Application of SVR optimized by Modified Simulated Annealing (MSA-SVR) air conditioning load prediction model

doi:10.1016/j.jii.2018.04.003

Journal of Industrial Information Integration

Volume 15, September 2019, Pages 247-251

https://doi.org/10.1016/j.jii.2018.04.003 Get rights and content

Abstract

In conditioning air load prediction model based on SVR model, the Simulated Annealing (SA) has been provided in order to surmount the disadvantage that the SVR model selects learning parameters depending on experience. The Modified Simulated Annealing (MSA) has been proposed to optimize the SVR prediction model, in which the annealing plan and disturbance range has been improved. Case researches in the paper show that MSA algorithm is of strong global optimization capability, good robustness and short calculation consumption. Compared with SA-SVR model and the VFSA-SVR model, MSA-SVR air conditioning load prediction model, the results show SVR model parameters obtained through MSA optimization can effectively improve the predication accuracy and stability of the air conditioning load prediction.

Introduction

For actual building in operation, the air conditioning load prediction is the long-term prediction on the cooling and heating energy necessary for the operation of air conditioning system in the future. It is intended to optimize the control service for the air conditioning system, determine optimal operation condition and set point based on predicted load distribution and establish optimal air conditioning operation strategy as to ensure the comfort of the air conditional room and the energy conservation of the air conditioning system operation. In order to accurately prediction, the air conditioning load, current researches mainly focus on the modeling technology and input parameters.

Air conditioning load prediction is the long-term prediction of the cooling and thermal energy required for the future operation of the air conditioning system. That is, the system will predict the real-time load and then turn on turn off the air-conditioner according, to the load prediction, in this way, the system will try to running the air-conditioner in daytime while the cost for power is higher as less as possible. Therefore, the power consumption and cost will be decreased. Air-conditioning load prediction has a large impact on air-conditioning's control strategy, and this control strategy has large impact on the consumption of an air-conditioning. In order to have a more ecofriendly usage of air-conditioner, an air-conditioning's load need to be accurately predicted. Air-conditioning Load Prediction plays an important part in air-conditioning system. A good air-conditioning load prediction system is able to correctly predict the load of an air-conditioning, accordingly, the temperature control of the air-conditioning will be adjusted. The accuracy of the predicting system decides whether the air-conditioning could reach the best temperature control as well as operating under the minimal consumption. Load predicting is vital for an air-conditioning to save energy.

The purpose of air conditioning load prediction is to optimize the control services of the air conditioning system, determine the optimal operating conditions and set load points based on the predicted load and to establish optimal air conditioning operation strategy to ensure the comfort of air-conditioned rooms and energy-saving of air-conditioning system.

Currently common air conditioning load predication includes duration model [1], Kalmanfilter model [2], Artificial neural network model [3], ARMA [4], and the like. The support vector regression machine (SVR) in SVM, as a machine learning with better non-linear fitting capacity, has been widely used in the research on the air conditioning load prediction in recent years [5], [6], [7]. SVR and ARMA models are respectively used in literature [8] to predict the air conditioning load. SVR model is used in the literature [9] to predict the air conditioning load, and the influence of the quantity of the modeling data on the prediction result is mainly study. The literature has [10] carried out the Burbank temperature change cycle combined particle swarm SVR short term load forecasting. Document [11] is based on the combination of time series and LS-SVR. Document [12] uses nonlinear time series and SVR to combine milk movement detection models. Empirical mode decomposition is used in the literatures to decompose the air conditioning load sequence into a series of relatively stable components and utilize SVR model to set up prediction models for each component. Selection of proper learning parameters is very significant for the learning performance and the generalization capability of the SVR model, directly affecting the prediction accuracy.

Simulated Annealing [13] (SA) is of strong global optimization capability, good robustness and short calculation consumption and thus has been successfully used in solve problems in parametric inversion and control correction of air conditioning system. In order to overcome the problems in relatively low operation efficiency and larger low-temperature disturbance of the simulated annealing algorithm and easily jumping out of the locally optimal solution, the SA has been provided with memory device, the annealing plan and disturbance range have been improved, the Modified Simulated Annealing (MSA) has been proposed, and MSA has been used to optimize the selection of learning parameters of SVR model in this paper. The example analysis verifies that, compared with SA-SVR model of SA and the VFSA-SVR model of Very Fast Simulated Annealing [14] (VFSA), MSA-SVR model is of better prediction accuracy and stability in the air conditioning load prediction.

Support Vector Machine (SVM) is a classification and regression machine learning method proposed by Vapnik. It can solve a series of problems such as small sample, non-linearity, high dimension and local minimum. At the same time, SVM also make up for the other methods in the sample data volume defects, can be used for non-linear regression and prediction of air conditioning load, which has strong practical advantages. Support Vector Regression (SVR) technology has been successfully applied to wind speed prediction, power load forecasting and other fields, and also has a small part of the application in the air-conditioning direction.

At the same time, the regression prediction method based on the SVM is less efficient than the traditional deductive deduction method on the terms of deriving from training samples to predictive samples, and it is a simplified solution of regression problem. Therefore SVM has the advantage of performance in comparison with the traditional method.

Section snippets

Model of support vector regression machine

The basic idea of the SVR is that: assuming the training sample sets are ${(x_{i}, y_{i}) i = 1, 2, . . ., n} {(x_{i}, y_{i}) i = 1, 2, . . ., n}, | x \in R^{d} and y \in R . |$

A function f (x) is obtained from the training regression of the SVR. $f (x) = (w \cdot φ (x)) + b$ where, w is the weight and b is the intolerance option. ε-insensitive loss function is introduced to generalize the SVM classification principle to the regression estimate problem. $ξ_{ɛ} = {\begin{matrix} 0 & i f | y - f (x) | \leq ɛ \\ | y - f (x) | - ɛ & i f | y - f (x) | > ɛ \end{matrix}$

The problem of finding ω and b is described in the mathematical

Simulated annealing algorithm

Simulated annealing algorithm (SA) was first applied to the combination optimization field by Kirkpatrick et al. [18]. It is a stochastic searching optimization algorithm based on Monte–Carto iterative solving strategy, by reference to the principle of physical metal annealing, and that is, the thermodynamics theory is applied mechanically to the solving strategy and its starting point is based on the similarity between the annealing phase of solid matter in physics and common combined

Data normalization

Air conditioning load data value is greatly fluctuant and it is required to conduct normalization processing for the data before training and prediction and the normalization equation is as follows (10). $\hat{x_{i}} = \frac{x_{i} - x_{\min}}{x_{\max} - x_{\min}}, i = 1, 2 \dots, n$ where, x_i the original data value, x_max the maximum value in the air conditioning load data, x_min the minimum value in the air conditioning load data and ${\hat{x}}_{i}$ the data value upon normalization.

Evaluation index

Parameter optimization must have the corresponding parameter evaluation

Research object and parameter

Data of this paper is from a hotel in Hainan Province, and this hotel collected hourly data from July in 2011 to August in 2012 (9528 h totally). Data of last 2 months (July and August in 2012) was chosen as test sample and data of other 12 months (from July in 2011 to June in 2016) was chosen as training sample.

This Paper respectively conducts 24-h prediction of air conditioning load of July and August in 2012. Respectively SA algorithm and VFSA (very fast Simulated Annealing) algorithm and

Conclusion

With the surge in air-conditioning power consumption [19], traditional air-conditioning load forecasting and traditional support vector machine air-conditioning load forecasting can no longer meet the needs of air-conditioning control. Support vector regression [20] (SVR, Support Vector Regression) has shown excellent performance in all aspects of regression. Based on the research results of other scholars, this dissertation carries out a detailed study and design of the air conditioner load

Funding

This work was supported by the Beijing municipal commission of education technology [grant numbers KM201210005026]; the State key research and development plan [grant number 2I025001201701].

Conflict of interest

None declared.

Declarations of interest

None.

References (20)

S.S.K Kwok et al.
A study of the importance of occupancy to building cooling load in prediction by intelligent approach
Energy Convers. Manage.
(2011)
H. Hahn et al.
Electric load forecasting methods: tools for decision making
Eur. J. Operat. Res.
(2009)
Q. Li et al.
Applying support vector machine to predict hourly cooling load in the building
Appl. Energy
(2009)
A. Selakov et al.
Hybrid PSO–SVM method for short-term load forecasting during periods with significant temperature variations in city of Burbank
Appl. Soft Comput.
(2014)
S.P. Zhang et al.
LS-SVR and AGO based time series prediction method
Int. Equip. Eng. Manage. (English Edition)
(2016)
X. Li et al.
A very fast simulated re-annealing (VFSA) approach for land data assimilation
Comput. Geosci.
(2004)
M. Sakawa et al.
Cooling load prediction in a district heating and cooling system through simplified robust filter and multi-layered neural network
M. Kawashima et al.
Hourly thermal load prediction for the next 24 hours by ARIMA, EWMA, LR and an artificial neural network
A.D. Papalexopoulos et al.
A regression-based approach to short-term system load forecasting
IEEE Trans. Power Syst.
(1990)
L. Xuemei et al.
Building cooling load forecasting model based on LS-SVM

There are more references available in the full text version of this article.

Cited by (40)

Chaos-based support vector regression for load power forecasting of excavators
2024, Expert Systems with Applications
The accurate prediction of digging load serves as a fundamental cornerstone for advancing the development of intelligent and unmanned excavators. Given the complex nonlinear dynamics of digging load, this paper proposes a novel prediction model for excavator load power based on the chaos theory and support vector regression (SVR). The presence of chaos in the dynamic digging load system is detected through phase space reconstruction. SVR is utilized for nonparametric modeling and prediction, with the reconstructed phase space capturing the essential characteristics of excavator load and serving as inputs for SVR. To optimize the hyperparameters, an improved particle swarm optimization (IPSO) algorithm is presented. Excavation experiments conducted under two typical load conditions demonstrate the superiority of the proposed chaos-based IPSO-SVR model in terms of prediction accuracy. This research lays a solid foundation for practical load prediction in industrial excavator settings.
Air conditioning load prediction based on hybrid data decomposition and non-parametric fusion model
2023, Journal of Building Engineering
Accurate prediction of air conditioning load is the pivotal problem of air conditioning optimization control, which is of great significance for reducing building energy consumption. In order to improve the prediction accuracy of air conditioning load, this paper proposes a load prediction method of air conditioning based on hybrid data decomposition (HRD) and non-parametric fusion model (NPFM). First, the historical load data is decomposed into linear term and nonlinear term by using HRD, which are employed to combine the temporal convolutional network (TCN) and long short term memory (LSTM) method to achieve the priori prediction of air conditioning load, and are regarded as the state transition equation. Secondly, the strong correlation index based on correlation screening is used as the observation parameter, and the nonlinear relationship between air conditioning load and real-time operating parameters of air conditioning equipment is mapped by back propagation (BP) neural network, which works as the observation equation, and forming a set of NPFM. Thirdly, the particle filter (PF) algorithm is introduced to correct the prior load prediction, suppress the noise disturbance, and realize the closed-loop control. Finally, the effectiveness of the proposed method is validated by two examples, and the results indicate that the proposed method can provide more accurate and robust air conditioning load prediction compared with existing methods.
Multi-step wind speed prediction based on LSSVM combined with ESMD and fractional-order beetle swarm optimization
2023, Energy Reports
Accurate and stable wind speed prediction can alleviate the uncertain impacts of wind power generation caused by nonlinear characteristics of wind speed, and then improve the reliability of wind power. In this paper, a hybrid model for wind speed prediction based on mode decomposition, parameter optimization and basic prediction model is proposed. First, the extreme-point symmetric mode decomposition (ESMD) is employed to adaptively decompose the denoised wind speed time series into sub-sequences with different frequencies. Second, a fractional-order beetle swarm optimization (FO-BSO) for parameter optimization of the Least squares support vector machine (LSSVM) is proposed. Through benchmark functions and non-parametric statistical test, the advantages of the FO-BSO in accuracy, stability and convergence speed are verified. Subsequently, the ESMD-FO-BSO-LSSVM prediction model is established, and three groups of wind speed datasets with different sampling locations and sampling frequencies are selected for simulation experiments. The results show that the coefficient of determination of 1-step prediction of the proposed model in three datasets are 0.9856, 0.9713, 0.9940, which has 2.43%, 3.38%, 3.08% average promotion than that of 7 comparative models. And the accuracy and stability of ESMD-FO-BSO-LSSVM model in multi-step wind speed prediction have also achieved better performance than 7 competitors.
A novel machine learning-based model predictive control framework for improving the energy efficiency of air-conditioning systems
2023, Energy and Buildings
The dynamic optimization of key setpoints (e.g., supply water temperatures of chillers and cooling towers, or indoor temperature and humidity) can track the efficient performance point of the air-conditioning system (ACS), thus obtaining a great energy-saving effect. The model predictive control is an effective way to precisely adjust these setpoints. However, most of the existing studies focused on simple linear strategies, e.g., experience-based, rule-based, or component-based, which results in unsatisfactory energy efficiency. Aiming at this, the paper proposed a novel machine learning-based model predictive control (MLB-MPC) framework with predictive inputs of disturbances. The framework first employed the grid search and cross-validation to optimize the total energy consumption prediction models (TECPMs), which were built by data-driven models (e.g., multiple linear regression, artificial neural network, support vector regression (SVR), and random forest). And then the optimization performances of each TECPM to the controlled variables were obtained by using different optimization algorithms (e.g., genetic algorithm, particle swarm optimization (PSO), and simulated annealing). Finally, the optimal match of TECPMs and optimization algorithms was achieved by trade-off among prediction accuracies, optimization accuracies, and optimization time. The case studies demonstrated that the optimal match for the MLB-MPC is SVR and PSO as they had the highest prediction accuracy (mean absolute percentage error of 2.5%) and shortest optimization time (41 ms/per) and optimization accuracies with little difference. After adopting the optimal match, the MLB-MPC with predictive disturbing inputs achieved a great energy-saving ratio of 7.1% for the ACS, which was only less than the MLB-MPC with ideal disturbing inputs by 0.4%. This indicated that exact predictions of disturbances (e.g., cooling load and outdoor wet-bulb temperature) are important for the MLB-MPC. The proposed MLB-MPC framework would provide a method for improving the energy efficiency of ACSs.
Qualitative and quantitative assessment of flavor quality of Chinese soybean paste using multiple sensor technologies combined with chemometrics and a data fusion strategy
2023, Food Chemistry
Citation Excerpt :
SVR is the regression analysis of support vector machine (SVM), a learning algorithm, which was originally proposed based on statistical learning theory (SLT) by Cortes and Vapnik (Cortes & Vapnik, 1995). SVM was widely employed in many researches (Tao, Yan, Gao, Sun, & Li, 2019; Xu, Wang, & Gu, 2019) owing to that it could exhibit a good performance in generalization ability, global optimal solution and solving non-linear problem (Han, Huang, Teye, Gu, & Gu, 2014). Data fusion strategy is a means of combining information from different sources to achieve a more accurate result, which focuses on the optimization of the implementation results to obtain more effective information, so as to achieve a more accurate description of the perceived object (Buratti, Malegori, Benedetti, Oliveri, & Giovanelli, 2018; Orlandi et al., 2019).
Multiple sensor technologies including electronic nose (E-nose), electronic tongue (E-tongue), colorimeter and texture analyzer combined with chemometrics and dada fusion strategies were applied to characterize the flavor quality of traditional Chinese fermented soybean paste. Principal components analysis (PCA) was performed to divide the selected soybean pastes into three clusters which was not completely consistent with geographical regions of selected samples. Support vector machine regression (SVR) outperformed partial least squares regression (PLSR) in quantitatively predicting sensory attributes. Additionally, prediction of overall flavor of soybean paste based on data fusion of multiple sensor information, with a correlation coefficient of prediction (Rp) of 0.9636 based on SVR, was better than prediction of E-nose and E-tongue data fusion (Rp = 0.9267). This study suggested multiple sensor technologies coupled with chemometrics can be a promising tool for flavor assessment and characterization of fermented soybean paste or other food matrixes.
A closed-loop data-fusion framework for air conditioning load prediction based on LBF
2022, Energy Reports
Accurate air conditioning load prediction is a key component of intelligent building management system for ensuring energy saving and safe operation of air conditioning system. In order to improve the prediction accuracy, a particle filter (PF) load prediction fusion estimation method based on long short-term memory (LSTM) and back propagation neural network (BP) is proposed. Firstly, spearman correlation analysis is used to select the influencing factors with high correlation as feature input. Aiming at the problem that the original signal is easy to be disturbed by noise and the data features are not obvious, locally weighted scatterplot smoothing (LOWESS) method is used to denoise the data to improve the data quality for further accurate prediction. Secondly, the data-driven air conditioning load state-space representation is established, which takes air conditioning load as the state variable and takes the load features collected by the sensor in real-time as the input variables. Thirdly, combined with the space representation method of air conditioning load based on LSTM-BP, PF is introduced to estimate the air conditioning load by using the fusion model. Meanwhile, the output load value of BP is fed back to the fusion model as the observation value to update the state-space representation of air conditioning load. Finally, two practical cases are used to verify the effectiveness of the method. The results indicate that the proposed method can provide more accurate and robust air conditioning load prediction.

View all citing articles on Scopus

View full text

Application of SVR optimized by Modified Simulated Annealing (MSA-SVR) air conditioning load prediction model

Abstract

Introduction

Section snippets

Model of support vector regression machine

Simulated annealing algorithm

Data normalization

Evaluation index

Research object and parameter

Conclusion

Funding

Conflict of interest

Declarations of interest

Energy Convers. Manage.

Eur. J. Operat. Res.

Appl. Energy

Appl. Soft Comput.

Int. Equip. Eng. Manage. (English Edition)

Comput. Geosci.

Cooling load prediction in a district heating and cooling system through simplified robust filter and multi-layered neural network

Hourly thermal load prediction for the next 24 hours by ARIMA, EWMA, LR and an artificial neural network

A regression-based approach to short-term system load forecasting

IEEE Trans. Power Syst.

Building cooling load forecasting model based on LS-SVM