng, CHEN Shun, LIANG Ruduo, ZHANG Zheng

2021, 43(8): 74-82. doi:10.3969/j.issn.1674-1951.2021.08.011

Asbtract ( 247 )

HTML ( 2)

PDF (1004KB) ( 384 )

Figures and Tables | References | Related Articles | Metrics

To solve the optimal power flow in power systems,new particle swarm optimization（PSO）algorithm based on Q learning and crisscross search is proposed.The improved algorithm introduces crossover operator into PSO mode to enhance the global convergence ability.At the same time,introducing the exploration mode of Q learning into the improved algorithm makes the algorithm explore in the known solution space,so as to better balance the relationship between exploration and utilization.In order to solve the dimension disaster of Q learning algorithm,the method of state-action chain is used.Simulation results of IEEE57 and IEEE118 node systems show that the proposed algorithm can enhance the global convergence of the traditional PSO algorithm,and effectively solve large-scale optimal power flow problems.

Table of Content