Grey Neural Network Research Articles

Improved gray neural network model for healthcare waste recycling forecasting

This paper addresses the problem of predicting multiple factors of health care waste recycling for promoting the construction of ecological civilization against the background of a comprehensive implementation of the “healthy China” strategy. In this paper, an improved gray neural network model is developed for recovery prediction. The one order single variable gray model and triple exponential smoothing are used for predicting the factors affecting recovery. A particle swarm optimization optimized back propagation (BP) neural network is trained by selecting higher-prediction results by precision comparison, and the trained BP neural network is used to predict the recovery of health care waste. Taking Shanghai as an example, this paper uses the actual data about Shanghai health care waste during 2013–2017, and the historical data of 11 factors affecting the recovery amount; these are used for empirical analysis. The results of this research show that the gray neural network performs better than other benchmark models and traditional predictive models. An accurate prediction of the amount of health care waste recovered can help decision-makers to implement recycling, and can serve as a reference for governmental departments, helping them to formulate relevant laws and regulations, develop a variety of tasks, and rationally allocate resources.

Research on Combination Forecast Model of Aerospace Relay’s Storage Life

With the rapid development of China space industry, relay is widely used in aerospace control system, thus it has a higher request for the relay reliability. The space relay is stored for a long time and may have a storage life of 15-20 years or much longer. In order to ensure that the relay always keeps in the activated state, it has very practical significance to predict the storage life of relay. Firstly, a single prediction model of space relay storage life is established, such as gray system theory prediction model and neural network prediction model. Then the prediction results are analyzed and the model error evaluation is carried out. According to the results of comparison, the advantages and disadvantages are obtained and improved. Finally, the storage life combination model of space relay is established, and Matlab software is used for the simulation and prediction of gray neural network to improve the accuracy of the prediction model

Degradation prediction of proton exchange membrane fuel cell based on grey neural network model and particle swarm optimization

The degradation prediction is an effective tool for the long-lasting operation of the proton exchange membrane fuel cells (PEMFC). In this paper, a novel grey neural network model (GNNM) method combined with the particle swarm optimization (PSO) and the moving window method is presented to forecast the degradation of PEMFC under different operating conditions. The proposed method considers the influence of load current, inlet temperature, inlet hydrogen pressure, and inlet relative humidity. The degradation prediction model of PEMFC is established by a grey neural network. The initial weight and threshold of established GNNM are optimized by the PSO. The optimized PSO-GNNM is iteratively trained based on the moving window method through several newly measured data. The influence of different moving window sizes on the degradation prediction of the PEMFC under the static load current is investigated. Then, a comparison between the proposed method and the adaptive neuro-fuzzy inference system method is discussed. Moreover, the influence of the load current on the degradation prediction of PEMFCs in postal fuel cell electric vehicle operating under real conditions is analyzed. Finally, the proposed method is validated using 3 PEMFC aging experiments under different conditions. The results show that the proposed method can precisely forecast the degradation for PEMFC on different applications.

The Application of a Time Series Forecasting Model in Air Passenger Volume Prediction Research

This paper presents a new forecasting model of time series (NFMTS). NFMTS is a set of time series forecasting models. In forecasting air passenger volume in Chengdu from 1993 to 2001, take point A (0.0002,0.8,0.0002) as the starting point of calculation, by calculating and screening, the automatic optimization search method of NFMTS can be applied to screen the optimal time series prediction model in NFMTS as Ju (0.00002,0.8,0.00002). It can not only make the predicted value of Chengdu's air passenger volume reach MSE=0 and AFER=0, but also can predict the future air passenger volume of Chengdu from 2002 to 2003. The prediction model presented in this paper is simple in calculation and easy to use. In dealing with the air passenger volume of Chengdu from 1993 to 2003. It has advantages over traditional grey model, grey neural network model and grey support vector machine model.

Prediction of District Heating Load Based on Grey Neural Network Model

In this paper, the gray correlation analysis method is used to evaluate the factors affecting the district heating load, and the gray prediction is combined with the BP neural network algorithm to establish a gray neural network structure, which can screen the factors affecting the heating load and predict the heating load. The heating load forecasting and verification of a district heating load is carried out. By comparing the prediction results and errors of the gray correlation GM(1, N) model, it is shown that the gray neural network model can select the appropriate influencing factors and exclude impact factors with low relevance in the heat load forecasting, improve the accuracy of heat load forecasting, and provide a theoretical basis for regional heat load forecasting.

Application of grey BP neural network in port logistics demand analysis

From the perspective of port cargo throughput, this paper firstly analyses the characteristics and influencing factors of port logistics demand. Secondly, considering the characteristics of nonlinear logistics demand and small sample modelling, the modelling adopts GM(1, 1) and the single prediction model of BP neural network for calculation. Then, based on the prediction results and the target of minimum fitting prediction square-error, the single model is given weight, and the combined prediction model is constructed. Finally, taken Qingdao Port as an example, the port logistics demand is simulated by MATLAB software. The results show that the combined forecasting model has higher accuracy and stronger stability than the single forecasting model, which can effectively reduce the error rate and make the forecasting result closer to reality, thus having guiding significance for the future port logistics development planning.

Air Target Threat Assessment Based on Improved Moth Flame Optimization‐Gray Neural Network Model

Air target threat assessment is a key issue in air defense operations. Aiming at the shortcomings of traditional threat assessment methods, such as one‐sided, subjective, and low‐accuracy, a new method of air target threat assessment based on gray neural network model (GNNM) optimized by improved moth flame optimization (IMFO) algorithm is proposed. The model fully combines with excellent optimization performance of IMFO with powerful learning performance of GNNM. Finally, the model is trained and evaluated using the target threat database data. The simulation results show that compared with the GNNM model and the MFO‐GNNM model, the proposed model has a mean square error of only 0.0012 when conducting threat assessment, which has higher accuracy and evaluates 25 groups of targets in 10 milliseconds, which meets real‐time requirements. Therefore, the model can be effectively used for air target threat assessment.

Modeling and application for air quality evaluation based on improved grey clustering and neural networks classification

According to the air quality monitoring date of the six urban evaluation sites arranged in six main administrative districts in Beijing during the four seasons from 2013 to 2014, an evaluation system based on air quality indexes of Chinese was formulated. And, next, based on concentration limits of six main pollutants including PM10, PM2.5, O3, CO, SO2 and NO2, analysis of the improved grey clustering method and the neural networks classification method were used for evaluation by using the detailed evaluation procedures. An improved grey clustering method based on the whitening function of the exponential types was adopted to improve the calculation precision of the fuzzy factors. It could be seen that the method adopted could more fully reflect the comprehensive effects of all the pollutants on the air quality. Meanwhile, two classification methods based on neural networks were adopted to classify the pollutants into different levels by MATLAB tools. By these methods the comprehensive influence of the pollutants on ambient air could be determined accurately and quickly. The evaluation results demonstrated that the improved grey clustering method which modified weight and extend the whitening function scope could greatly improve information utilization. By considering the weights of whitening values and standard values in different grey classes, the cluster result was more comparable. The neural networks classification method effectively improved the network generalization by the massive data of training samples and testing samples constructed by LINSPACE function in MATLAB tools. The evaluation results obtained by BP and RBF neural network learning showed high accuracy. Among them, the evaluation results of RBF neural network classification and grey clustering modified were more similar that the air quality level of six main administrative districts in Beijing was generally in grade 2. According to the evaluation purposes of the ambient air quality and the practical evaluation demands, it was possible to combine these two methods from different perspectives to gain a more reasonable evaluation result. Thus, it could be concluded that the evaluation of the pollutant contents in the air was of great practical value for eventual heightening of the new ambient air quality standard.

Civil aviation safety evaluation based on deep belief network and principal component analysis

Civil aviation in modern industries is becoming increasingly automatic, precise, and efficient. Serious accidents and unsafe incidents are used to describe and investigate the safety level. Therefore, effectively predicting big data from these features and accurately identifying the safety level with advanced theories are new issues in civil aviation. The prediction of serious flight incident rate for unsafe events is proposed on the basis of deep learning considering the characteristics of big data. In this method, deep belief network (DBN) is combined with Principal Component Analysis (PCA). The deep architecture is beneficial for safety prediction because each layer learns more complex features than the layers before. Compared with the previous prediction based on historical accident data, The DBN predicts the serious flight incident rate based on the results from PCA. Results indicate that the prediction data of PAC-DBN is consistent with the actual data of serious flight incident rate. The proposed method is superior to forecasting the serious flight incident rate compared with the gray neural network method, support vector regression, DBN. Simultaneously, the main influencing factors can be extracted to reduce flight incident rate.

Adaptive regulating of automotive mono-tube hydraulic adjustable dampers using gray neural network–based compensation system

With the development of the controllable suspension systems, the mono-tube hydraulic adjustable damper has attracted great public attention with the advantages such as good heat dissipation, less power, fast response, durable, reliable, and simple structure. However, the unknown regulating mechanism modeling impedes the practical application of the mono-tube hydraulic adjustable damper. To model the regulating mechanism, this paper analytically studies the behavior of the mono-tube hydraulic adjustable damper via developing an analytical model and thermal effect equations for the use of engineering design. Then, the mono-tube hydraulic adjustable damper is tested in an integral shock absorber testing system to verify the accuracy of model and equations. On the basis of the verified analytical model and thermal effect equations, a compensation system with gray neural network algorithm is originally designed to model the regulating mechanism of the mono-tube hydraulic adjustable damper, thus achieving the desired damping force adaptively and accurately at various working conditions by obtaining the required rotary angle of the adjustment rod. The simulation results and experimental results show that the characteristic analyses of mono-tube hydraulic adjustable damper are reliable. Meanwhile, the simulation results of the gray neural network algorithm also indicate that the proposed compensation system can provide an exact regulating mechanism model for the mono-tube hydraulic adjustable damper and the proposed gray neural network algorithm is superior to the traditional neural network algorithm.

A hybrid model using signal processing technology, econometric models and neural network for carbon spot price forecasting

Carbon spot price forecasting result is important for both policymakers and market participants. However, because of the complex features of carbon spot price, accurate forecasting is very difficult. To achieve a better prediction precision, a hybrid model combined with complete ensemble empirical mode decomposition (CEEMD), co-integration model (CIM), generalized autoregressive conditional heteroskedasticity model (GARCH), and grey neural network (GNN) optimized by ant colony algorithm (ACA) is proposed. Then it is validated by using data collected from European Union emission trading scheme (EU ETS). The results indicate that the performance of the chosen model is remarkably better than that of other models. Therefore, the hybrid model could be used more frequently for carbon spot price forecasting in the future.

Forecasting of power grid investment capability based on grey neural network combination model

Under the market environment, power generation investment has become an investment behavior aimed at maximizing the interests of investors, and the construction of power generation projects presents the characteristics of uncertainty, so it is difficult to balance power supply and demand. In this context, how to make their own optimal investment decisions to reduce risk has become a very important issue for a single power generation company. Therefore, from the point of view of the whole system, this paper focuses on the causal relationship between the variables of the power generation investment system under the market environment, and on the basis of power supply and demand balance, puts forward a power grid investment capacity prediction model based on the grey neural network combination model. In this paper, the investment of power grid in a province of China is taken as the research object, the grey neural network prediction model is constructed, the investment demand and the internal and external influence indexes of the investment capacity of power grid are analyzed, and the investment demand and the investment capacity in the next six years are predicted. The prediction results verify the correctness and effectiveness of the grey neural network model in the investment prediction of power grid.

Practical multi-objective control for automotive semi-active suspension system with nonlinear hydraulic adjustable damper

In recent, the development of the automotive semi-active suspension (SAS) system with the nonlinear hydraulic adjustable damper (HAD) has attracted widely attention due to its superiorities of low energy consumption, fast response, durable, reliable and simple structure. Many existing SAS researches focus on control algorithms for limited objectives and assume the control signal is the damping force, but very few of them discusses how to realize the calculated damping force. To solve this problem, the unknown regulating mechanism of the nonlinear HAD should be modeled and a complex multi-objective control of the SAS system should be designed. Aiming at modeling the regulating mechanism of the nonlinear HAD, this work constructs a novel compensation system by using fuzzy neural network or grey neural network algorithm in order to realize the desired damping force by regulating the stepper motor angle adaptively. Meanwhile, a new multi-objective control strategy, particle swarm optimization based sliding mode control method, is originally proposed to calculate the desired damping force to achieve exact damping force control of the SAS system. To realize the practical application of the SAS system with the nonlinear HAD, an online hardware-in-the-loop test is conducted on a quarter car test rig. Numerical and experimental results illustrate that the modeled regulating mechanism of the nonlinear HAD and the proposed multi-objective control of SAS system are effective to improve the vertical performance of the vehicle and feasible in practical applications.

Grey Wolf Optimization Trained Feed Foreword Neural Network for Breast Cancer Classification

Breast cancer is the most common invasive cancer in females worldwide and is major cause of deaths. The diagnoses of breast cancer include mammograms, breast ultrasound, magnetic resonance imaging (MRI), ductogram and biopsy. Biopsy is best and only way to know if the breast tumor is cancerous. Report says that positive detection of breast cancer through biopsy can reach as low as 10%. So many statistical techniques and cognitive science approaches like artificial intelligence are used to detect the type of breast cancer in a patient for getting more accuracy. This article presents the breast cancer classification using feed foreword neural network trained by grey wolf optimization algorithm. The superiority of the GWO-FFNN is shown by experimenting Wisconsin Hospital data set (Breast Cancer Wisconsin) and comparing recently reported results. The evaluations show that the proposed approach is very robust, effective and gives better correct classification as compared to other classifiers.

Application of Grey Neural Network with Particle Swarm Optimization on Adaptive Setting Value Selection in Phase-selection Controller

The power system is affected by different levels of transient shock while the AC filter is switching on. In order to minimize the transient impact, the circuit breaker is operated by the phase-selection controller using phase separation control which makes each phase breaker switching at the zero-crossing point of voltage. Therefore the accuracy of setting value selection in phase-selection controller will influence the switching impact. Statistics found that mechanical properties and electrical properties varied with mechanical wear and the aging of the components, which made the actual closing time change over time and deviate from the fixed value. Thus, a PSO-based grey neural network model is built to forecast the actual closing time. This model is optimized by PSO algorithm using the historical data. Finally, the predicted results show the effectiveness and feasibility of this method.

Integrated buffer monitoring and control based on grey neural network

Classic buffer monitoring methods only consider the buffer consumption amount, while the subsequent trend information of buffer consumption is ignored. In this paper, we propose an integrated buffer monitoring method. First, the prediction model based on grey neural network is established, and the follow-up buffer consumption is predicted quantitatively according to the past and present performance data at the project monitoring points. Second, considering the relationship between the buffered consumed and the follow-up buffer consumption, a buffer integrated monitoring system is formed based on the integrated quantitative analysis on the buffer consumed and the subsequent trend information at each monitoring point. Finally, the Monte Carlo simulation experiment is carried out to validate the model system. The results show that, as opposed to the classic buffer monitoring methods, the proposed method can control the progress of the project comprehensively and reduce the fluctuation of project duration, thus achieving the double optimisation of project duration and cost under the premise of guaranteeing the completion probability.

Design of Order Demand Forecasting System Based on Neural Network

As the market demand is seasonal and random, the traditional data model can not describe the order change rules accurately. In order to improve the prediction accuracy, neural network and gray theory are combined to construct a gray neural network order forecasting method. Gray system theory can capture the regularity of data sequence effectively. The optimal neural network parameters are obtained by training the data set. In the neural network model, enter a small amount of data to achieve accurate prediction of the order. The simulation results show that the improved gray neural network improves the prediction accuracy of the order requirement and provides the basis for the forecast of the market demand compared with the traditional forecasting method.

Prediction model for railway freight volume with GCA-genetic algorithm-generalized neural network: empirical analysis of China

Reasonable and scientific prediction for railway freight volume has an important impact on railway network planning and railway transportation resources allocation. However, to predict the future railway freight volume is complicated and difficult, because it is influenced by many factors, such as macro economy, industrial structure, and supply capacity, etc. In this paper, an improved prediction model is proposed, named as GCA-GA-GNN. GCA is short for grey correlation analysis, which was adopted to select the key factors which have great influence on railway freight volume instead of subjective factors. GNN is the main body of the prediction model, which combines grey prediction model and neural networks to take the advantages of linear and nonlinear modeling capabilities. Moreover, genetic algorithm is used in GNN to improve calculating speed. Then, the validity of the model was verified by the empirical case of China. The results of five different prediction models showed that the model proposed in this paper has faster convergence speed and higher prediction accuracy. Moreover, according to the downward trend of China railway freight volume from the year 2017 to 2020, some suggestions are proposed to reverse the downward trend and increase railway corporation’s profits.

Gray BP neural network based prediction of rice protein interaction network

In order to improve the effectiveness of the network prediction result of rice protein interaction, the network prediction method of rice protein interaction based on gray BP neural network has been proposed. Firstly, a series of key features about interaction sites of rice protein such as the series spectrum, conserved weight, entropy value, accessible exterior product of compound and sequence rate, etc., should be extracted firstly. Then the gray BP neural network and their integration will be applied to the training and test of these sample set. Ten times of cross validation are applied to the training and test, and four groups of feature combinations of rice protein interaction with comparability are created. As for each time of the addition of new features in the experiment, the accurate prediction rate would be improved, especially in case of the addition of exterior product and sequence rate features, the accuracy can be improved greatly, which means that in case the combination of multiple features is adopted, the method of predicting the interaction of rice protein by combining the gray BP neural network algorithm is accurate and effective.

Grey neural network prediction model based on fruit fly optimisation algorithm and its application

The vinyl acetate (VAC) polymerisation rate is considered as an important quality index in the production of polyvinyl alcohol. However, the quality of polyvinyl alcohol can not be controlled effectively because it can not be measured online. Therefore, how to design reliable estimation of VAC polymerisation rate is crucial. The fruit fly optimisation algorithm (FOA), as a novel meta-heuristic and evolutional algorithm, has several merits, such as higher prediction accuracy, having few parameters to be adjusted and able to achieve global optimum. This paper proposes a grey neural network prediction model to improve the prediction performance by taking advantage of FOA to optimise the 'whitening' parameters of this grey neural network. Based on the data from a real plant, the proposed grey neural network prediction model combined with FOA (FOA_GNN) is evaluated and proved to be valid. Its simulation experimental results also confirm the advantage of FOA_GNN algorithm to the traditional grey neural network model (GNN), the adaptive compete genetic neural network prediction model (ACGA) and radial basic function (RBF) neural network model.