Study On Dynamic Benchmarking Methods For Computer And Air Conditioning Energy Consumption Evaluation In Office Buildings

As one of the important methods to reduce building energy consumption,building energy benchmarking has been widely used in practice.In order to provide real-time evaluation of the energy consumption in the process of building operation and targeted energy-saving suggestions to users,it is necessary to establish corresponding dynamic energy benchmarking based on the characteristics of different energy-related behavior.However,so far only few studies have been conducted to contribute to energy benchmarking methods for different types of energy-related behavior.Moreover,they are established based on historical data without considering the changes of occupancy,leading to the difficulties in dynamic evaluation and thus cannot ensure the rationality and effectiveness of benchmarking.Taking office buildings as an example,this paper establishes the dynamic benchmarking method for computer and air conditioning energy consumption evaluation by considering the characteristics of different types of energy-related behavior and the changes of real-time occupancy information.Regarding the dynamic benchmarking method for computer energy consumption evaluation,a cluster analysis method is first adopted to identify different occupancy modes and computer energy consumption modes.Then,the information entropy method is used to analyze the correlation between different occupancy modes and computer energy consumption modes.Based on the results of information entropy analysis,a dynamic benchmarking model by using neural network algorithms is established for computer energy consumption evaluation.Regarding the dynamic benchmarking method for air conditioning energy consumption evaluation,after identifying the different indoor and outdoor temperature difference patterns by cluster analysis,a decision tree model is established to evaluate the turn on/off status of air conditioners.Based on the turn on/off status,this paper further produces the sample data under different indoor thermal comfort conditions simulated in the EnergyPlus environment.Then,the cluster analysis method used to identify the indoor and outdoor temperature difference modes and the air conditioning energy consumption modes under the different thermal comfort sample data.Based on the results of cluster analysis,the Pearson coefficients are method is used to quantify the correlation between different indoor and outdoor temperature difference modes and air conditioning energy consumption modes.Finally,for different Predicted Mean Vote(PMV)intervals,this paper adopts the gaussian process regression algorithm to separately establish dynamic benchmarking models to evaluate air conditioning energy consumption and energy-saving potentials under the different thermal comfort conditions and offers corresponding energy saving suggestions.The main conclusions are as follows:(1)The cluster analysis technology can effectively identify the characteristics of different occupancy behaviors and computer energy consumption in different periods(such as the holistic variation trend,variation range,peak time and duration),and thus can obtain typical variation modes of corresponding clusters.Based on the pattern recognition,the information entropy technology can also effectively quantify the correlation between different occupancy modes and computer energy consumption modes,and extract the sample data for the training of the dynamic benchmarking model of computer energy consumption.(2)The proposed dynamic benchmarking method for computer energy consumption evaluation based on typical occupancy models can dynamically adjust the model prediction baseline according to the changes of real-time occupancy,so as to more accurately and reasonably evaluate the computer energy consumption.The RMSE,MAE and R-Squared values of the computer benchmarking model trained in this paper are 0.047,0.036 and 0.879,respectively.And the model verification results show that the relative error of the validation data are within 15%at 129(90%)time points.The above results show that the computer benchmarking model has good accuracy in predicting computer energy consumption.The variation range of the daily cumulative energy consumption waste of the studied office is from 1.13 to 2.43KWh,and the average value is 1.78 KWh.In the practical application process,the user can judge the energy consumption of the computer at a certain time by comparing the predicted baseline value with the actual computer energy consumption value,and can provide relevant energy consumption suggestions based on the energy consumption characteristics and modes of different groups of people.(3)The cluster analysis technology can effectively identify the characteristics of different indoor and outdoor temperature difference and air conditioning energy consumption in different periods(such as the holistic variation trend,variation range,peak time and duration).and thus can obtain typical variation modes of corresponding clusters.Based on the pattern recognition,the Pearson coefficients technology can effectively quantify the correlation between different indoor and outdoor temperature difference modes and air conditioning energy consumption modes,and extract the sample data for the training of the dynamic benchmarking model of air conditioning energy consumption.(4)The evaluation method based on the decision tree algorithm can effectively evaluate the status of air conditioning in different periods of indoor and outdoor environment.The prediction accuracy of the decision tree model is 94%,and the prediction accuracy of the model to the validation data is 95%.The results show that the prediction accuracy of the decision tree model is acceptable.The evaluation results of the test data show that 3 days of sample data have air conditioning energy consumption waste,which are 6.02kwh,8.37kwh and 10.43kwh,respectively.Therefore,in practical applications,it is necessary to further evaluate the energy consumption of air conditioners in the operation process under the premise of judging whether the air conditioner needs to be turned on at a certain time.(5)The proposed dynamic benchmark evaluation method of air conditioning energy consumption can effectively evaluate the energy use of air conditioners during operation and distinguish the difference of energy saving potentials of air conditioners under different thermal comfort conditions.For the PMV interval value at[-1,1](Model 1),the MSE,R-Squared and MAE values of the trained model are 0.170,0.893 and 0.102,respectively.The validation results of this model show that the RMSE error is within 15%at 82%of time points in the validation set.For the PMV interval value at[-2,2](Model 2),the MSE,R-Squared and MAE values of the trained model are 0.174,0.873 and 0.102,respectively.The validation results of this model show that the RMSE error of 81%of time points in the validation set is within 15%.The above results show that the established benchmarking model with PMV values in the interval of[-1,1]and[-2,2]has good accuracy and stability in predicting air conditioning energy consumption.For this office,the cumulative daily energy waste of test set 1 evaluated by Model 1 varied from 2.31 to 8.04KWh,with an average of 4.43 KWh.In model 2,the variation range of cumulative daily energy waste in test set 2 was 2.379.14KWh,with an average of 4.59 KWh.For the same sample data in the two test sets,the total cumulative energy saving of the evaluated sample data in Model 2 is 9.33 KWh more than that in Model 1,with a relative increase of 30%.This indicates that the energy saving potential of air conditioning operation is improved correspondingly through the adjustment of adaptive behaviors of users.In practical applications,users can select the corresponding benchmarking model according to the specific thermal comfort requirements to evaluate the energy consumption during the operation of air conditioners and provide corresponding energy consumption suggestions,so as to further reduce the energy consumption of air conditioners.