how to write a pyomo optimization to select optimal volume of renewables?
Question:
Background
I am trying to write a pyomo optimization which takes in a customer’s electricity load and the generation data of several renewable projects, then optimally solves for the lowest cost selection of renewable projects to minimize electricity consumption, subject to a few constraints.
What I’ve tried
Using the pyomo readthedocs and stackoverflow. I’ve written my first attempt (below), but I am having two issues.
Problem
- ERROR: Rule failed for Expression ‘d_spill_var’ with index 0: PyomoException:
Cannot convert non-constant Pyomo expression
I think this is because I am trying to return a max(expr, 0) value for one of my dependent Expresions. However even if I change this, I still get issue 2 below;
- RuntimeError: Cannot write legal LP file. Objective ‘objective’ has nonlinear terms that are not quadratic.
Help Requested
Could someone please point me in the right direction to solving the above two issues? Any help would be greatly appreciated!
Code OLD v1.0
import os
import pandas as pd
from pyomo.environ import *
import datetime
def model_to_df(model, first_period, last_period):
# Need to increase the first & last hour by 1 because of pyomo indexing
periods = range(model.T[first_period + 1], model.T[last_period + 1] + 1)
spot = [value(model.spot[i]) for i in periods]
load = [value(model.load[i]) for i in periods]
slr1 = [value(model.slr1_size[i]) for i in periods]
slr2 = [value(model.slr2_size[i]) for i in periods]
slr3 = [value(model.slr3_size[i]) for i in periods]
wnd1 = [value(model.wnd1_size[i]) for i in periods]
wnd2 = [value(model.wnd2_size[i]) for i in periods]
wnd3 = [value(model.wnd3_size[i]) for i in periods]
d_slrgen_var = [value(model.d_slrgen_var[i]) for i in periods]
d_wndgen_var = [value(model.d_wndgen_var[i]) for i in periods]
d_spill_var = [value(model.d_spill_var[i]) for i in periods]
d_selfcons_var = [value(model.d_selfcons_var[i]) for i in periods]
df_dict = {
'Period': periods,
'spot': spot,
'load': load,
'slr1': slr1,
'slr2': slr2,
'slr3': slr3,
'wnd1': wnd1,
'wnd2': wnd2,
'wnd3': wnd3,
'd_slrgen_var': d_slrgen_var,
'd_wndgen_var': d_wndgen_var,
'd_spill_var': d_spill_var,
'd_selfcons_var': d_selfcons_var
}
df = pd.DataFrame(df_dict)
return df
LOCATION = r"C:cbc-win64"
os.environ["PATH"] = LOCATION + ";" + os.environ["PATH"]
df = pd.DataFrame({
'hour': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23],
'load': [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
'spot': [65.4, 62.7, 60.9, 60.3, 61.8, 64.5, 65.9, 57.9, 39.7, 28.3, 20.9, 16.3, 18.1, 23.9, 32.3, 43.2, 59.3, 76.3, 80.5, 72.5, 73.1, 69.0, 67.9, 67.7],
'slr1': [0.00, 0.00, 0.00, 0.00, 0.00, 0.04, 0.20, 0.44, 0.60, 0.69, 0.71, 0.99, 1.00, 0.66, 0.75, 0.63, 0.52, 0.34, 0.14, 0.02, 0.00, 0.00, 0.00, 0.00],
'slr2': [0.00, 0.00, 0.00, 0.00, 0.03, 0.19, 0.44, 0.68, 1.00, 0.83, 0.90, 0.88, 0.98, 0.94, 0.83, 0.70, 0.36, 0.11, 0.02, 0.00, 0.00, 0.00, 0.00, 0.00],
'slr3': [0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.03, 0.17, 0.39, 0.87, 0.91, 1.00, 0.89, 0.71, 0.71, 0.85, 0.63, 0.52, 0.32, 0.12, 0.02, 0.00, 0.00, 0.00],
'wnd1': [1.00, 0.72, 0.74, 0.94, 0.69, 0.90, 0.92, 0.76, 0.51, 0.35, 0.31, 0.34, 0.37, 0.28, 0.35, 0.40, 0.39, 0.32, 0.42, 0.48, 0.74, 0.63, 0.80, 0.97],
'wnd2': [0.95, 0.67, 0.82, 0.48, 0.51, 0.41, 0.33, 0.42, 0.34, 0.30, 0.39, 0.29, 0.34, 0.55, 0.67, 0.78, 0.84, 0.73, 0.77, 0.89, 0.76, 0.97, 1.00, 0.91],
'wnd3': [0.32, 0.35, 0.38, 0.28, 0.33, 0.38, 0.41, 0.38, 0.51, 0.65, 0.54, 0.88, 0.93, 0.89, 0.90, 1.00, 0.90, 0.76, 0.76, 0.92, 0.71, 0.56, 0.52, 0.40]
})
first_model_period = df['hour'].iloc[0]
last_model_period = df['hour'].iloc[-1]
# **********************
# Build Model
# **********************
model = ConcreteModel()
# Fixed Paramaters
model.T = Set(initialize=df.index.tolist(), doc='hourly intervals', ordered=True)
model.load_v = Param(model.T, initialize=df.load, doc='customers load', within=Any)
model.spot_v = Param(model.T, initialize=df.spot, doc='spot price for each interval', within=Any)
model.slr1 = Param(model.T, initialize=df.slr1, doc='1MW output solar farm 1', within=Any)
model.slr2 = Param(model.T, initialize=df.slr2, doc='1MW output solar farm 2', within=Any)
model.slr3 = Param(model.T, initialize=df.slr3, doc='1MW output solar farm 3', within=Any)
model.wnd1 = Param(model.T, initialize=df.wnd1, doc='1MW output wind farm 1', within=Any)
model.wnd2 = Param(model.T, initialize=df.wnd2, doc='1MW output wind farm 2', within=Any)
model.wnd3 = Param(model.T, initialize=df.wnd3, doc='1MW output wind farm 3', within=Any)
# Variable Parameters
model.slr1_flag = Var(model.T, doc='slr 1 on / off', within=Binary, initialize=0)
model.slr2_flag = Var(model.T, doc='slr 2 on / off', within=Binary, initialize=0)
model.slr3_flag = Var(model.T, doc='slr 3 on / off', within=Binary, initialize=0)
model.wnd1_flag = Var(model.T, doc='wnd 1 on / off', within=Binary, initialize=0)
model.wnd2_flag = Var(model.T, doc='wnd 2 on / off', within=Binary, initialize=0)
model.wnd3_flag = Var(model.T, doc='wnd 3 on / off', within=Binary, initialize=0)
model.slr1_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr2_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr3_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd1_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd2_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd3_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.total_gen = Var(model.T, initialize=0, within=NonNegativeReals)
# Dependent Expression Parameters
def dependent_solar_gen(model, t):
"Total selected solar Generation"
return (model.slr1[t] * model.slr1_flag[t] * model.slr1_size[t]) +
(model.slr2[t] * model.slr2_flag[t] * model.slr2_size[t]) +
(model.slr3[t] * model.slr3_flag[t] * model.slr3_size[t])
model.d_slrgen_var = Expression(model.T, rule=dependent_solar_gen)
def dependent_wind_gen(model, t):
"Total selected wind Generation"
return (model.wnd1[t] * model.wnd1_flag[t] * model.wnd1_size[t]) +
(model.wnd2[t] * model.wnd2_flag[t] * model.wnd2_size[t]) +
(model.wnd3[t] * model.wnd3_flag[t] * model.wnd3_size[t])
model.d_wndgen_var = Expression(model.T, rule=dependent_wind_gen)
def dependent_spill(model, t):
"Volume of energy not consumed by customer (spilled into grid)"
expr = (model.d_slrgen_var[t] + model.d_wndgen_var[t]) - model.load_v[t]
return max(0, expr)
model.d_spill_var = Expression(model.T, rule=dependent_spill)
def dependent_self_cons(model, t):
"Volume of energy consumed by customer"
expr = (model.d_slrgen_var[t] + model.d_wndgen_var[t]) - model.d_spill_var[t]
return expr
model.d_selfcons_var = Expression(model.T, rule=dependent_self_cons)
# -----------------------
# Constraints
# -----------------------
def min_spill(model, t):
"Limit spill renewables to 10% of total"
return model.d_spill_var[t] <= 0.1 * (model.d_slrgen_var[t] + model.d_wndgen_var[t])
model.min_spill_c = Constraint(model.T, rule=min_spill)
def load_match(model, t):
"contract enough renewables to offset 100% load, even if its not time matched"
return (model.d_slrgen_var[t] + model.d_wndgen_var[t]) >= model.load_v[t]
model.load_match_c = Constraint(model.T, rule=load_match)
# **********************
# Define the income, expenses, and profit
# **********************
green_income = sum(model.spot_v[t] * model.d_spill_var[t] for t in model.T)
black_cost = sum(model.spot_v[t] * (model.load_v[t] - model.d_selfcons_var[t]) for t in model.T)
slr_cost = sum(40 * model.d_slrgen_var[t] for t in model.T)
wnd_cost = sum(70 * model.d_wndgen_var[t] for t in model.T)
profit = green_income - black_cost - slr_cost - wnd_cost
model.objective = Objective(expr=profit, sense=maximize)
# Solve the model
# solver = SolverFactory('glpk')
solver = SolverFactory('cbc')
solver.solve(model, timelimit=10)
results_df = model_to_df(model, first_period=first_model_period, last_period=last_model_period)
print(results_df)
Solved
Thanks @airliquid, I managed to solve the issue thanks to your advice.
What I had to do was linearize the max constraint, redefine dependent expressions as constraints, remove some redundant variables, and change the last two constraints to summations
It’s not the prettiest answer, but it works!
UPDATED CODE (V2)
import os
import pandas as pd
from pyomo.environ import *
import datetime
def model_to_df(model, first_period, last_period):
# Need to increase the first & last hour by 1 because of pyomo indexing
periods = range(model.T[first_period + 1], model.T[last_period + 1] + 1)
spot = [value(model.spot_v[i]) for i in periods]
load = [value(model.load_v[i]) for i in periods]
slr1 = [value(model.slr1_size[i]) for i in periods]
slr2 = [value(model.slr2_size[i]) for i in periods]
slr3 = [value(model.slr3_size[i]) for i in periods]
wnd1 = [value(model.wnd1_size[i]) for i in periods]
wnd2 = [value(model.wnd2_size[i]) for i in periods]
wnd3 = [value(model.wnd3_size[i]) for i in periods]
slr1_gen = [value(model.slr1_gen[i]) for i in periods]
slr2_gen = [value(model.slr2_gen[i]) for i in periods]
slr3_gen = [value(model.slr3_gen[i]) for i in periods]
wnd1_gen = [value(model.wnd1_gen[i]) for i in periods]
wnd2_gen = [value(model.wnd2_gen[i]) for i in periods]
wnd3_gen = [value(model.wnd3_gen[i]) for i in periods]
total_gen = [value(model.total_gen[i]) for i in periods]
spill_gen = [value(model.spill_gen[i]) for i in periods]
spill_gen_sum = [value(model.spill_gen_sum[i]) for i in periods]
spill_binary = [value(model.spill_binary[i]) for i in periods]
self_cons = [value(model.self_cons[i]) for i in periods]
df_dict = {
'Period': periods,
'spot': spot,
'load': load,
'slr1': slr1,
'slr2': slr2,
'slr3': slr3,
'wnd1': wnd1,
'wnd2': wnd2,
'wnd3': wnd3,
'slr1_gen': slr1_gen,
'slr2_gen': slr2_gen,
'slr3_gen': slr3_gen,
'wnd1_gen': wnd1_gen,
'wnd2_gen': wnd2_gen,
'wnd3_gen': wnd3_gen,
'total_gen': total_gen,
'spill_gen': spill_gen,
'self_cons': self_cons,
'spill_gen_sum': spill_gen_sum,
'spill_binary': spill_binary
}
df = pd.DataFrame(df_dict)
return df
LOCATION = r"C:cbc-win64"
os.environ["PATH"] = LOCATION + ";" + os.environ["PATH"]
df = pd.DataFrame({
'hour': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23],
'load': [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
'spot': [65.4, 62.7, 60.9, 60.3, 61.8, 64.5, 65.9, 57.9, 39.7, 28.3, 20.9, 16.3, 18.1, 23.9, 32.3, 43.2, 59.3, 76.3, 80.5, 72.5, 73.1, 69.0, 67.9, 67.7],
'slr1': [0.00, 0.00, 0.00, 0.00, 0.00, 0.04, 0.20, 0.44, 0.60, 0.69, 0.71, 0.99, 1.00, 0.66, 0.75, 0.63, 0.52, 0.34, 0.14, 0.02, 0.00, 0.00, 0.00, 0.00],
'slr2': [0.00, 0.00, 0.00, 0.00, 0.03, 0.19, 0.44, 0.68, 1.00, 0.83, 0.90, 0.88, 0.98, 0.94, 0.83, 0.70, 0.36, 0.11, 0.02, 0.00, 0.00, 0.00, 0.00, 0.00],
'slr3': [0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.03, 0.17, 0.39, 0.87, 0.91, 1.00, 0.89, 0.71, 0.71, 0.85, 0.63, 0.52, 0.32, 0.12, 0.02, 0.00, 0.00, 0.00],
'wnd1': [1.00, 0.72, 0.74, 0.94, 0.69, 0.90, 0.92, 0.76, 0.51, 0.35, 0.31, 0.34, 0.37, 0.28, 0.35, 0.40, 0.39, 0.32, 0.42, 0.48, 0.74, 0.63, 0.80, 0.97],
'wnd2': [0.95, 0.67, 0.82, 0.48, 0.51, 0.41, 0.33, 0.42, 0.34, 0.30, 0.39, 0.29, 0.34, 0.55, 0.67, 0.78, 0.84, 0.73, 0.77, 0.89, 0.76, 0.97, 1.00, 0.91],
'wnd3': [0.32, 0.35, 0.38, 0.28, 0.33, 0.38, 0.41, 0.38, 0.51, 0.65, 0.54, 0.88, 0.93, 0.89, 0.90, 1.00, 0.90, 0.76, 0.76, 0.92, 0.71, 0.56, 0.52, 0.40]
})
# df.to_csv('example.csv', index=False)
first_model_period = df['hour'].iloc[0]
last_model_period = df['hour'].iloc[-1]
# **********************
# Build Model
# **********************
model = ConcreteModel()
# Fixed Paramaters
model.T = Set(initialize=df.index.tolist(), doc='hourly intervals', ordered=True)
model.load_v = Param(model.T, initialize=df.load, doc='customers load', within=Any)
model.spot_v = Param(model.T, initialize=df.spot, doc='spot price for each interval', within=Any)
model.slr1 = Param(model.T, initialize=df.slr1, doc='1MW output solar farm 1', within=Any)
model.slr2 = Param(model.T, initialize=df.slr2, doc='1MW output solar farm 2', within=Any)
model.slr3 = Param(model.T, initialize=df.slr3, doc='1MW output solar farm 3', within=Any)
model.wnd1 = Param(model.T, initialize=df.wnd1, doc='1MW output wind farm 1', within=Any)
model.wnd2 = Param(model.T, initialize=df.wnd2, doc='1MW output wind farm 2', within=Any)
model.wnd3 = Param(model.T, initialize=df.wnd3, doc='1MW output wind farm 3', within=Any)
# Variable Parameters
model.slr1_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr2_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr3_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd1_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd2_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd3_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr1_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.slr2_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.slr3_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd1_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd2_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd3_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.total_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.spill_gen_sum = Var(model.T, initialize=0, within=Reals)
model.spill_binary = Var(model.T, doc='to get max', within=Binary, initialize=0)
model.spill_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.self_cons = Var(model.T, initialize=0, within=NonNegativeReals)
# -----------------------
# Constraints
# -----------------------
# SIZE CONSTRAINTS
def slr1_size(model, t):
"slr1 size"
return model.slr1_size[t] == model.slr1_size[1]
model.slr_size1_c = Constraint(model.T, rule=slr1_size)
def slr2_size(model, t):
"slr2 size"
return model.slr2_size[t] == model.slr2_size[1]
model.slr_size2_c = Constraint(model.T, rule=slr2_size)
def slr3_size(model, t):
"slr3 size"
return model.slr3_size[t] == model.slr3_size[1]
model.slr_size3_c = Constraint(model.T, rule=slr3_size)
def wnd1_size(model, t):
"wnd1 size"
return model.wnd1_size[t] == model.wnd1_size[1]
model.wnd_size1_c = Constraint(model.T, rule=wnd1_size)
def wnd2_size(model, t):
"wnd2 size"
return model.wnd2_size[t] == model.wnd2_size[1]
model.wnd_size2_c = Constraint(model.T, rule=wnd2_size)
def wnd3_size(model, t):
"wnd3 size"
return model.wnd3_size[t] == model.wnd3_size[1]
model.wnd_size3_c = Constraint(model.T, rule=wnd3_size)
# GENERATION EXPRESSIONS / CONSTRAINTS
def slr1_gen(model, t):
"solar 1 generation"
return model.slr1_gen[t] == model.slr1[t] * model.slr1_size[t]
model.slr_gen1_c = Constraint(model.T, rule=slr1_gen)
def slr2_gen(model, t):
"solar 2 generation"
return model.slr2_gen[t] == model.slr2[t] * model.slr2_size[t]
model.slr_gen2_c = Constraint(model.T, rule=slr2_gen)
def slr3_gen(model, t):
"solar 3 generation"
return model.slr3_gen[t] == model.slr3[t] * model.slr3_size[t]
model.slr_gen3_c = Constraint(model.T, rule=slr3_gen)
def wnd1_gen(model, t):
"wind 1 generation"
return model.wnd1_gen[t] == model.wnd1[t] * model.wnd1_size[t]
model.wnd_gen1_c = Constraint(model.T, rule=wnd1_gen)
def wnd2_gen(model, t):
"wind 2 generation"
return model.wnd2_gen[t] == model.wnd2[t] * model.wnd2_size[t]
model.wnd_gen2_c = Constraint(model.T, rule=wnd2_gen)
def wnd3_gen(model, t):
"wind 3 generation"
return model.wnd3_gen[t] == model.wnd3[t] * model.wnd3_size[t]
model.wnd_gen3_c = Constraint(model.T, rule=wnd3_gen)
# TOTAL GENERATION
def total_gen(model, t):
"sum of generation"
return model.total_gen[t] == model.slr1_gen[t] + model.slr2_gen[t] + model.slr3_gen[t] +
model.wnd1_gen[t] + model.wnd2_gen[t] + model.wnd3_gen[t]
model.total_gen_c = Constraint(model.T, rule=total_gen)
# SPILL GENERATION
def spill_gen_sum(model, t):
"X >= x1"
return model.spill_gen_sum[t] == model.total_gen[t] - model.load_v[t]
model.spill_gen_sum_c = Constraint(model.T, rule=spill_gen_sum)
def spill_check_one(model, t):
"X >= x1"
return model.spill_gen[t] >= model.spill_gen_sum[t]
model.spill_check_one_c = Constraint(model.T, rule=spill_check_one)
def spill_check_two(model, t):
"X >= x2"
return model.spill_gen[t] >= 0
model.spill_check_two_c = Constraint(model.T, rule=spill_check_two)
def spill_binary_one(model, t):
"X <= x1 + M(1-y)"
return model.spill_gen[t] <= model.spill_gen_sum[t] + 9999999*(1-model.spill_binary[t])
model.spill_binary_one_c = Constraint(model.T, rule=spill_binary_one)
def spill_binary_two(model, t):
"X <= x2 + My"
return model.spill_gen[t] <= 9999999*model.spill_binary[t]
model.spill_binary_two_c = Constraint(model.T, rule=spill_binary_two)
# SELF CONS
def self_cons(model, t):
"X <= x2 + My"
return model.self_cons[t] == model.total_gen[t] - model.spill_gen[t]
model.self_cons_c = Constraint(model.T, rule=self_cons)
# ACTUAL CONSTRAINTS
def min_spill(model, t):
"Limit spill renewables to 10% of total"
return sum(model.spill_gen[t] for t in model.T) <= 0.2 * sum(model.total_gen[t] for t in model.T)
model.min_spill_c = Constraint(model.T, rule=min_spill)
def load_match(model, t):
"contract enough renewables to offset 100% load, even if its not time matched"
return sum(model.total_gen[t] for t in model.T) >= sum(model.load_v[t] for t in model.T)
model.load_match_c = Constraint(model.T, rule=load_match)
# **********************
# Define the battery income, expenses, and profit
# **********************
green_income = sum(model.spot_v[t] * model.spill_gen[t] for t in model.T)
black_cost = sum(model.spot_v[t] * (model.load_v[t] - model.self_cons[t]) for t in model.T)
slr_cost = sum(40 * (model.slr1_gen[t] + model.slr2_gen[t] + model.slr3_gen[t]) for t in model.T)
wnd_cost = sum(70 * (model.wnd1_gen[t] + model.wnd2_gen[t] + model.wnd3_gen[t]) for t in model.T)
cost = black_cost + slr_cost + wnd_cost - green_income
model.objective = Objective(expr=cost, sense=minimize)
# Solve the model
# solver = SolverFactory('glpk')
solver = SolverFactory('cbc')
solver.solve(model)
# , timelimit=10
results_df = model_to_df(model, first_period=first_model_period, last_period=last_model_period)
print(results_df)
results_df.to_csv('temp.csv', index=False)
Answers:
Your model is well written (clear and organized), but shows some significant Linear Programming issues. A couple pointers to help you along the way….
- You are making this model non-linear by multiplying variables together, which adds HUGE complication for the solver and is not necessary in this case. You can use a couple of constraints, including a "big-M" constraint to remove the portion where you are multiplying your flag times capacity (both variables…hence the nonlinearity). You are essentially doing this:
output[t] = energy[t] * size[t] * running[t]
where energy is a parameter, the others are variables, and output is rolled up into an expression. You can linearize that by just making output a variable and using 3 constraints:
output[t] >= energy[t] * size[t] - (1 - running[t]) * M
output[t] <= energy[t] * size[t]
output[t] <= running[t] * M
Where M is some logical upper bound on energy*size. Sometimes the upper + the lower are not needed, but you probably need this to enforce the "equality" constraint you want because of the overage constraint.
I would make output a new variable to alleviate the complex expressions you have.
- In your attempt to make the "rolled up expressions" you are instead making a set of "indexed expressions". Just use a summation statement over model.T. This:
model.d_slrgen_var = Expression(model.T, rule=dependent_solar_gen)
makes a set of expressions, not a summation
max() is a nonlinear operator, and you cannot return or use that in a LP expression. So…. Introduce another variable similar to:
spill[t] ∈ {non-negative REALS }
spill[t] >= ... for each t in model.T
After you manage to get all that ironed out ( 😉 ) and working, you might want to re-organize the model and double-index your main variables by [source, time] instead of naming them all by source, which is a bit cleaner, but window dressing for later.
Comment back if you are stuck…
Background
I am trying to write a pyomo optimization which takes in a customer’s electricity load and the generation data of several renewable projects, then optimally solves for the lowest cost selection of renewable projects to minimize electricity consumption, subject to a few constraints.
What I’ve tried
Using the pyomo readthedocs and stackoverflow. I’ve written my first attempt (below), but I am having two issues.
Problem
- ERROR: Rule failed for Expression ‘d_spill_var’ with index 0: PyomoException:
Cannot convert non-constant Pyomo expression
I think this is because I am trying to return a max(expr, 0) value for one of my dependent Expresions. However even if I change this, I still get issue 2 below;
- RuntimeError: Cannot write legal LP file. Objective ‘objective’ has nonlinear terms that are not quadratic.
Help Requested
Could someone please point me in the right direction to solving the above two issues? Any help would be greatly appreciated!
Code OLD v1.0
import os
import pandas as pd
from pyomo.environ import *
import datetime
def model_to_df(model, first_period, last_period):
# Need to increase the first & last hour by 1 because of pyomo indexing
periods = range(model.T[first_period + 1], model.T[last_period + 1] + 1)
spot = [value(model.spot[i]) for i in periods]
load = [value(model.load[i]) for i in periods]
slr1 = [value(model.slr1_size[i]) for i in periods]
slr2 = [value(model.slr2_size[i]) for i in periods]
slr3 = [value(model.slr3_size[i]) for i in periods]
wnd1 = [value(model.wnd1_size[i]) for i in periods]
wnd2 = [value(model.wnd2_size[i]) for i in periods]
wnd3 = [value(model.wnd3_size[i]) for i in periods]
d_slrgen_var = [value(model.d_slrgen_var[i]) for i in periods]
d_wndgen_var = [value(model.d_wndgen_var[i]) for i in periods]
d_spill_var = [value(model.d_spill_var[i]) for i in periods]
d_selfcons_var = [value(model.d_selfcons_var[i]) for i in periods]
df_dict = {
'Period': periods,
'spot': spot,
'load': load,
'slr1': slr1,
'slr2': slr2,
'slr3': slr3,
'wnd1': wnd1,
'wnd2': wnd2,
'wnd3': wnd3,
'd_slrgen_var': d_slrgen_var,
'd_wndgen_var': d_wndgen_var,
'd_spill_var': d_spill_var,
'd_selfcons_var': d_selfcons_var
}
df = pd.DataFrame(df_dict)
return df
LOCATION = r"C:cbc-win64"
os.environ["PATH"] = LOCATION + ";" + os.environ["PATH"]
df = pd.DataFrame({
'hour': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23],
'load': [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
'spot': [65.4, 62.7, 60.9, 60.3, 61.8, 64.5, 65.9, 57.9, 39.7, 28.3, 20.9, 16.3, 18.1, 23.9, 32.3, 43.2, 59.3, 76.3, 80.5, 72.5, 73.1, 69.0, 67.9, 67.7],
'slr1': [0.00, 0.00, 0.00, 0.00, 0.00, 0.04, 0.20, 0.44, 0.60, 0.69, 0.71, 0.99, 1.00, 0.66, 0.75, 0.63, 0.52, 0.34, 0.14, 0.02, 0.00, 0.00, 0.00, 0.00],
'slr2': [0.00, 0.00, 0.00, 0.00, 0.03, 0.19, 0.44, 0.68, 1.00, 0.83, 0.90, 0.88, 0.98, 0.94, 0.83, 0.70, 0.36, 0.11, 0.02, 0.00, 0.00, 0.00, 0.00, 0.00],
'slr3': [0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.03, 0.17, 0.39, 0.87, 0.91, 1.00, 0.89, 0.71, 0.71, 0.85, 0.63, 0.52, 0.32, 0.12, 0.02, 0.00, 0.00, 0.00],
'wnd1': [1.00, 0.72, 0.74, 0.94, 0.69, 0.90, 0.92, 0.76, 0.51, 0.35, 0.31, 0.34, 0.37, 0.28, 0.35, 0.40, 0.39, 0.32, 0.42, 0.48, 0.74, 0.63, 0.80, 0.97],
'wnd2': [0.95, 0.67, 0.82, 0.48, 0.51, 0.41, 0.33, 0.42, 0.34, 0.30, 0.39, 0.29, 0.34, 0.55, 0.67, 0.78, 0.84, 0.73, 0.77, 0.89, 0.76, 0.97, 1.00, 0.91],
'wnd3': [0.32, 0.35, 0.38, 0.28, 0.33, 0.38, 0.41, 0.38, 0.51, 0.65, 0.54, 0.88, 0.93, 0.89, 0.90, 1.00, 0.90, 0.76, 0.76, 0.92, 0.71, 0.56, 0.52, 0.40]
})
first_model_period = df['hour'].iloc[0]
last_model_period = df['hour'].iloc[-1]
# **********************
# Build Model
# **********************
model = ConcreteModel()
# Fixed Paramaters
model.T = Set(initialize=df.index.tolist(), doc='hourly intervals', ordered=True)
model.load_v = Param(model.T, initialize=df.load, doc='customers load', within=Any)
model.spot_v = Param(model.T, initialize=df.spot, doc='spot price for each interval', within=Any)
model.slr1 = Param(model.T, initialize=df.slr1, doc='1MW output solar farm 1', within=Any)
model.slr2 = Param(model.T, initialize=df.slr2, doc='1MW output solar farm 2', within=Any)
model.slr3 = Param(model.T, initialize=df.slr3, doc='1MW output solar farm 3', within=Any)
model.wnd1 = Param(model.T, initialize=df.wnd1, doc='1MW output wind farm 1', within=Any)
model.wnd2 = Param(model.T, initialize=df.wnd2, doc='1MW output wind farm 2', within=Any)
model.wnd3 = Param(model.T, initialize=df.wnd3, doc='1MW output wind farm 3', within=Any)
# Variable Parameters
model.slr1_flag = Var(model.T, doc='slr 1 on / off', within=Binary, initialize=0)
model.slr2_flag = Var(model.T, doc='slr 2 on / off', within=Binary, initialize=0)
model.slr3_flag = Var(model.T, doc='slr 3 on / off', within=Binary, initialize=0)
model.wnd1_flag = Var(model.T, doc='wnd 1 on / off', within=Binary, initialize=0)
model.wnd2_flag = Var(model.T, doc='wnd 2 on / off', within=Binary, initialize=0)
model.wnd3_flag = Var(model.T, doc='wnd 3 on / off', within=Binary, initialize=0)
model.slr1_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr2_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr3_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd1_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd2_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd3_size = Var(model.T, bounds=(0, 1500), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.total_gen = Var(model.T, initialize=0, within=NonNegativeReals)
# Dependent Expression Parameters
def dependent_solar_gen(model, t):
"Total selected solar Generation"
return (model.slr1[t] * model.slr1_flag[t] * model.slr1_size[t]) +
(model.slr2[t] * model.slr2_flag[t] * model.slr2_size[t]) +
(model.slr3[t] * model.slr3_flag[t] * model.slr3_size[t])
model.d_slrgen_var = Expression(model.T, rule=dependent_solar_gen)
def dependent_wind_gen(model, t):
"Total selected wind Generation"
return (model.wnd1[t] * model.wnd1_flag[t] * model.wnd1_size[t]) +
(model.wnd2[t] * model.wnd2_flag[t] * model.wnd2_size[t]) +
(model.wnd3[t] * model.wnd3_flag[t] * model.wnd3_size[t])
model.d_wndgen_var = Expression(model.T, rule=dependent_wind_gen)
def dependent_spill(model, t):
"Volume of energy not consumed by customer (spilled into grid)"
expr = (model.d_slrgen_var[t] + model.d_wndgen_var[t]) - model.load_v[t]
return max(0, expr)
model.d_spill_var = Expression(model.T, rule=dependent_spill)
def dependent_self_cons(model, t):
"Volume of energy consumed by customer"
expr = (model.d_slrgen_var[t] + model.d_wndgen_var[t]) - model.d_spill_var[t]
return expr
model.d_selfcons_var = Expression(model.T, rule=dependent_self_cons)
# -----------------------
# Constraints
# -----------------------
def min_spill(model, t):
"Limit spill renewables to 10% of total"
return model.d_spill_var[t] <= 0.1 * (model.d_slrgen_var[t] + model.d_wndgen_var[t])
model.min_spill_c = Constraint(model.T, rule=min_spill)
def load_match(model, t):
"contract enough renewables to offset 100% load, even if its not time matched"
return (model.d_slrgen_var[t] + model.d_wndgen_var[t]) >= model.load_v[t]
model.load_match_c = Constraint(model.T, rule=load_match)
# **********************
# Define the income, expenses, and profit
# **********************
green_income = sum(model.spot_v[t] * model.d_spill_var[t] for t in model.T)
black_cost = sum(model.spot_v[t] * (model.load_v[t] - model.d_selfcons_var[t]) for t in model.T)
slr_cost = sum(40 * model.d_slrgen_var[t] for t in model.T)
wnd_cost = sum(70 * model.d_wndgen_var[t] for t in model.T)
profit = green_income - black_cost - slr_cost - wnd_cost
model.objective = Objective(expr=profit, sense=maximize)
# Solve the model
# solver = SolverFactory('glpk')
solver = SolverFactory('cbc')
solver.solve(model, timelimit=10)
results_df = model_to_df(model, first_period=first_model_period, last_period=last_model_period)
print(results_df)
Solved
Thanks @airliquid, I managed to solve the issue thanks to your advice.
What I had to do was linearize the max constraint, redefine dependent expressions as constraints, remove some redundant variables, and change the last two constraints to summations
It’s not the prettiest answer, but it works!
UPDATED CODE (V2)
import os
import pandas as pd
from pyomo.environ import *
import datetime
def model_to_df(model, first_period, last_period):
# Need to increase the first & last hour by 1 because of pyomo indexing
periods = range(model.T[first_period + 1], model.T[last_period + 1] + 1)
spot = [value(model.spot_v[i]) for i in periods]
load = [value(model.load_v[i]) for i in periods]
slr1 = [value(model.slr1_size[i]) for i in periods]
slr2 = [value(model.slr2_size[i]) for i in periods]
slr3 = [value(model.slr3_size[i]) for i in periods]
wnd1 = [value(model.wnd1_size[i]) for i in periods]
wnd2 = [value(model.wnd2_size[i]) for i in periods]
wnd3 = [value(model.wnd3_size[i]) for i in periods]
slr1_gen = [value(model.slr1_gen[i]) for i in periods]
slr2_gen = [value(model.slr2_gen[i]) for i in periods]
slr3_gen = [value(model.slr3_gen[i]) for i in periods]
wnd1_gen = [value(model.wnd1_gen[i]) for i in periods]
wnd2_gen = [value(model.wnd2_gen[i]) for i in periods]
wnd3_gen = [value(model.wnd3_gen[i]) for i in periods]
total_gen = [value(model.total_gen[i]) for i in periods]
spill_gen = [value(model.spill_gen[i]) for i in periods]
spill_gen_sum = [value(model.spill_gen_sum[i]) for i in periods]
spill_binary = [value(model.spill_binary[i]) for i in periods]
self_cons = [value(model.self_cons[i]) for i in periods]
df_dict = {
'Period': periods,
'spot': spot,
'load': load,
'slr1': slr1,
'slr2': slr2,
'slr3': slr3,
'wnd1': wnd1,
'wnd2': wnd2,
'wnd3': wnd3,
'slr1_gen': slr1_gen,
'slr2_gen': slr2_gen,
'slr3_gen': slr3_gen,
'wnd1_gen': wnd1_gen,
'wnd2_gen': wnd2_gen,
'wnd3_gen': wnd3_gen,
'total_gen': total_gen,
'spill_gen': spill_gen,
'self_cons': self_cons,
'spill_gen_sum': spill_gen_sum,
'spill_binary': spill_binary
}
df = pd.DataFrame(df_dict)
return df
LOCATION = r"C:cbc-win64"
os.environ["PATH"] = LOCATION + ";" + os.environ["PATH"]
df = pd.DataFrame({
'hour': [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23],
'load': [100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100, 100],
'spot': [65.4, 62.7, 60.9, 60.3, 61.8, 64.5, 65.9, 57.9, 39.7, 28.3, 20.9, 16.3, 18.1, 23.9, 32.3, 43.2, 59.3, 76.3, 80.5, 72.5, 73.1, 69.0, 67.9, 67.7],
'slr1': [0.00, 0.00, 0.00, 0.00, 0.00, 0.04, 0.20, 0.44, 0.60, 0.69, 0.71, 0.99, 1.00, 0.66, 0.75, 0.63, 0.52, 0.34, 0.14, 0.02, 0.00, 0.00, 0.00, 0.00],
'slr2': [0.00, 0.00, 0.00, 0.00, 0.03, 0.19, 0.44, 0.68, 1.00, 0.83, 0.90, 0.88, 0.98, 0.94, 0.83, 0.70, 0.36, 0.11, 0.02, 0.00, 0.00, 0.00, 0.00, 0.00],
'slr3': [0.00, 0.00, 0.00, 0.00, 0.00, 0.00, 0.03, 0.17, 0.39, 0.87, 0.91, 1.00, 0.89, 0.71, 0.71, 0.85, 0.63, 0.52, 0.32, 0.12, 0.02, 0.00, 0.00, 0.00],
'wnd1': [1.00, 0.72, 0.74, 0.94, 0.69, 0.90, 0.92, 0.76, 0.51, 0.35, 0.31, 0.34, 0.37, 0.28, 0.35, 0.40, 0.39, 0.32, 0.42, 0.48, 0.74, 0.63, 0.80, 0.97],
'wnd2': [0.95, 0.67, 0.82, 0.48, 0.51, 0.41, 0.33, 0.42, 0.34, 0.30, 0.39, 0.29, 0.34, 0.55, 0.67, 0.78, 0.84, 0.73, 0.77, 0.89, 0.76, 0.97, 1.00, 0.91],
'wnd3': [0.32, 0.35, 0.38, 0.28, 0.33, 0.38, 0.41, 0.38, 0.51, 0.65, 0.54, 0.88, 0.93, 0.89, 0.90, 1.00, 0.90, 0.76, 0.76, 0.92, 0.71, 0.56, 0.52, 0.40]
})
# df.to_csv('example.csv', index=False)
first_model_period = df['hour'].iloc[0]
last_model_period = df['hour'].iloc[-1]
# **********************
# Build Model
# **********************
model = ConcreteModel()
# Fixed Paramaters
model.T = Set(initialize=df.index.tolist(), doc='hourly intervals', ordered=True)
model.load_v = Param(model.T, initialize=df.load, doc='customers load', within=Any)
model.spot_v = Param(model.T, initialize=df.spot, doc='spot price for each interval', within=Any)
model.slr1 = Param(model.T, initialize=df.slr1, doc='1MW output solar farm 1', within=Any)
model.slr2 = Param(model.T, initialize=df.slr2, doc='1MW output solar farm 2', within=Any)
model.slr3 = Param(model.T, initialize=df.slr3, doc='1MW output solar farm 3', within=Any)
model.wnd1 = Param(model.T, initialize=df.wnd1, doc='1MW output wind farm 1', within=Any)
model.wnd2 = Param(model.T, initialize=df.wnd2, doc='1MW output wind farm 2', within=Any)
model.wnd3 = Param(model.T, initialize=df.wnd3, doc='1MW output wind farm 3', within=Any)
# Variable Parameters
model.slr1_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr2_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr3_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd1_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd2_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.wnd3_size = Var(model.T, bounds=(0, 1000), doc='selected size in MWs', initialize=0, within=NonNegativeIntegers)
model.slr1_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.slr2_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.slr3_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd1_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd2_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.wnd3_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.total_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.spill_gen_sum = Var(model.T, initialize=0, within=Reals)
model.spill_binary = Var(model.T, doc='to get max', within=Binary, initialize=0)
model.spill_gen = Var(model.T, initialize=0, within=NonNegativeReals)
model.self_cons = Var(model.T, initialize=0, within=NonNegativeReals)
# -----------------------
# Constraints
# -----------------------
# SIZE CONSTRAINTS
def slr1_size(model, t):
"slr1 size"
return model.slr1_size[t] == model.slr1_size[1]
model.slr_size1_c = Constraint(model.T, rule=slr1_size)
def slr2_size(model, t):
"slr2 size"
return model.slr2_size[t] == model.slr2_size[1]
model.slr_size2_c = Constraint(model.T, rule=slr2_size)
def slr3_size(model, t):
"slr3 size"
return model.slr3_size[t] == model.slr3_size[1]
model.slr_size3_c = Constraint(model.T, rule=slr3_size)
def wnd1_size(model, t):
"wnd1 size"
return model.wnd1_size[t] == model.wnd1_size[1]
model.wnd_size1_c = Constraint(model.T, rule=wnd1_size)
def wnd2_size(model, t):
"wnd2 size"
return model.wnd2_size[t] == model.wnd2_size[1]
model.wnd_size2_c = Constraint(model.T, rule=wnd2_size)
def wnd3_size(model, t):
"wnd3 size"
return model.wnd3_size[t] == model.wnd3_size[1]
model.wnd_size3_c = Constraint(model.T, rule=wnd3_size)
# GENERATION EXPRESSIONS / CONSTRAINTS
def slr1_gen(model, t):
"solar 1 generation"
return model.slr1_gen[t] == model.slr1[t] * model.slr1_size[t]
model.slr_gen1_c = Constraint(model.T, rule=slr1_gen)
def slr2_gen(model, t):
"solar 2 generation"
return model.slr2_gen[t] == model.slr2[t] * model.slr2_size[t]
model.slr_gen2_c = Constraint(model.T, rule=slr2_gen)
def slr3_gen(model, t):
"solar 3 generation"
return model.slr3_gen[t] == model.slr3[t] * model.slr3_size[t]
model.slr_gen3_c = Constraint(model.T, rule=slr3_gen)
def wnd1_gen(model, t):
"wind 1 generation"
return model.wnd1_gen[t] == model.wnd1[t] * model.wnd1_size[t]
model.wnd_gen1_c = Constraint(model.T, rule=wnd1_gen)
def wnd2_gen(model, t):
"wind 2 generation"
return model.wnd2_gen[t] == model.wnd2[t] * model.wnd2_size[t]
model.wnd_gen2_c = Constraint(model.T, rule=wnd2_gen)
def wnd3_gen(model, t):
"wind 3 generation"
return model.wnd3_gen[t] == model.wnd3[t] * model.wnd3_size[t]
model.wnd_gen3_c = Constraint(model.T, rule=wnd3_gen)
# TOTAL GENERATION
def total_gen(model, t):
"sum of generation"
return model.total_gen[t] == model.slr1_gen[t] + model.slr2_gen[t] + model.slr3_gen[t] +
model.wnd1_gen[t] + model.wnd2_gen[t] + model.wnd3_gen[t]
model.total_gen_c = Constraint(model.T, rule=total_gen)
# SPILL GENERATION
def spill_gen_sum(model, t):
"X >= x1"
return model.spill_gen_sum[t] == model.total_gen[t] - model.load_v[t]
model.spill_gen_sum_c = Constraint(model.T, rule=spill_gen_sum)
def spill_check_one(model, t):
"X >= x1"
return model.spill_gen[t] >= model.spill_gen_sum[t]
model.spill_check_one_c = Constraint(model.T, rule=spill_check_one)
def spill_check_two(model, t):
"X >= x2"
return model.spill_gen[t] >= 0
model.spill_check_two_c = Constraint(model.T, rule=spill_check_two)
def spill_binary_one(model, t):
"X <= x1 + M(1-y)"
return model.spill_gen[t] <= model.spill_gen_sum[t] + 9999999*(1-model.spill_binary[t])
model.spill_binary_one_c = Constraint(model.T, rule=spill_binary_one)
def spill_binary_two(model, t):
"X <= x2 + My"
return model.spill_gen[t] <= 9999999*model.spill_binary[t]
model.spill_binary_two_c = Constraint(model.T, rule=spill_binary_two)
# SELF CONS
def self_cons(model, t):
"X <= x2 + My"
return model.self_cons[t] == model.total_gen[t] - model.spill_gen[t]
model.self_cons_c = Constraint(model.T, rule=self_cons)
# ACTUAL CONSTRAINTS
def min_spill(model, t):
"Limit spill renewables to 10% of total"
return sum(model.spill_gen[t] for t in model.T) <= 0.2 * sum(model.total_gen[t] for t in model.T)
model.min_spill_c = Constraint(model.T, rule=min_spill)
def load_match(model, t):
"contract enough renewables to offset 100% load, even if its not time matched"
return sum(model.total_gen[t] for t in model.T) >= sum(model.load_v[t] for t in model.T)
model.load_match_c = Constraint(model.T, rule=load_match)
# **********************
# Define the battery income, expenses, and profit
# **********************
green_income = sum(model.spot_v[t] * model.spill_gen[t] for t in model.T)
black_cost = sum(model.spot_v[t] * (model.load_v[t] - model.self_cons[t]) for t in model.T)
slr_cost = sum(40 * (model.slr1_gen[t] + model.slr2_gen[t] + model.slr3_gen[t]) for t in model.T)
wnd_cost = sum(70 * (model.wnd1_gen[t] + model.wnd2_gen[t] + model.wnd3_gen[t]) for t in model.T)
cost = black_cost + slr_cost + wnd_cost - green_income
model.objective = Objective(expr=cost, sense=minimize)
# Solve the model
# solver = SolverFactory('glpk')
solver = SolverFactory('cbc')
solver.solve(model)
# , timelimit=10
results_df = model_to_df(model, first_period=first_model_period, last_period=last_model_period)
print(results_df)
results_df.to_csv('temp.csv', index=False)
Your model is well written (clear and organized), but shows some significant Linear Programming issues. A couple pointers to help you along the way….
- You are making this model non-linear by multiplying variables together, which adds HUGE complication for the solver and is not necessary in this case. You can use a couple of constraints, including a "big-M" constraint to remove the portion where you are multiplying your flag times capacity (both variables…hence the nonlinearity). You are essentially doing this:
output[t] = energy[t] * size[t] * running[t]
where energy is a parameter, the others are variables, and output is rolled up into an expression. You can linearize that by just making output a variable and using 3 constraints:
output[t] >= energy[t] * size[t] - (1 - running[t]) * M
output[t] <= energy[t] * size[t]
output[t] <= running[t] * M
Where M is some logical upper bound on energy*size. Sometimes the upper + the lower are not needed, but you probably need this to enforce the "equality" constraint you want because of the overage constraint.
I would make output a new variable to alleviate the complex expressions you have.
- In your attempt to make the "rolled up expressions" you are instead making a set of "indexed expressions". Just use a summation statement over model.T. This:
model.d_slrgen_var = Expression(model.T, rule=dependent_solar_gen)
makes a set of expressions, not a summation
max()is a nonlinear operator, and you cannot return or use that in a LP expression. So…. Introduce another variable similar to:
spill[t] ∈ {non-negative REALS }
spill[t] >= ... for each t in model.T
After you manage to get all that ironed out ( 😉 ) and working, you might want to re-organize the model and double-index your main variables by [source, time] instead of naming them all by source, which is a bit cleaner, but window dressing for later.
Comment back if you are stuck…