BP Neural Network Method Research Articles

Study on annual rainfall forecast of Xishan station in Wenzhou City

Wenzhou city is Located in the southeast of Zhejiang Province, it belongs to subtropical climate. The main factors that affect the climate in Wenzhou include strong solar radiation and the regulation of marine water body. The annual variation of rainfall in Wenzhou City is very large. The disastrous weather in Wenzhou mainly includes typhoon, rain and drought. The ways of Typhoon Affecting Wenzhou include direct landing and adjacent area landing. Direct landing is more harmful. Rainfall is the most important factor affecting flood and drought disasters in Wenzhou City, so the analysis of rainfall can provide a reference for flood control and drought relief work in Wenzhou City. In this paper, BP neural network method is used to analyze and forecast the rainfall of Xishan station in the main urban area of Wenzhou City. The results show that the precipitation in Wenzhou City will change greatly in the next ten years. Flood disaster may occur in 2022 and 2028, and drought may occur in 2023.

Research on the Prediction Model of the Friction Coefficient of Asphalt Pavement Based on Tire‐Pavement Coupling

The correlations between pavement texture and tire pressure with the actual tire‐road contact area were first investigated according to the tire‐road static contact characteristics; on this basis, the influence mechanisms of speed and pavement texture on the pavement friction coefficient were systematically explored from the angle of tire‐road coupling system dynamics via the self‐developed dynamic testing system of tire‐pavement friction. By integrating the above influence factors, the BP neural network method was applied to the regression of the prediction model for the asphalt pavement friction coefficient. Through the comparison between the model measured value and estimated value, their correlation coefficient R2 reached 0.73, indicating that this model is of satisfactory prediction accuracy and applicable to the antiskid design of asphalt pavement.

Research on DBN-based Evaluation of Distribution Network Reliability

In order to accurately and efficiently analyze the reliability of distribution network, this paper proposes a method of analyzing the reliability of distribution network based on a deep belief network. The Deep Belief Network (DBN) is composed of limiting Boltzmann machine layer-by-layer stacking. It has a strong advantage of automatic feature extraction, which overcomes the shortcomings of traditional neural networks in extracting data features. The entire training process of DBN can be roughly divided into two stages: pre-training and fine-tuning.First of all, the pre-training of the DBN model is realized by training the Restricted Boltzmann Machine (RBM) layer by layer, then the BP algorithm is used for reverse fine-tuning to complete the training process of the entire network. finally, the reliability analysis of distribution network is performed by the trained DBN. Compared with the BP neural network method and the traditional Monte Carlo simulation method, it is verified that the proposed model of distribution network reliability analysis has high accuracy.

New stability forecasting model for goaf slope based on the AHP–TOPSIS theory

The goaf slopes commonly found in China have serious stability problems. These areas experience several landslides yearly, which results in casualties, property damage, and economic loss. Therefore, investigating the accurate identification and prediction of the stability state of goaf slopes is important to prevent disasters. In this study, a new prediction model of goaf slope stability is established based on the introduced theories of the analytic hierarchy process (AHP) and technique for order preference by similarity to ideal solution (TOPSIS). The factors affecting goaf slope stability are classified, and a multilevel evaluation structure is constructed with the target, criterion, and index layers. Then, the weight vectors of evaluation indicators are determined scientifically using the exponential scale method. The TOPSIS method is introduced to construct a stability judgment matrix using the indicator values of critical levels. Finally, the relative closeness values are calculated to evaluate and predict the stability level of goaf slopes comprehensively. A goaf slope example is used as the case study to demonstrate this model, and results are compared with common evaluation methods (i.e., Swedish slice, Bishop, BP neural network, Janbu, limit equilibrium, and Spencer method of strength reduction). Their calculated safety factors all exceed 1.25. The goaf slope stability level is grade II and in the stable state. The proposed model has an engineering application value in predicting goaf slope states, showing good performance, applicability, and rationality.

Based on BP neural network method to establish higher education health evaluation system

This paper is here to develop a model that can assess the health status of the higher education system in any country. Then select a country and propose a set of policies that will move that country from its current state to its target state. Five indicators for the evaluation of the education system are formulated. Then, Python crawler is used to capture the data of different countries under these five indicators from the Internet. Then, K-means++ clustering method is used to classify countries into three classes. The significance and specific results of classification are shown in 4.2.1 and Table3. Then, according to different types of countries, TOPSIS based on entropy weight method is used to establish a system that can evaluate the health status of higher education in any country.

基于GA-BP优化模型的中小河流健康评价研究

PDF HTML阅读 XML下载 导出引用 引用提醒 基于GA-BP优化模型的中小河流健康评价研究 DOI: 10.5846/stxb202003070437 作者: 作者单位: 作者简介: 通讯作者: 中图分类号: 基金项目: 国家自然科学基金面上项目（51979149） Health assessment of small and medium rivers based on GA-BP optimization model Author: Affiliation: Fund Project: 摘要 | 图/表 | 访问统计 | 参考文献 | 相似文献 | 引证文献 | 资源附件 | 文章评论 摘要:河流健康评价是河流生态修复的前提和依据，但目前中小河流还缺乏相应的评价标准。为客观评估中小河流健康状况，以桥边河为例，构建了河流水生生物、水环境、水文、形态、社会服务功能5个要素17个指标的健康评价体系，利用遗传算法（GA）优化BP神经网络方法建立GA-BP模型确定其指标权重，采用赋分法进行评价。GA-BP模型输出结果显示，指标权重中水环境指标赋值最大，其次是河流水文指标>河流水生生物指标>河流形态指标>河流社会服务功能指标。通过建立的评价体系对桥边河进行评价，结果表明有3个点健康等级评价为Ⅰ级，3个点评价为Ⅱ级，2个点评价为Ⅲ级。总体来看，桥边河流域基本处于健康状态，上游健康状况明显优于下游，从上游到下游越靠近河口位置，河流的健康状况越差。 Abstract:River health assessment is the premise and basis of river ecological restoration, but up to date there is no corresponding health assessment standard for small and medium-sized rivers. In order to objectively evaluate the health status of small and medium-sized rivers, this paper takes Qiaobian River as an example and constructs a health evaluation system around 5 elements, namely the river aquatic organisms, water environment, hydrology, morphology and social service functions, which contains 17 indexes. Then we capitalize genetic algorithm (GA) to optimize BP neural network method to establish GA-BP model to confirm the weight of these indexes, and use the scoring method to evaluate. According to the output of GA-BP model, the river's water environment has the largest value in the weights of indexes, sequentially followed by river hydrological index, river aquatic organism index, river shape index, and river social service function indicators. The result of evaluating in Qiaobian River through the established evaluation system shows that 3 points of them have a health rating of level I, 3 points are rated level Ⅱ, and 2 points are rated level Ⅲ. Overall, the Qiaobian River is basically in a healthy state, and the health status of the upper reaches of the basin is significantly better than that of the downstream. The closer to the estuary from upstream to downstream, the worse the health of the river is. 参考文献 相似文献 引证文献

Research on recognition method of cloud precipitation particle shape based on BP neural network

In order to solve the problem that the shape of cloud particle images measured by airborne cloud imaging probe (CIP) cannot be automatically recognized, this paper proposes an automatic recognition method of cloud and precipitation particle shape based on BP neural network. This method mainly uses a set of geometric parameters which can better describe the shape characteristics of cloud precipitation particles. Based on the cloud precipitation particle images measured by CIP in the precipitation stratiform clouds in northern China, a particle shape data training set and a testing set were constructed to train and verify the effect of the selected BP neural network model. The selected BP neural network model can classify the cloud particle image into tiny, column, needle, dendrite, aggregate, graupel, sphere, hexagonal and irregular. Utilizing the field campaign data measured by CIP, the habit identified results by the improved Holroyd method and by the selected BP neural network model were compared, which shows that the accuracy of BP neural network method is better than that of improved Holroyd method.

Fault Diagnosis of Commutation Failure Using Wavelet Transform and Wavelet Neural Network in HVDC Transmission System

Commutation failure is the most common fault on the inverter side of the high voltage direct current (HVDC) transmission system. In the actual engineering operation, there are many reasons for the commutation failure. In order to diagnose the causes in real-time and effectively, this article proposes a new method for the fault diagnosis of the commutation failure based on the combination of the wavelet transform and wavelet neural network (WNN). First, by analyzing the changes in electrical characteristics caused by the failure of commutation, the voltage signal of the inverter side commutation bus is finally selected as the feature signal for diagnosis. Then, the wavelet transform is used as a tool for extracting fault feature quantities, and the root mean square (rms) value of the wavelet detail coefficients, wavelet energy skewness, and wavelet energy spectrum information entropy of the inverter side commutation bus voltage signal after wavelet decomposition are used as three feature parameters. At the same time, wavelet packet decomposition is performed on the inverter side commutation bus voltage signal, and the wavelet packet subband energy, energy entropy, and energy skewness are extracted as the other three feature vectors. Finally, the simulation test data shows that the proposed fault diagnosis method can accurately identify the specific fault causes and has high reference value and specific engineering guiding significance for improving the reliability of the HVDC transmission system. Compared with the traditional BP neural network method for fault diagnosis of commutation failure, the proposed diagnosis method considers the fault type more comprehensively, the diagnosis speed is faster, and the recognition accuracy rate is higher in the case of a small number of samples.

Comparative study of metamodeling methods for modeling the constitutive relationships of the TC6 titanium alloy

It is important to model the flow behavior of materials before the design and optimization of the forming process. In this work, the TC6 titanium alloy was subjected to a temperature range of 800 °C to 1000 °C and a strain range of 0.01 s−1 to 10 s−1 in isothermal compression experiments, after which friction and temperature corrections were performed. Based on the corrected data, three metamodeling methods—the response surface method, the Kriging method, and the BP neural network—were used to model the constitutive relationships of the TC6 titanium alloy. The results show that the three models can better predict the flow stress of the TC6 titanium alloy than the Arrhenius model. In addition, by comparing the fitting accuracy of the three methods using different numbers of training samples, we found that the accuracy of the BP neural network method is very dependent on the number of training samples, while the fitting accuracy of the response surface method and the Kriging method is always maintained at a high level; when the number of training samples is between 30 and 240, the curved surfaces fitted by these two methods do not change significantly. Compared with the Arrhenius model, the response surface method and the Kriging method can reach higher accuracy with fewer training samples, which is significant for reducing the number of experiments and calculation time required to model the constitutive relationship.

Evaluation model of green supply chain cooperation credit based on BP neural network

More and more enterprises hope to achieve cooperation and win–win. However, many companies often have problems such as insufficient partner credit, which seriously affects the quality of cooperation. In order to effectively evaluate the credit, this paper constructs a personal credit evaluation model. The model compares the weight adjustment method with BP neural network and other methods. Compared with the BP neural network weight adjustment algorithm, the improved algorithm has obvious advantages in accuracy and convergence speed. The simulation results show that the green supply chain cooperation credit evaluation model can better evaluate the environmental behavior of enterprises. The BP neural network can better solve the problem of slow convergence and premature convergence, and can search data more accurately. The algorithm has good robustness. The evaluation model has high optimization accuracy, which shows that BP neural network can better learn and evaluate the credit of green supply chain at different levels.

Effect of Saline Soil Cracks on Satellite Spectral Inversion Electrical Conductivity

The dehydration cracking of saline soil is a kind of common natural phenomenon, and the cracks of saline soil will affect the satellite spectrum, and then affect the accuracy of satellite spectral inversion of electrical conductivity (EC). This study introduces the concept of crack rate (CR) to describe the crack information of saline soil, and quantifies the influence of saline soil crack on the EC of satellite spectral inversion. In 2014 and 2020, the satellite-ground synchronous observation experiments of soda-type inland saline soil and coastal chlorinated-type saline soil were carried out, and the CR of surface cracked saline soil was extracted by an image processing algorithm. For the saline soil spectrum data, the correlation analysis method is used to establish the best band combination that characterizes the relationship between the different saline soil spectrum data and salinity, and the EC inversion model is established using the BP neural network method. The results show that: after the CR is introduced, the determination coefficient (R2) for the EC of soda-type saline soil satellite spectral inversion increased from 0.59 to 0.67, with an increase of 14.42%, and the mean square error (MSE) reduced from 0.20 to 0.16, with a decrease of 19.49%. The R2 for the EC of coastal chlorinated-type saline soil satellite spectral inversion increased from 0.64 to 0.75, an increase of 17.73%, and the MSE decreased from 0.16 to 0.12, a decrease of 25.15%. The study proved the influence of the cracks in the saline soil on the satellite spectrum and provided a new way to improve the accuracy of the satellite spectrum inversion of the EC of the cracked saline soil.

Analysis of influencing factors of hospitalization cost based on BP neural network model*

In view of the present social status quo, “cost of medical care”, according to a national disease surveillance center of 2700 medical record sample data, using BP neural network method, through SPSS software operation, get in 26 factors affecting hospitalization cost, age, level of hospital, hospital source code, disease types, cost the importance of the five factors relative to the highest, and the accuracy rate of more than 85%, so as to provide basis for government decision-making departments to formulate corresponding health policy.

Application of genetic algorithm and BP neural network in supply chain finance under information sharing

The supply chain finance industry will generate the flow of funds and commodities when providing financing services to small and medium-sized enterprises (SMEs). At this time, banks will face multiple risks such as policy, operation, market and credit. The investigation on supply chain finance under information sharing from the aspect of credit risk assessment will be conducted. The genetic algorithm combined with support vector machine and BP neural network is selected to evaluate the credit risk of supply chain finance. In the support vector machine method, the parameter selection method adopts genetic algorithm. In the included data, the gap in growth capacity of SMEs is relatively large. The standard deviations of main business, net profit and total assets are all above 30%, and the standard deviations of current ratio and quick ratio are small, which means that the two are more stable and healthier. In addition, among all the investigated enterprises, the cost gap is large, and the standard deviation of the inventory decline price reserve is small, which means that most enterprises have good inventory quality. After classification, 32 high-quality enterprises, 46 neutral enterprises and 55 risk enterprises are obtained in the total sample. In the test sample, there are 21 high-quality enterprises, 12 neutral enterprises, and 26 risk enterprises. The overall classification accuracy of the support vector machine method optimized by genetic algorithm is relatively lower than that of the BP neural network method. The classification accuracy of the support vector machine method optimized by genetic algorithm is 76.27%, and the classification accuracy of BP neural network method is 89.83%. The supply chain financial risk assessment of SMEs is mainly explored from the perspective of banks. The results can provide theoretical support for reducing the probability of bank’s profit damage and increasing the bank’s profitability.

Research on prediction model of construction period of overhead line project based on BP neural network

With the stable development of economy and society, the power load increases year by year, so the power transmission company needs to build enough transmission lines to ensure the power supply. Overhead line project is the most commonly used form of transmission line, but the overhead line project often encounters the situation that cannot be completed on time, which affects the satisfaction of users’ electricity demand. Therefore, it is necessary to reasonably predict the construction period of the overhead line project, so as to provide a reasonable reference for the formulation of the production plan of the grid company. Firstly, this paper analyses the key factors that affect the construction period of overhead line project. Based on the BP neural network method, the construction period prediction model of overhead line project is established. The accuracy of the prediction is high.

Establishment of vibration wear model for turning large-pitch thread tools and its wear suppression method

Large-pitch thread parts are the key components of heavy-duty equipment and their function is the transmission of torque and accuracy, resulting in high precision and reliability requirements for their threaded faces. Since the vibration and tool wear are related to dynamics and tribology, respectively, directly establishing the mathematical relationship between them is complicated. Considering that the cutting force is not only the result of vibration effect but also the direct cause of tool wear, the relationship between the two is constructed in relation to the instantaneous cutting force in the process of machining. Firstly, the instantaneous cutting force models for different cutting speeds and axial feeds under vibration are constructed and verified. Secondly, the mathematical models of cutting force and rear face wear prediction are established based on the BP neural network method. Finally, based on these models, the vibration wear model of the large-pitch thread tool is obtained iteratively to describe the relationship between vibration and wear in the time domain, and the model is verified by experiments. The error in the model is within 20%. At the same time, based on the genetic algorithm and multi-objective optimization model, the optimal values of cutting parameters that minimize the impact of cutting vibration on tool wear are obtained, and a method of reducing the vibration to improve tool wear is proposed.

An Integrated quantitative modeling approach for fault-related fractures in tight sandstone reservoirs

Fault-related fractures widely exist in tight sandstone reservoirs, which greatly improve the heterogeneity of reservoir. Unraveling the spatial distribution of fractures is most important for enhancing oil and gas recovery. In this study, an integrated quantitative modeling approach is proposed and applied to the Boxing sag in Jiyang Depression, Bohai Basin, China. From observations of drilled cores and well log interpretations we construct the internal structure of the fault damage zone and establish its geological conceptual model. A single well rock mechanics model is built by applying BP neural network method on rock mechanics experimental data. With the sedimentary facies guided lithofacies model, a three-dimensional heterogeneous rock mechanics model is constructed by integrating three geostatistical methods (i.e. sequential Gaussian simulation method, probability volume modeling method and trend modeling method). This workflow assigns the geological concept and rock mechanics characteristics to the geological volume. Based on the energy conservation principle, the relationship between fracture density and strain energy is derived using different fracture criteria. Based on simulation and calculation results, the fracture density model is established by deterministic modeling. Finally, these properties are incorporated into a DFN model. The model results show that, at the regions near the faults, the fracture development is mainly controlled by faulting. The induced fracture zone is a fracture-intensive area. However, there are few fractures within the fault core zone. The regions far from the faults are mainly controlled by the rock mechanics properties of geological materials, which show zonal characteristics. This model is validated by data from five wells in this region, which shows a strong correlation between predicted fracture density and water cut in produced fluid.

A new method to production prediction for the shale gas reservoir

ABSTRACT It is a common phenomenon that shale gas production differs significantly among wells, and it is difficult to find a relationship between engineering variables and gas rate by utilizing a simple plotting of production data against well and treatment variables. Data such as completion method, flow back, and gas production have been collected for 180 shale gas wells. Firstly, the collected data were standardized, and then related or dependent variables were removed by gray correlation analysis. Finally, principal component analysis and BP Neural-network methods are applied separately to analyze the collected data. It is the first-of-its-kind study that was conducted to define the relationship between gas production and engineering parameters for Fuling shale gas reservoir in China. By use of principal component analysis, it is concluded that the number of stages, the lateral length, the distance between clusters, and the average volume of slick-water are the main factors that impact shale gas production. The results showed that stage numbers account for 42.9% of the total variance, while the horizontal length accounts for about 21.4%; therefore, these two factors account for 64.3% among all of the factors. However, the next important factors are the average distance between clusters and the average volume of slick-water, both account for 24.1% of the total variance. The importance of each variable was also defined by the BP Neural-network method. The number of stages and the horizontal length are still the two major factors, and both account for 52%. The volume of slick-water and the proppant size of 40/70 mesh for both ceramic and sand are considered the second important factor and both account for 37%. This implies that shale gas well configuration and treatment parameters are the dominant factors for gas production in shale reservoirs. Therefore, in order to enhance gas production, it is crucial to increase the number of stages and horizontal length firstly and to raise the volume of slick-water by using 40/70 mesh for both ceramic and sand. However, another driver to accelerate production is to decrease the distance of clusters. These measures will be helpful to improve production in the future shale gas development of Fuling reservoir. This study is believed to be the first-of-its-kind to define the relationship between gas production and engineering parameters for Fuling shale gas reservoir in China.

A prediction model for local scour depth based on BP and GA-BP neural network

A prediction model for the local scouring depth at bridge piers is built using BP and GA-BP neural network method. Measured data is collected from laboratory tests and field observation to work as train set of the predicting model which is verified by the measured data of Hangzhou Bay Bridge. The result shows that the predicted values of the local scour depth of the piers obtained by the two models are in good agreement with the measured values, which is more suitable than the existing general formula. What’s more, GA-BP model performs better than BP model.

Research on Wind Power Ultra-short-term Forecasting Method Based on PCA-LSTM

Wind power ultra-short-term forecasting can provide the support for adjusting the intraday power generation plan, carrying out the incremental spot trading of wind power, and improving the utilization of wind power. In order to improve the forecast accuracy of wind power, a wind power ultra-short-term power forecast method based on long-term-term memory (LSTM) network is proposed. First, the principal component analysis method is used to reduce the multivariate meteorological time series dimension. Then by using the cyclic memory characteristics of LSTM network to model multi-dimensional time series, the nonlinear mapping relationship between meteorological data and power data is established, and the wind power forecast is finally realized. The actual data of the eastern China wind farm is used to verify the results. It shows the method established in this paper can effectively use the meteorological and power data to forecast the wind power, and compared with the traditional time series, BP neural network method, the method in this paper has higher forecast accuracy and has broad application potential.

Research on Sensor Correction Method for Monitoring in Early Warning System of Coal Spontaneous Combustion

According to statistics, there are as many as 4000 coal spontaneous combustion accidents every year, causing heavy casualties and property losses. By accurately detecting the change of index gas concentration, it can provide reliable criteria for identifying and warning the early hidden dangers of coal spontaneous combustion. The improvement of sensor precision is the basis of intelligent monitoring. Modification of wireless multi-parameter sensor used in coal mine is of great significance for accurate monitoring of gas concentration in coal mine. In order to modify the wireless multi-parameter sensor, the high and low temperature experiments of the sensor were carried out at -5∼45°C, and the real values of O2 concentration at 15%, 21%, 25% and CO concentration at 175ppm, 250ppm and 375ppm were measured. Measurement data was Non-linear fitting and correction by using SVM (Support Vector Machine), BP neural network and Elastic Network regression method. The experimental results show that the Elastic Network regression compensation reduces the O2 sensor from the original maximum mean absolute percentage error (MAPE) of 18.4% to 0.52%, and the average MAPE decreases from 13.89% to 0.21%. Using the SVM nonlinear compensation, the CO sensor is reduced from the original maximum MAPE of 43.2% to 1.6%, and the average MAPE is reduced from 21.38% to 1.4%. Elastic Network regression compensation excellent achieves nonlinear compensation of O2 sensor. The SVM better realizes the nonlinear compensation of the CO sensor. Through non-linear compensation, wireless multi-parameter sensor can better meet the requirements of coal mine gas concentration monitoring.