# Temperature distribution and evolution in a greenhouse

I am trying to mathematically model the temperature distribution and evolution in a greenhouse, which conserves heat due to the greenhouse effect. Here is a transverse schematic ($$H$$ is the height of the box, $$l$$ its width and $$L$$ is length but is not represented in this 2D schema) of the situation I'm trying to solve:

There are several radiative energy flows (visible $$F_d$$ from the sun, infrared $$F_i$$ from the atmosphere, infrared $$F_s=\sigma T_s^4$$ from the black soil and infrared $$F_p=\sigma T_p^4$$ from the plastic/glass wall). The plastic/glass absorbs infrared but not visible light, it is why $$F_d$$ goes directly into the black soil in the schema. There are also convective flows inside the greenhouse due to the airflow $$Q$$ but also due to natural convection between the two plates of different temperature ($$T_s > T_p$$). Adimensionnal numbers and correlations can be used to deduce the convective factor $$h$$.

For the moment, I made the hypothesis that $$F_d$$ and $$F_i$$ were constant. I also have considered a stationary situation where the energy does not accumulate, the outgoing energy is thus equal to the incoming energy. By doing such energy balance locally (air and soil) and globally, I have obtained the following algebraic system of non-linear equation, with three unknows $$P$$, $$T_s$$ and $$T_p$$, where $$P$$ is the heat given to the air inside the greenhouse:

$$\begin{array}{r c l} P = h(T_p-T) + h(T_s-T) \\ F_d + F_i = P + \sigma T_p^4 \\ F_d + \sigma T_p^4 = \sigma T_s^4 + h(T_s-T) \end{array}$$

Using the Newton-Raphson method, I was able to solve this non-linear system, but I would now like to refine my mathematical model. So here is my two problems:

$$F_d(t)=A\sin(\pi (t-t_r)/(t_s-t_r))$$ is not constant (where $$t_r$$ and $$t_s$$ are respectively the sunrise and sunset time). The hypothesis of a stationary problem is therefore erroneous. I tried to modify my equation system but I'm stuck because I don't know where to put the time derivatives and the appropriate constants (heat capacity, volumic mass ?). Once I have the differential equation system, I think I know how to proceed, using Crank-Nicolson discretization method and returning to an algebraic system that can be solved by Newton-Raphson.

In addition to being able to have the temporal evolution of temperatures $$T$$ (air), $$T_s$$ (black soil) and $$T_p$$ (plastic/glass plate), I would also like to have their spatial distribution. What I mean is the value of $$T(z,t)$$ for all position $$z$$ and for all time $$t$$, knowing the initial condition $$T(z,t_0)$$ for all $$z$$, and assuming a constant air flow $$Q$$. I guess because of the airflow $$Q$$, there will be a air transport equation involved, but I'm not sure how to include these new equations.

I will be infinitely grateful to anyone who can help. It is a personal project that I have set myself for educational purposes, I am not an expert but I try as best I can to carry out this project.

• I think I know how to proceed, using Crank-Nicolson discretization method and returning to an algebraic system that can be solved by Newton-Raphson. Newton-Raphson method is used to find the root of a function while Crank-Nicolson is a method of solving PDEs, so I think you have some misunderstandings to clear up. – Kyle Kanos Jan 9 at 11:00
• Crank-Nicolson discretizes a differential system that becomes an algebraic system with unknowns $x(t+\delta t)$. If linear, LU factorization can be used to solve it. In our case, it is not because $F(x)=0$ is a non-linear system, where $x(t+\delta t)$ is a vector of $3$ unknowns. It can be solved using several successive applications of the recurrent equation: $x_{n+1} = x_n + J_F(x_n)^{-1} F(x_n)$, where $J_F$ is the Jacobian matrix of the function $F$, will converge towards the exact solution $x(t+\delta t)$. It is the well-known Newton-Raphson method. – A. Reagan Jan 9 at 12:11
• It's also well-known that the "Thomas algorithm" solves the problem without resorting to root-finding... – Kyle Kanos Jan 9 at 12:13
• Thomas algorithm can solve linear system when the matrix is tridiagonal, the system I want to solve is non-linear due to the unknown temperature $T$ that appears in the system with an exponent 4 due to Stefan-Boltzman law. – A. Reagan Jan 9 at 12:18