## No full-text available

To read the full-text of this research,

you can request a copy directly from the authors.

This study evaluated if an Artificial Intelligence climate forecasting model can be considered as a useful tool for saving energy in semi-closed greenhouses. Preliminary results are presented on the 5-Minutes prediction of the internal air temperature and humidity modeled with Artificial Neural Networks (ANN). Since the final goal of the simulation is to integrate the predictions in a control system, the inputs were selected according to the standard signals in control theory: Set Points, Perturbations and Current State Vector. These inputs were: energy for heating, energy taken from cooling, ventilation opening, thermal screen opening, outside conditions (temperature, relative humidity, solar radiation, wind velocity) and current internal conditions (temperature and relative humidity). Data for the models were recorded in 2011, taken of 30-seconds-intervals. The ANN was created, trained and validated using different data sets. The prediction showed a very good fit to measured data and suggests that the ANN methods can be used to make short-term climate predictions, which are useful to take control actions before the trigger setpoints are reached.

To read the full-text of this research,

you can request a copy directly from the authors.

A procedure is presented for obtaining a dynamic linear model of auto-regression with exogenous variables (ARX) for predicting the behaviour of the air temperature inside a greenhouse. The ARX are dynamic mathematical models derived from the theory of Systems Identification. The input variables of the model were air temperature, solar radiation, wind velocity affecting the ventilation area of the greenhouse and relative humidity, quantified in a meteorological station located 700 m from a greenhouse in Chapingo, State of México. The response variable was the air temperature inside the greenhouse. Samples were taken of the input and output variables of the model every 5 min during a crop cycle. To determine the structure of the best model, as many as 100 000 ARX models were evaluated using the information criteria and final prediction error of Akaike. The adjustment between the simulated and observed temperatures, and the residual analysis, indicated that ARX models of second degree or above, adequately predict the behaviour of the temperature inside the greenhouse.

This work analyzes an energy consumption predictor for greenhouses using a multi-layer perceptron (MLP) artificial neural network (ANN) trained by means of the Levenbergh-Marquardt back propagation algorithm. The predictor uses cascade architecture, where the outputs of a temperature and relative humidity model are used as inputs for the predictor, in addition to time and energy consumption. The performance of the predictor was evaluated using real data obtained from a greenhouse located at the Queretaro State University, Mexico. This study shows the advantages of the ANN with a focus through analysis of variance (ANOVA). Energy consumption values estimated with an ANN were compared with regression-estimated and actual values using ANOVA and mean comparison procedures. Results show that the selected ANN model gave a better estimation of energy consumption with a 95% significant level. The study resents an algorithm based in ANOVA procedures and ANN models to predict energy consumption in greenhouses.

In this paper, the characterization and modelling of the most relevant convective transfers contributing to the elaboration of the greenhouse climate are reviewed. Convective transfers include heat and mass transfers between air and solid surfaces (walls, roof, leaves) along with air, heat, water vapour and tracer gas transfers to or from the inside air. Adopting the assumption that the greenhouse is a perfectly stirred tank, the specific characterization methods associated with this approach are reviewed. The perfectly stirred tank approach requires the assumption of uniform temperature, humidity and CO2 content inside the greenhouse and uses a ‘big leaf’ model to treat the plant canopy and describe the exchanges of latent and sensible heat with inside air. The simulation of the ventilation processes associated with this simplified approach is based on the Bernoulli equation and on the experimental determination of semi-empirical parameters by means of air exchange rate measurements. The techniques used to measure temperature and air exchange rates measurements pertaining to the whole greenhouse volume are presented. A complete panorama of the studies in relation to the transfer coefficients between the different surfaces together with the ventilation performances of various greenhouse types are also presented.This paper is the first part of a review of the convective transfers in greenhouses and in the second paper, a similar study based on the approach of the distributed climate is presented.

In this study three models are developed, one for prediction of the inside temperature 5 and 10 min ahead of time, other for prediction of CO2 concentration 5 min in advance, and the results of this model are linked with a third model for photosynthesis prediction 5 min ahead. Data were collected from two compartments at the experimental greenhouse (Humboldt University). Artificial Neural Networks (ANN) were used because of their ability to capture the nonlinear relationships governing the greenhouse environment. Matlab's Neural Networks Toolbox was used to train, validate and test the NN models. A data pattern of 7800 and 11 input variables were used for prediction of the inside temperature. A linear regression was performed between actual and predicted values with coefficients of determination (CD) of 0.997 and 0.994 and mean square errors (MSE) of 3.48 and 5.84 for 5 and 10 min temperature predictions. To evaluate the performance of the ANN, a different 200 data set was fed into the NN (October 28, 2 to 6:35 pm), and the predictions were very precise with MSE between actual and predicted values of 0.088 and 0.029. Eight input variables and a 1800 data set were used for predicting CO2 concentration 5 min ahead of time. The CD of the linear regression between actual and predicted values was 0.994. Again the ANN was fed with 200 different input data (22 June, 3:20 to 19:55 pm) and MSE between actual and predicted values was 535. The results from the CO2 model were used as an input in the photosynthesis model. In this last model seven variables were used and the predictions were very precise in both cases for photosynthesis 5 and 10 min ahead. The sensitivity analysis performed shows that relative humidity is one of the most important variables affecting photosynthesis prediction.

The use of greenhouses for vegetable production has expanded rapidly in recent years in México. The most important aspect for the success of these agro-businesses is the improvement of production efficiency, higher quality and productivity, which are related to specific climatic conditions of each region. The present study introduces the analysis and simulation of a mathematical model of greenhouse climate. The model is formulated on basic principles of mass and energy transfer processes, and simulates the greenhouse air, soil, roof and crop temperatures, as well as relative humidity. The dynamic simulation of the mathematical model was made using the SIMULINK tool of MATLAB software. The simulation results were calibrated and validated with the measured data collected in a 1000 m2 experimental greenhouse developed at the university of Querétaro. The results from the validation were prediction equations for the roof temperature (R2=0.955), inner air temperature (R2=0.964), crop temperature (R2=0.835), soil temperature (R2=0.714) and relative humidity (R2=0.960). The magnitude of the coefficients indicates that the model can be used to predict the greenhouse climate with a high level of confidence, and it is a tool for supporting the analysiś of the necessary conditions for greenhouse vegetable production under climatic conditions of the Central Region of México.

The adequacy of radial basis function neural networks to model the inside air temperature of a hydroponic greenhouse as a function of the outside air temperature and solar radiation, and the inside relative humidity, is addressed. As the model is intended to be incorporated in an environmental control strategy both oo-line and on-line methods could be of use to accomplish this task. In this paper known hybrid oo-line training methods and on-line learning algorithms are analyzed. An oo-line method and its application to on-line learning is proposed. It exploits the linear–non-linear structure found in radial basis function neural networks.

A new greenhouse type has been designed to study ways of decreasing water use by horticulture in semi-arid regions. To control the greenhouse a model-based control design is required. To this end a model is needed to predict the systems behaviour (1 day ahead), without much computational effort. A physics-based model is developed, based on enthalpy and mass balances. The (lumped) key parameters of the model are identified with a controlled random search algorithm. To increase estimation accuracy and reduce computation time, estimation in parts was applied, that is only a part of the whole model was used in combination with measured data for state values of neighbouring compartments. This results in parameter estimates that converge well. In order to keep the model information needs limited, the underlying process details were aggregated into a lumped parameter description, at the expense of time-varying parameters over the seasons. The parameter fluctuation over the year was studied by repeated monthly parameter estimations. Since parameters fluctuate significantly, further research will focus on the use of adaptive mechanisms to facilitate model-based control.

Outliers are observations that do not follow the statistical distribution of the bulk of the data, and consequently may lead to erroneous results with respect to statistical analysis. Many conventional outlier detection tools are based on the assumption that the data is identically and independently distributed. In this paper, an outlier-resistant data filter-cleaner is proposed. The proposed data filter-cleaner includes an on-line outlier-resistant estimate of the process model and combines it with a modified Kalman filter to detect and “clean” outliers. The advantage over existing methods is that the proposed method has the following features: (a) a priori knowledge of the process model is not required; (b) it is applicable to autocorrelated data; (c) it can be implemented on-line; and (d) it tries to only clean (i.e., detects and replaces) outliers and preserves all other information in the data.

A brief description of the relation between systems and models as well as the modelling process is the starting point for a classification of potential application fields for models in horticulture. The horticulture production process can be characterised as an open and highly complex system affected by weather, soil, insects, diseases, weeds, nutrition, prices and interactions of these many factors. At the moment, knowledge of the whole system is rather limited and models describing their behaviour are incomplete approximations of the real system that they attempt to simulate. Nevertheless, it is possible to identify different types of problems, a grower might be confronted with in horticulture (operational, tactical, strategic decisions). In order to implement models for decision support, it is not sufficient to know the potential problems, it is also necessary to understand the decision making process which is described from a more theoretical point of view. A review of the evolution of computer-based systems for supporting decision making completes the preceding descriptions. In confronting the different types of real problems with the available technical possibilities, the discussion about implementation problems will be opened, including the question, who should or will apply models to derive answers to problems. It is concluded that the use of models in practice will only increase if the models deal with problems faced by the decision makers and if it becomes obvious to the farmers that they can derive answers to their problems on a more efficient way using specific models.

Artificial neural networks possess several properties that make them particularly attractive for applications to modelling and control of complex non-linear systems. Among these properties are their universal approximation ability, their parallel network structure and the availability of on- and off-line learning methods for the interconnection weights. However, dynamic models that contain neural network architectures might be highly non-linear and difficult to analyse as a result. Artificial Neural Networks for Modelling and Control of Non-Linear Systems investigates the subject from a system theoretical point of view. However the mathematical theory that is required from the reader is limited to matrix calculus, basic analysis, differential equations and basic linear system theory. No preliminary knowledge of neural networks is explicitly required. The book presents both classical and novel network architectures and learning algorithms for modelling and control. Topics include non-linear system identification, neural optimal control, top-down model based neural control design and stability analysis of neural control systems. A major contribution of this book is to introduce NLq Theory as an extension towards modern control theory, in order to analyze and synthesize non-linear systems that contain linear together with static non-linear operators that satisfy a sector condition: neural state space control systems are an example. Moreover, it turns out that NLq Theory is unifying with respect to many problems arising in neural networks, systems and control. Examples show that complex non-linear systems can be modelled and controlled within NLq theory, including mastering chaos. The didactic flavor of this book makes it suitable for use as a text for a course on Neural Networks. In addition, researchers and designers will find many important new techniques, in particular NLq Theory, that have applications in control theory, system theory, circuit theory and Time Series Analysis.

System identification is the art and science of building mathematical models of dynamic systems from observed input–output data. It can be seen as the interface between the real world of applications and the mathematical world of control theory and model abstractions. As such, it is an ubiquitous necessity for successful applications. System identification is a very large topic, with different techniques that depend on the character of the models to be estimated: linear, nonlinear, hybrid, nonparametric, etc. At the same time, the area can be characterized by a small number of leading principles, e.g. to look for sustainable descriptions by proper decisions in the triangle of model complexity, information contents in the data, and effective validation. The area has many facets and there are many approaches and methods. A tutorial or a survey in a few pages is not quite possible. Instead, this presentation aims at giving an overview of the “science” side, i.e. basic principles and results and at pointing to open problem areas in the practical, “art”, side of how to approach and solve a real problem.

Six prototypes plastic greenhouses were built in the tropical lowlands of Indonesia. The geometrical dimensions were designed using computational fluid dynamics (CFD) by taking local climate parameters as static reference boundary conditions. It is necessary to evaluate the climate dynamics inside the greenhouse during varying climatological conditions. A greenhouse climate model was developed to optimise cover properties and ventilation rate as main parameters, calculating only three state variables: average greenhouse air temperature TAir, average greenhouse air water vapour pressure (expressed as air water vapour pressure deficit DAir), and average canopy temperature TCan. Solar radiation distribution, air exchange by ventilation, and crop transpiration constituted the backbones of the model. The climate outdoor and inside the test greenhouses with crops having leaf area index from 0.02 to 4.10 were measured for one growing season. Measurements and calculations of TAir and DAir agreed satisfactorily, with less than 5% errors. It is concluded that the model is robust and could be used as a design tool for the tropical lowland greenhouses

We propose a robust learning algorithm and apply it to recurrent neural networks. This algorithm is based on filtering outliers from the data and then estimating parameters from the filtered data. The filtering removes outliers from both the target function and the inputs of the neural network. The filtering is soft in that some outliers are neither completely rejected nor accepted. To show the need for robust recurrent networks, we compare the predictive ability of least squares estimated recurrent networks on synthetic data and on the Puget Power Electric Demand time series. These investigations result in a class of recurrent neural networks, NARMA(p,q), which show advantages over feedforward neural networks for time series with a moving average component. Conventional least squares methods of fitting NARMA(p,q) neural network models are shown to suffer a lack of robustness towards outliers. This sensitivity to outliers is demonstrated on both the synthetic and real data sets. Filtering the Puget Power Electric Demand time series is shown to automatically remove the outliers due to holidays. Neural networks trained on filtered data are then shown to give better predictions than neural networks trained on unfiltered time series.

Greenhouse climate models: an overview. EFITA Conference. Debreecen, Hungary 5-9 July

- Literature Cited Boaventura-Cunha

Literature Cited
Boaventura-Cunha, J. 2003. Greenhouse climate models: an overview. EFITA
Conference. Debreecen, Hungary 5-9 July. p.823-829.

Greenhouse climate models: an overview

- J Literature Cited Boaventura-Cunha

Literature Cited
Boaventura-Cunha, J. 2003. Greenhouse climate models: an overview. EFITA
Conference. Debreecen, Hungary 5-9 July. p.823-829.