*2.3. Control*

While assessing lighting class selection for a given street segmen<sup>t</sup> there are certain measurable coefficients that influence this process. One of them is traffic intensity, or volume. Depending on the number of vehicles passing, there could be up to three different lighting classes assigned. Varying lighting classes lead to different dimming settings of the luminaires, which results in energy savings. While assigning a lighting class, the maximum traffic volume should be considered. For such a class, power levels of the luminaires should be selected accordingly by means of photometric calculations. If information about a current traffic volume is available, then changing the classes for a given street segmen<sup>t</sup> can be performed efficiently [17,18], which leads to energy savings [19] and CO2 emission reduction.

Table 2 presents the 24-h lighting class assignment structure and the comparison of obtained savings for various control approaches. The table is made for the arterial road with a dominant lighting class ME2.


**Table 2.** Different lighting control strategies and corresponding energy saving (lighting classes according CEN/TR 13201-1:2004). Any lighting class change is triggered by a varying traffic volume.

**Remark 1.** *It should be noted that the new release of the standard CEN 13201 was issued in 2014 but its Polish localization was published two years later, in 2016. For that reason, the calculations presented below was initially made in accordance with CEN 13201:2004.*

The analysis was based on 100 luminaires and one year worth of traffic data. The *calendar* control strategy assumes that ME2 is assigned regardless of an actual traffic intensity and lighting is turned on at sunset and turned off at sunrise, totaling 4292 on-hours. The *dynamic* assumes reading traffic intensity every 15 min. The lighting classes are adjusted accordingly which leads to the 34% energy saving comparing to the *calendar* approach. ME2 is used 27% of the time, while the lower class, ME3b, is assigned 14%, and ME4a 59%.

In some situations, a statistic approach is also used. In this case, the control is not based on actual traffic volume but on statistical analysis of its historical data records—separately for working days, holidays, Saturdays, and Sundays. Even though it is commonly used, it very often leads to lighting class violations because of a high traffic variability. An actual traffic volume can surpass the statistics based on historical or incomplete data. An example for the aforementioned arterial road is also presented for comparison. For more diverse road and street infrastructure, actual savings obtained by application of dynamic control are expected to be at the level of 27% [20].

The above findings regard CEN/TR 13201-1:2004. The current release of the standard (since 2014) alters number of regulations, lighting class names and their assignment methods, among others. It also introduces a notion of an adaptive control. This gives greater flexibility for dynamic control applicability. A comparison of energy savings achieved thanks to application of a dynamic control, made for a slightly larger and more representative area with variable, high volume traffic, is given in Table 3. There are 210 luminaires installed, with the total power of 26,933 W. The total annual operation time is 4292 h. In this table, we compare the performance structure for both 2004 and 2014 releases of CEN/TR 13201-1. As can be seen, following the 2014 revision leads to significant savings, from 35% to 46%, which gives the motivation to upgrading a lighting infrastructure performance to the new release of the standard. It is mainly due to introduction of maximum traffic capacity indicator on which assignment of lighting classes is based now. The M2 class is assigned 8% of the time only.

**Table 3.** Dynamic lighting control energy saving for ME2 and M2 base lighting classes; comparison between CEN/TR 13201-1 releases 2004 and 2014.


The energy saving is also influenced by the traffic intensity parameters, in particular how often the data are read from sensors and how wide is the averaging time window [20].
