Simultaneous charge and discharge

[Ethan-Russell]

I am still seeing simultaneous charging and discharging with the full size grid. Need to get creative with a way to limit this with LP hopefully without changing the objective.

[sallyrobson]

I am still seeing simultaneous charging and discharging with the full size grid. Need to get creative with a way to limit this with LP hopefully without changing the objective.

I'm thinking this might be a good question for Ray if we want to reach out, I don't think we've used up our time for him.

[Ethan-Russell]

A brain dump.

If we say that storage is on the generator side:

power discharged should be added to pgen_bus
power charged should be subtracted from pgen_bus

That is theoretically what is happening in line 456-457 of Storage.jl

Ok. Let's find an example of when things are messed up.

julia> out_path = raw"L:\Project-Gurobi\Workspace3\E4ST_Output\baseline_testing\bar_hom\230906_140425427"

julia> data = read_processed_results(out_path)

julia> pdischarge = get_raw_result(data, :pdischarge_stor);

julia> pcharge = get_raw_result(data, :pcharge_stor);

julia> p = pdischarge .* pcharge;

julia> count(>(1), p)
1254 # This means there are 1254 times for which there is simultaneous charging & discharging

julia> findmax(p)
(1.1764705813597874e6, CartesianIndex(209, 1, 43))

julia> day = get_table_col(data, :hours, :e4st_day)[43]
12

julia> hr_idxs = get_hour_idxs(data, :e4st_day=>day)
2-element Vector{Int64}:
 43
 44

julia> pcharge[209, 1, hr_idxs]
2-element Vector{Float64}:
 1176.4705851141678
    9.690407329034632e-7

julia> pdischarge[209, 1, hr_idxs]
2-element Vector{Float64}:
 999.9999968087767
   1.3619506173416022e-6

julia> 999.9999968087767 / 1176.4705851141678
0.849999999542474 # Round trip storage efficiency of 85%

# Yup, we are charging a lot here.  First thing to check is LMP's
julia> get_table_col(data, :storage, :lmp_e)[209][ 1, hr_idxs]
2-element Vector{Float64}:
 45.72981667241463
 19.777874344097867

# Just to make sure we're getting the correct LMP here:
julia> get_table_col(data, :bus, :lmp_elserv_preloss)[get_table_col(data, :storage, :bus_idx)[209]][1, hr_idxs]
2-element Vector{Float64}:
 45.72981667241463
 19.777874344097867

# Rats.  Positive LMP means that increasing the power served would increase the objective, so I don't know why we're seeing this.

# Maybe let's take a look at the `pgen_bus` value
julia> pgen_bus[get_table_col(data, :storage, :bus_idx)[209], 1, hr_idxs]
2-element Vector{Float64}:
 1999.2075872742473
    8.155446399178975e-7

julia> gen_idxs = get_table_row_idxs(data, :gen, :bus_idx=>stor_bus_idx)
5-element Vector{Int64}:
  8227
 10024
 18493
 20760
 20761

julia> pgen_gen[gen_idxs, 1, hr_idxs]
5×2 Matrix{Float64}:
    4.18739e-8  3.86825e-7
 2175.68        0.0
    8.4827e-8   4.22689e-9
    1.18931e-8  1.34263e-8
    1.56859e-8  1.81568e-8

# Ah, all the generation is from a single generator, 10024.  Let's check out its cost of operation:
julia> obj_coef = get_raw_result(data, :obj_coef);
julia> obj_pgen_gen = obj_coef[:pgen_gen];
julia> obj_pgen_gen_unweighted = unweight_hourly(data, obj_pgen_gen)
julia> obj_pgen_gen_unweighted[10024, 1, hr_idxs]
julia> obj_pgen_gen_unweighted[10024, 1, hr_idxs]
2-element Vector{Float64}:
 -22.38
 -22.38

# That looks like an IRA ptc value.
julia> gen.pcap[10024]
1-element view(::Matrix{Float64}, 1, :) with eltype Float64:
 2999.9999999493302

julia> gen.af[10024][hr_idxs]
2-element Vector{Float64}:
 0.725226058959961
 0.0

julia> gen.pcap[10024] .*  0.725226058959961
1-element Vector{Float64}:
 2175.6781768431356

# Ah.  So there is a generator with maxed out capacity and generation with negative gen cost

# Let's check and see why we wouldn't be discharging after - could be that e0 is maxed out
julia> e0 = get_raw_result(data, :e0_stor)
julia> e0[stor_idx,1]
3999.9999976746653

# Yup.  Seems like our yearly e0 is hurting us a bit.

julia> sum(weight_hourly(data, pdischarge) .* (pdischarge .* pcharge .> 1))
1.950544207651207e7

julia> sum(weight_hourly(data, pdischarge))
7.695451029947007e7

So in conclusion, our initial stored energy in the battery, e0, is sometimes getting maxed out to the point where it cannot generate more

[Ethan-Russell]

Ok so we need the following:

• e0 should become a daily variable because our hours representation isn't that sequential.
• (optional) constrain pcharge + pdischarge to be <= max of (pcharge_max and pdischarge_max)

[Ethan-Russell]

After making those two changes, we are still seeing simultaneous charging and discharging, albeit less frequently.

Let's take a look at another case

julia> data = read_processed_results("L:/Project-Gurobi/Workspace3/E4ST_Output/baseline_testing/bar_hom\\230914_162839691");

julia> pcharge = get_raw_result(data, :pcharge_stor);

julia> pdischarge = get_raw_result(data, :pdischarge_stor);

julia> p = pdischarge .* pcharge;

julia> count(>(10), p)
790

julia> findmax(p)
(345261.40185765864, CartesianIndex(218, 1, 36))

julia> stor_idx = 218
218

julia> hr_idxs^C

julia> day = get_table_col(data, :hours, :e4st_day)[36]
8

julia> hr_idxs = get_hour_idxs(data, :e4st_day=>day)
2-element Vector{Int64}:
 35
 36

julia> get_table(data, :hours, hr_idxs)
2×14 SubDataFrame
 Row │ hour_id  src_file                           description                        hours    year   month  day    hour   season   e4st_day  hrs_per_day  hr_idx  e4st_day_hour_duration  ozone_season 
     │ String3  String                             String                             Float64  Int64  Int64  Int64  Int64  String7  Int64     Int64        Int64   Int64                   Int64        
─────┼──────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────────
   1 │ h35      L:\\Project-Gurobi\\Workspace3\\…  us-can: load weighted average of…  76.9171   2008      8      6     17  summer          8            2      35                       4             1
   2 │ h36      L:\\Project-Gurobi\\Workspace3\\…  us-can: load weighted average of…  76.9171   2008      8      6     13  summer          8            2      36                       4             1

julia> pcharge[stor_idx, 1, hr_idxs]
2-element Vector{Float64}:
 656.072149189513
 615.7875374234399

julia> pdischarge[stor_idx, 1, hr_idxs]
2-element Vector{Float64}:
 520.398061636471
 560.6826719850344

julia> pdischarge[stor_idx,:, hr_idxs] .+ pcharge[stor_idx, :, hr_idxs]
1×2 Matrix{Float64}:
 1176.47  1176.47

# Ok, so our new constraint is not being violated here, and we're seeing some slight differences in the amount of charge/discharge in each hour

julia> sum(pdischarge[stor_idx, 1, hr_idxs]) / sum(pcharge[stor_idx, 1, hr_idxs])
0.8500000000003894 # No violation of our efficiency constraint

julia> pcharge[stor_idx,1, hr_idxs] .- pdischarge[stor_idx, 1, hr_idxs]
2-element Vector{Float64}:
 135.674087553042
  55.104865438405454

# A bit more load added to the grid in the first hour than in the 2nd hour.  Let's dig into the LMP.
julia> get_table_col(data, :bus, :lmp_elserv_preloss)[get_table_col(data, :storage, :bus_idx)[stor_idx]][1, hr_idxs]
2-element Vector{Float64}:
 0.08949863770090605
 0.08949863774134087

julia> stor_bus_idx = get_table_col(data, :storage, :bus_idx)[stor_idx]
218

julia> pgen_bus = get_raw_result(data, :pgen_bus); pgen_gen = get_raw_result(data, :pgen_gen);

julia> pgen_bus[stor_bus_idx, 1, hr_idxs]
2-element Vector{Float64}:
 -135.67405792881732
  -55.10483585548184

julia> stor_bus_idx = get_table_col(data, :storage, :bus_idx)[stor_idx]
218

julia> pgen_bus = get_raw_result(data, :pgen_bus); pgen_gen = get_raw_result(data, :pgen_gen);

julia> pgen_bus[stor_bus_idx, 1, hr_idxs]
2-element Vector{Float64}:
 -135.67405792881732
  -55.10483585548184

julia> gen_idxs = get_table_row_idxs(data, :gen, :bus_idx=>stor_bus_idx);

julia> pgen_gen[gen_idxs, 1, hr_idxs]
23×2 Matrix{Float64}:
 2.3553e-8   2.39795e-8
 2.91271e-5  2.90053e-5
 2.37537e-8  2.4189e-8
 0.0         0.0
 3.55872e-8  3.59832e-8
 1.77827e-8  1.80557e-8
 2.41219e-8  2.45704e-8
 3.66558e-8  3.76026e-8
 0.0         0.0
 2.09853e-8  2.13719e-8
 1.77358e-8  1.80055e-8
 2.05869e-8  2.09049e-8
 3.51248e-8  3.5477e-8
 1.98514e-8  1.99209e-8
 2.07689e-8  2.11016e-8
 2.16953e-8  2.20374e-8
 3.59679e-8  3.62593e-8
 4.77814e-8  4.93427e-8
 3.39119e-8  1.02346e-7
 3.81812e-8  3.85238e-8
 0.0         4.70806e-9
 1.15352e-8  1.16289e-8
 1.15328e-8  1.1627e-8

# Huh, no generation at this bus.  But very small LMP.

# Just to keep track of how much discharging is happening while there is charging happening too, I will come up with a metric:
julia> sum(weight_hourly(data, pdischarge) .* (pdischarge .* pcharge .> 1))
1.0157812171984944e7

julia> sum(weight_hourly(data, pdischarge))
7.739219724255703e7

# That's about 10 million TWh out of 77 million TWh of total battery discharge.

[Ethan-Russell]

I think I'd like to add a small VOM to our storage to see if that significantly reduces the problem