Data collection for calculations and result processing

Four different CFB boilers and one BFB boiler were chosen to the calculation case matrix.

These boilers use distinct fuel mixtures and vary also structurally, which together mean also diverse experiences of fouling issues in them. The boilers were selected based on specific and documented combustibility campaigns, guarantee test or probe test periods that had taken place in them. This was done to be able to link the calculation results to detailed fuel mixture information that would not be accessible afterwards from regular operation periods, when only basic systematic analyses are conducted on the fuel at the most.

The measurement data was retrieved from the distributed control system (DCS) log files and processed to fit the in-house calculation tool. Alongside the required flows and tem-peratures, other operational parameters were also collected from the raw data, including calculated fuel loads, soot blowing steam flows, and pressure changes on the flue gas side

of the ducts. A summary of the measurements can be found in Table 6. Original data had one minute measurement intervals in each boiler, but since the timespan of the calculation cases extended over several days, a 10-minute interval was concluded to be sufficiently representative. In addition, the longer intervals had lower required computation time as well, and so the one minute data was averaged into 10-minute intervals before calculation (work package 3. in Figure 13).

Table 6 Selected DCS and process modelling data for the calculations

Measured variables and constants Raw unit in DCS

or modelling Source Flue gas temperatures in and out (for each

heat exchanger) °C DCS

Steam or feedwater temperatures in and

out (for each heat exchanger) °C DCS

Steam or feedwater mass flow (for each

heat exchanger) t/h or kg/s DCS

Soot blowing steam flow t/h or kg/s DCS

Pressure drops in flue gas side Pa or mbar DCS Main steam parameters °C, bar and kg/s DCS

Heat exchange surface area m² Boiler design

data Heat exchanger tube and element pitches mm Boiler design

data Reference value for overall heat transfer

coefficient W/m²K Boiler design

data

Mass flow of steam was measured only after all superheaters by default, e.g. from the main steam tube before the turbine. Water spray flows added to this main steam in attem-perators after primary and secondary superheaters were acknowledged by subtracting them respectively from the main steam mass flow values. In most cases, the feedwater flow measurement before economizer took place before the separation of water to the attemperators, so the total flows to attemperators had to be subtracted from the feedwater flow measurement as well to give precise input for the fouling calculations of economiz-ers.

Geometrical data, including tube and element pitches, and the reference heat transfer co-efficients for each heat exchanger element were gathered from the designed process mod-elling data of each boiler. The elementary composition of the flue gas, that was utilized in the molar mass calculations for determining the specific heat capacities of the flue gas, was also fetched from the process modelling results, since it was thought to be a suffi-ciently constant parameter in each boiler’s case individually, despite how the fired fuel feed changes the composition slightly.

Air preheaters were left out of examination because of assumed significant inaccuracy related to overly high equivalence of temperature values on the hot and cold, or flue gas and clean air sides of air preheaters. For the evaluated superheaters and economizers, the measurement of steam temperatures were assumed to be more precise than possibly var-ying temperatures of flue gas, but the measurement of the volumetric flue gas flow was considered to be even more uncertain than that of the temperatures. Therefore, the flue gas flow was set as the variable to be calculated in the in-house program. The evaluated test matrix is shown in Table 7, where the abbreviations SSH, ECO and PSH refer to secondary superheater, economizer and primary superheater respectively.

Table 7 Test matrix of the fouling examinations

Boiler case number Boiler type Examined heat exchangers

1 CFB SSH, ECO

2 CFB SSH, ECO

3 CFB SSH, ECO

4 CFB PSH, ECO

5 BFB PSH

Despite the similarity of the evaluated superheaters appears to be a bit low in Table 7, it should be noted that since the phenomenon to be validated is the outer surface fouling of the tube elements, the flue gas side temperature range at the given heat exchangers is of higher interest than the actual superheating stage of the steam. The studied boilers have varying superheater and economizer configurations, meaning that a universal definition of the flue gas temperature range by each superheating or feedwater heating stage cannot be made. A refined categorization based on distinct temperature ranges is presented in the results section.

Research targets included estimating the applicability of the heat transfer method and the correlation strength with selected operational parameters, and so basic statistical correla-tion analyses were conducted. Uncertainty could have been caused by various issues.

Continuous flue gas temperature measurement is a good example of the calculation tool input values that have high susceptibility for measurement error. As the thermoelements are typically rather near the duct walls, the temperature difference between the measure-ment point and the centerline of the flue gas flow facing a superheater or economizer can be even 100 °C. Varying flow patterns make implementation of a constant correction factor to this unavoidable error challenging. Partly due to these measurement errors, ra-ther low attention was paid on the raw and discrete calculation values themselves. Instead,

time-dependency and the rate of the calculated thermal resistance values by each test point of the combustion campaigns were studied further.

Time series analysis on the calculated values was done by line fitting with linear regres-sion to catch the rate of thermal resistance build-up. Pearson’s coefficient of determina-tion was also calculated to evaluate the differences in the calculated fouling rates; some cases were found to express almost asymptotic behavior in thermal resistance build-up, whereas others showed steadily rising resistance values until the following soot blowing pulse. Microsoft Excel was used in the linear regression and Pearson product moment correlation coefficient (PPMCC) calculations, using LINEST and RSQ functions. It should be noted that while PPMCC is a measure of the applicability of the linear trendline formed over specific data points, it cannot analyze properly the nature of the deviation from a perfect line. In other words, a similar PPMCC value can be formed for a data set with considerable deviation between adjacent data points but strongly linear overall trend, and for a clearly non-linear set with minimal deviation.

Several comparisons using calculated thermal resistances were conducted in the compu-tational part. One of these was comparison against fired fuel mixture. To gain comparable data for the fuel mixture examinations, the effect of boiler load change was formulated into alternative fouling rate figures, mainly because of one considerable load change in boiler 1. The steam power adjustment was performed for the boiler cases by dividing thermal resistances by the ratio of corresponding steam flow measurements to average steam flow, as per equation

𝑟_𝑠.𝑎. = ^𝑟

( ^{𝑃𝑠𝑡𝑒𝑎𝑚}

𝑃𝑠𝑡𝑒𝑎𝑚,𝑎𝑣𝑔), (13)

where r expresses an original thermal resistance value, Psteam is the calculated steam power value at the timestamp of r, Psteam,avg means average steam power from the whole campaign period and rs.a. is the resulting steam power adjusted r value. Effect of this equation is discussed in Chapter 5.5.