Low energy climate control - the sequel
This sequel to the design project for new store rooms for the National Museum of Denmark was published as:
Tim Padfield, 'Low Energy Climate Control in Museum Stores - a postscript', Proceedings of the ICOM-CC Conference, Edinburgh, September 1996, vol 1 pp 68-71.
A large museum storage building whose relative humidity was controlled by warming to reduce relative humidity showed a violently oscillating temperature with a period of about a week, as predicted by a computer model. The principle of humidity control by heating a building when the RH rises above the set point does not work in buildings that have a small air exchange rate and are filled with water absorbent materials. Such buildings must be controlled by dehumidification or by heating at a slowly varying energy input, depending on the time of year rather than on the currently measured RH.
At the ICOM-CC meeting in Dresden in 1990 Poul Jensen and I presented a paper (1) in which we analysed various methods of relative humidity control for museum stores. The unusual feature of this analysis was the inclusion of the stored objects in the calculation of heat and moisture variation. The analysis was a contribution to the planning of new storage rooms for the National Museum. It was made entirely on a computer, without the possibility of experimental verification on the real buildings at that time.
Two of the methods described in that paper, warming to constant relative humidity (RH) and dehumidification to constant RH, have now been tested in one of these storage buildings. This building (fig.1) has an average height of 7.5 metres and a ground plan of 1700 square metres, divided into three halls of equal size. Its walls are of lightweight concrete, 250 mm thick, clad with tongue and groove timber painted black. The space between concrete and wood is filled with 50 mm of insulation. The roof is black tar paper on insulating panels. The stored collection is very varied but includes many large wooden objects. The ventilation system gives one air change per hour. It is provided with heaters to control the relative humidity.
Fig.1 A view from the south of the National Museum's storage building for large objects, in Ørholm, north of Copenhagen. The building is 60 metres long and 24 metres across the gable end. The wall is 5 m high. The walls are of expanded concrete blocks clad with wooden boarding painted black. The roof is of tar paper on insulating panels. The sectioned south aisle shows the pattern of air circulation which transfers moisture evaporated from the heated south wall and causes a high relative humidity at the RH sensor shown on the north wall. The process is described in detail in the text.
One prediction of the computer analysis was that a room filled with moisture absorbent materials such as paper, wood, leather and textiles, cannot reliably be controlled by heating to constant relative humidity, because raising the temperature to counteract a RH rising above the set point will lead to a further rise in RH, through moisture release from the stored objects (see the note below). The process runs out of control because of this 'positive feedback'. The relative humidity remains only slightly above the set point but the absolute humidity of the air rises with the temperature, fed by the drying objects. The runaway heating eventually stops because there is always some infiltration of outside air. This air will have a much lower moisture content than the inside air, even if it is raining, because its temperature is lower.
The RH will first fall below the set point when the low moisture content of the air coming in from outside compensates for the water contributed by the objects, which must draw water from ever deeper layers. Then the whole process goes into reverse: the temperature falls, the objects absorb water, further reducing the RH. The cooling eventually stops when the outside air, now entering a cooler building, contains more water vapour than the inside air. The cycle then repeats.
Several conservators have expressed scepticism over the possibility that this process can occur outside a computer's memory, pointing out that RH control by a humidistat that calls for warmth when the RH increases over a set point, is used successfully in many places, particularly in historic buildings.
This scepticism was shared by the building committee. It decided to continue with the planned climate control method: warming to constant relative humidity, using outside air for cooling. Our advice was, however, partly heeded in later buildings, which have full air conditioning, scaled down to take advantage of the stabilising influence of the stored objects.
It seems that the heating to constant RH soon proved unstable, even though there were few objects in the vast interior. A dehumidifier was installed but the authors of the computer simulation were not told and were therefore denied an opportunity to check the detailed predictions of their model against reality as the building steadily filled with objects.
Fortunately, for us, an opportunity came two years later when the dehumidifier broke down. The original control program had been left in the silicon memory that steers the air conditioning equipment. The heating system again took over the job of relative humidity control. The result for a two month period is shown in fig.2. The room temperature shows cycles of heating and cooling over a period of about a week. The steeply rising and falling dotted line is the temperature of the air blown in from the air conditioning unit. Notice that the rise in temperature does not cause a fall in RH.
Fig.2 The climate in the building during April and May 1995. The line marked 'Inlet T' is the temperature of the air blown in from the climate control equipment.
Figure 3 shows the computer model results for the same period in the 'Test Reference Year', using exactly the same data and calculations as in the 1990 paper. The pattern of episodes of heating for about five days, accompanied by a relative humidity stubbornly above the set point, followed by rapid cooling, is strikingly similar in both graphs. The cooling periods are more abrupt in the model because the temperature was actively driven down, whereas in the real building the heating was simply shut off and fresh air let in, if its temperature and dew point were suitable. The cooling pattern is therefore not identical between model and reality.
Fig.3 Corresponding data from the computer model, as described in detail in reference 1. The chaotic cycling of the inside temperature shows a similar pattern to fig.2 but the steepness of the cooling part of the cycle is different, because in the computer model the air was forcibly cooled. In the real building the heating is simply switched off and outside air let in.
Notice that in both graphs the trigger that starts a cycle of instability is a rise in the outside temperature. This temperature rise penetrates the insulation and slightly heats the objects, which then desorb moisture to send the inside RH over the set point.
The predicted good climate regulation by dehumidification resumed as soon as the dehumidifer was repaired. This is shown in the data for June and July, in fig.4. The dramatic temperature cycling has stopped but now we see a daily cycle in RH in the building, out of phase with the daily fall in outside RH during the morning and much too large to attribute to the reaction of the stored objects to the rising temperature.
Fig.4 The climate in the building during June and July 1995, after the dehumidifier had been repaired. Notice the less dramatic variation in inside temperature, which follows the outside temperature. The outside RH (dotted line) is superimposed on the inside RH to show that they vary out of phase. The inside RH follows the outside temperature, probably because of evaporation of water into the interior from the cellular concrete wall as it is warmed by solar radiation absorbed by the black wooden cladding (see text and Fig. 1)
I attribute this phenomenon to warming of the concrete block wall by heat transmission from the black timber cladding, which is fully exposed to the south. Water evaporates into the building from the porous concrete, which is coated on the inside with porous lime wash. Such a wall is a very efficient humidity buffer, at its surface temperature, as shown by a recent study of a church with a similarly porous wall surface (2). This warm air, with a higher moisture content than the air in the centre of the room, rises convectively to the roof and convects down again past the sensor, which is on the inner wall forming the north side of the south aisle. The mezzanine floor does not meet the walls and serves to keep the convective circulation close to the wall surface. The roof is coated on the inside with relatively impermeable acrylic paint, so the air retains its high water content as it descends past the cooler north side of the pitched roof. On the downward part of the cycle the air reaches temperature equilibrium with the room but not moisture equilibrium, so the sensor sees a high RH in the air streaming past it. The process is sketched in figure 1. Indirect support for this explanation is provided by the much more stable environment in the north aisle, which shares the same ventilation system but receives no solar radiation on its walls.
Radiant heating was not included in the computer model. I had not anticipated that the entire building would be painted black!
The short answer is that it works when the air exchange in the building is fast and when there is little humidity buffering from the building structure and furniture.
The RH in a sealed container rises with temperature if there is more than about 100g of small pieces of unvarnished wood per cubic meter of air. In a leaking case, such as a large building, a larger quantity of moisture absorbent material is needed to ensure that the RH rises with temperature. Furthermore, the moisture in the material must be readily available.
In the building described here there is an abundance of bare wood and cardboard containers, the walls are lightweight concrete with a significant buffer capacity, coated with a porous paint. there are no windows and the air conditioning is designed to stop sucking in outside air above a certain inside temperature, thus worsening the situation. In the historic houses that are warmed to constant relative humidity there is a much greater natural ventilation, through windows, doors and chimneys. The availability of moisture from furnishing and panelling is usually surprisingly small. Wood panelling and walls coated with oil paint or plastic paint have negligible buffer effect. In such houses the RH falls as the temperature rises and humidistatic control by heating is possible.
One should not, however, be complacent about extending this humidity control principle to storage rooms in buildings specially designed, or adapted for storage. In these buildings the natural air exchange is generally smaller and the risk of a rising RH with heating is significant. It is not practical to assume a constant or predictable air exchange in a building, because the air exchange is dependent on wind and on temperature gradient.
A computer model predicting an unstable climate in store rooms designed to be heated to attain constant humidity has been confirmed by the behaviour of a real, large building. Dehumidification is a better method for relative humidity control in store rooms, particularly those with a small natural air exchange rate. Dehumidification also results in a lower average temperature, which is good for the collection.
If relative humidity control by heating is preferred, there are several ways of preventing runaway heating. The feedback loop, in which the RH sensor relays ambiguous information to the control electronics, cannot be used. The temperature can be adjusted according to the time of year, to hold the temperature about 5C above the expected temperature, taken from averages over many years. The same average excess temperature can be achieved by heating with constant energy, after establishing by trial and error how many watts are needed. If the building is well insulated, well sealed and massive, with a very slow response to changes in the outdoor temperature and humidity, the problem can be avoided by installing a very low powered heating system, so that the temperature can only rise very slowly, ensuring that even a slow air exchange is enough to prevent a rise in RH (3).
In an isolated volume of air (or just space with water vapour) the relative humidity falls as the temperature rises. If one adds some absorbent material, such as cellulose, this change of RH will be partly compensated by buffering from the material: as the RH changes the material will absorb or emit water vapour to the space around it. However, if there is very little space compared to absorbent material, the RH will actually rise with temperature, overcompensating the change expected in an empty space. This is because of the thermodynamics of the affinity between water and cellulose. All absorbent materials show this phenomenon to some extent.
Figure 5. The curve A-A shows the way the RH changes with temperature in an empty room. As one adds absorbent material this curve is flattened until, at about 100 g cellulose per cubic metre of space, there is no change of RH with temperature (line B-B). If one adds more cellulose, the trend reverses, because the chemistry of the cellulose-water vapour interaction is now dominating the space (line C-C)
1. Tim Padfield and Poul Jensen, 'Low Energy Climate Control in Museum Stores', Triennial meeting of the International Commission of Museums, Committee for Conservation, Dresden 1990, pp 596-601.
2. Bent Eshøj and Tim Padfield, 'The use of porous building materials to provide a stable relative humidity', Triennial meeting of the International Commission of Museums, Committee for Conservation, Washington 1993, pp 605-609.
3. This suggestion is based on climate data from the building housing the State Archive of Schleswig-Holstein, Germany, collected by Lars Christoffersen, Birch & Krogboe K/S, Virum, Denmark.
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 3.0 License.