JAUE2017-094: Impact Analysis of Typical Demand Day Selection on CCHP Operational Optimization

Abstract

Optimizing the configuration and operation of a CCHP system for a whole year leads to very high and 
unfeasible computational time expenses due to high variability of energy demand. To overcome this 
problem, this paper presents a new and creative method to reduce a full year of demand data to a few 
representative days that adequately preserve significant demand characteristics of the yearly consumption 
profiles. Typical demand days are selected based on the use of kmeans clustering algorithm and average 
method, and their ability to resemble the original data is tested by means of mean absolute percentage error 
(MAPE) analysis. CCHP system is optimized with mixed-integer linear programming algorithm (MILP).
In order to select the optimal clustering number so as to confirm optimal number of typical days,
MAPE and annual total cost resulting from actual operation following with different clustering numbers are 
compared. To compare with the proposed method, this paper also imitates traditional method by choosing 
peak load days in each season and calculating 12-day load by classifying load in unit of month by 
proportional sharing method as contrast schemes. A case study of a Qingdao office building is discussed to 
demonstrate the proposed method. The results illustrate that the magnitude of MAPE between actual 
demand load and typical days load can affect actual operational effect. In conclusion, optimal number of 
typical days for actual CCHP operation can obtain very low MAPE and lowest annual total cost. 
In general, the method presented in this paper has important reference to CCHP operational
optimization and can be used in practice.