Determine potential utilisation at time slice level

[alexdewar]

Part of agent investment

[tsmbland]

I think this will go something like this:

---

Before beginning:

• we'll need utilizations in the previous year from the dispatch optimization (done, but I'm not sure we're storing these)
• we'll need to calculate marginal cost for all assets and processes in each timeslice (awaiting equation from Adam) Calculate marginal costs for assets/processes #516
• demand for SVD commodities is provided in the input, but we need to calculate a value for each agent based on commodity portions (included in the loop below, but we could pre-calculate). SED commodities won't have demand to begin with, but I believe this will be calculated as we go through.

---

Then

for commodity (important to run SVD commodities first)
	get agents that service this commodity
	for agent
		calculate demand share for this commodity/agent
		get assets owned by this agent
		for asset
			for timeslice
				calculate potential utlilization (see below)
		get processes in the search space
		for process
			for timeslice
				calculate potential utilization (see below)

	continue with investment...

---

Calculating potential utilization (for single asset/process in a single timeslice)

Inputs:

• demand (single value)
• marginal cost (single value)
• max utilization (single value): for assets this the limit imposed by the capacity of the asset and and the upper availability limit in the timeslice (is is straightforward to calculate this?); for processes I think we assume there's no limit (inf)
• asset marginal costs: marginal cost of every asset in the current year in this timeslice
• asset utilizations: utilization of every asset in the previous year in this timeslice

Algorithm:

1. Sum up the utilizations from assets with lower (or equal) marginal cost
2. Subtract this from the demand
3. Apply max utilization upper limit

[alexdewar]

I've opened an issue for checking for demand loops between commodities: #519

I've marked it as on hold in case we can figure out a better approach before the v0.2.0 release, but if we don't have anything by then, I think we should just prevent users from creating these loops in the first place.

[alexdewar]

I think we should do #519 before this, but we can do this in parallel with #516 with a placeholder function for marginal cost that just returns zero (or whatever).