Heat Pump Flexibility#

This tutorial covers heat pump modeling and flexibility mechanisms in PyPSA-GB, including demand disaggregation, thermal storage, and demand shifting.

What You’ll Learn#

  • How heat pump electricity demand is calculated from FES data

  • The two space heating flexibility mechanisms: TANK and COSY

  • How COP (Coefficient of Performance) varies with temperature

  • How to analyze heat pump flexibility results

  • PyPSA components created for heat flexibility

1. Setup#

[1]:

import pypsa
import pandas as pd
import numpy as np
import matplotlib.pyplot as plt
import warnings

warnings.filterwarnings('ignore')
plt.style.use('seaborn-v0_8-whitegrid')
plt.rcParams['figure.figsize'] = [14, 6]
plt.rcParams['figure.dpi'] = 100

print(f"PyPSA version: {pypsa.__version__}")

PyPSA version: 1.0.7

2. Understanding Heat Pump Flexibility#

2.1 Why Heat Pump Flexibility Matters#

Heat pumps are a key technology for decarbonizing heating in Great Britain. They convert electricity into heat at high efficiency (COP 2-5 depending on temperature), but they add significant new electricity demand. By 2035, FES projects that heat pumps could add 20-40 TWh of electricity demand.

Flexibility allows this demand to be shifted in time, helping to:

  • Reduce peak demand on the electricity grid

  • Better utilize renewable generation when it’s available

  • Reduce consumer energy costs by heating at off-peak times

2.2 Two Space Heating Flexibility Mechanisms#

PyPSA-GB models two complementary flexibility mechanisms. Both are for space heating, but use different physical storage:

Mechanism

Physical Storage

PyPSA Load Carrier

How It Works

Typical Capacity

TANK

Hot water cylinder

hot water demand

Pre-heats water in tank, releases heat when needed

10-50 kWh thermal

COSY

Building thermal mass

space heating

Pre-heats building fabric (walls, floors), slowly releases heat

2-10 kWh/100m²

TANK Mode (Hot Water Tank Storage)#

                    COP
Electricity ──▶ [Heat Pump] ──▶ [Hot Water Tank] ──▶ Space Heating
    (MW)         (Link)           (Store)              (Load)

  • Heat pump converts electricity to heat at COP efficiency

  • Heat stored in insulated hot water tank (can be charged/discharged)

  • Tank releases heat to space heating system when needed

  • Heat loss: ~1-3% per hour depending on insulation

COSY Mode (Building Thermal Inertia)#

                    COP
Electricity ──▶ [Heat Pump] ──▶ [Building Mass] ──▶ Space Heating
    (MW)         (Link)           (Store)             (Load)

  • Heat pump converts electricity to heat at COP efficiency

  • Heat absorbed by building fabric (concrete, brick, furniture)

  • Building slowly releases stored heat, maintaining comfort

  • Heat loss depends on building insulation (EPC rating)

3. Load a Network with Heat Pump Flexibility#

[2]:

# Load a solved network with heat pump flexibility enabled
# HT35_flex is a future scenario (2035, Holistic Transition) with heat flexibility
n = pypsa.Network("../../../resources/network/HT35_flex_solved.nc")

print("Network loaded")
print(f"  Buses: {len(n.buses)}")
print(f"  Loads: {len(n.loads)}")
print(f"  Links: {len(n.links)}")
print(f"  Stores: {len(n.stores)}")
print(f"  Snapshots: {len(n.snapshots)}")
print(f"  Period: {n.snapshots[0]} to {n.snapshots[-1]}")

INFO:pypsa.network.io:Imported network 'HT35_flex (Clustered)' has buses, carriers, generators, lines, links, loads, storage_units, stores, sub_networks

Network loaded
  Buses: 924
  Loads: 877
  Links: 1221
  Stores: 627
  Snapshots: 168
  Period: 2035-01-13 00:00:00 to 2035-01-19 23:00:00

4. Heat Pump Components in the Network#

4.1 Identifying Heat Pump Loads#

[3]:

# Heat pump loads are identified by their carrier, not by naming convention
# TANK mode (hot water demand)
tank_loads = n.loads[n.loads['carrier'] == 'hot water demand']
print(f"TANK mode loads (hot water demand): {len(tank_loads)}")

# COSY mode (space heating)
cosy_loads = n.loads[n.loads['carrier'] == 'space heating']
print(f"COSY mode loads (space heating): {len(cosy_loads)}")

# Base electricity loads (non-flexible)
base_loads = n.loads[n.loads['carrier'] == '']
print(f"Base electricity loads (non-flexible): {len(base_loads)}")

# Total heat pump loads
total_hp_loads = len(tank_loads) + len(cosy_loads)
print(f"\nTotal heat pump loads: {total_hp_loads}")

# Show sample loads from each type
if len(tank_loads) > 0:
    print("\nSample TANK loads (hot water demand):")
    print(tank_loads[['bus', 'carrier']].head())

if len(cosy_loads) > 0:
    print("\nSample COSY loads (space heating):")
    print(cosy_loads[['bus', 'carrier']].head())

TANK mode loads (hot water demand): 313
COSY mode loads (space heating): 313
Base electricity loads (non-flexible): 0

Total heat pump loads: 626

Sample TANK loads (hot water demand):
                                           bus           carrier
name
ABHA11 hot water demand tank  heat_ABHA11 heat  hot water demand
ABNE3- hot water demand tank  heat_ABNE3- heat  hot water demand
ABTH11 hot water demand tank  heat_ABTH11 heat  hot water demand
ALNE3J hot water demand tank  heat_ALNE3J heat  hot water demand
ALST31 hot water demand tank  heat_ALST31 heat  hot water demand

Sample COSY loads (space heating):
                                                         bus        carrier
name
ABHA11 space heating  thermal inertia_ABHA11 thermal inertia  space heating
ABNE3- space heating  thermal inertia_ABNE3- thermal inertia  space heating
ABTH11 space heating  thermal inertia_ABTH11 thermal inertia  space heating
ALNE3J space heating  thermal inertia_ALNE3J thermal inertia  space heating
ALST31 space heating  thermal inertia_ALST31 thermal inertia  space heating

4.3 Thermal Storage (Stores)#

[6]:

# Identify thermal buses and stores
thermal_buses = n.buses[n.buses['carrier'] == 'heat']
print(f"Thermal buses (for heat): {len(thermal_buses)}")

# TANK stores (hot water tanks)
tank_stores = n.stores[n.stores['carrier'] == 'hot water']
print(f"TANK stores (hot water tanks): {len(tank_stores)}")

# COSY stores (building thermal mass)
cosy_stores = n.stores[n.stores['carrier'] == 'thermal inertia']
print(f"COSY stores (building thermal mass): {len(cosy_stores)}")

if len(tank_stores) > 0:
    print(f"\nTotal TANK storage capacity: {tank_stores['e_nom'].sum():.1f} MWh")
    print(f"  Mean per location: {tank_stores['e_nom'].mean():.2f} MWh")
    print(f"  Sample TANK stores:")
    print(tank_stores[['bus', 'carrier', 'e_nom']].head())

if len(cosy_stores) > 0:
    print(f"\nTotal COSY storage capacity: {cosy_stores['e_nom'].sum():.1f} MWh")
    print(f"  Mean per location: {cosy_stores['e_nom'].mean():.2f} MWh")
    print(f"  Sample COSY stores:")
    print(cosy_stores[['bus', 'carrier', 'e_nom']].head())

Thermal buses (for heat): 313
TANK stores (hot water tanks): 313
COSY stores (building thermal mass): 313

Total TANK storage capacity: 1041.1 MWh
  Mean per location: 3.33 MWh
  Sample TANK stores:
                                         bus    carrier     e_nom
name
CLAC1Q hot water tank tank  heat_CLAC1Q heat  hot water  0.139533
CRAR1R hot water tank tank  heat_CRAR1R heat  hot water  0.620923
DUNO1Q hot water tank tank  heat_DUNO1Q heat  hot water  0.146510
DUNO1R hot water tank tank  heat_DUNO1R heat  hot water  0.146510
KEIT3- hot water tank tank  heat_KEIT3- heat  hot water  1.063942

Total COSY storage capacity: 895.8 MWh
  Mean per location: 2.86 MWh
  Sample COSY stores:
                                                                 bus  \
name
CLAC1Q building thermal mass  thermal inertia_CLAC1Q thermal inertia
CRAR1R building thermal mass  thermal inertia_CRAR1R thermal inertia
DUNO1Q building thermal mass  thermal inertia_DUNO1Q thermal inertia
DUNO1R building thermal mass  thermal inertia_DUNO1R thermal inertia
KEIT3- building thermal mass  thermal inertia_KEIT3- thermal inertia

                                      carrier     e_nom
name
CLAC1Q building thermal mass  thermal inertia  0.122213
CRAR1R building thermal mass  thermal inertia  0.534907
DUNO1Q building thermal mass  thermal inertia  0.128008
DUNO1R building thermal mass  thermal inertia  0.128008
KEIT3- building thermal mass  thermal inertia  0.917529

5. Coefficient of Performance (COP)#

5.1 Understanding COP#

The Coefficient of Performance (COP) determines how efficiently a heat pump converts electricity to heat:

\[\text{Thermal Output (kW)} = \text{Electrical Input (kW)} \times \text{COP}\]

COP varies with temperature:

  • Milder weather (10°C): COP ~4-5 (very efficient)

  • Cold weather (0°C): COP ~2-3 (less efficient)

  • Very cold (-5°C): COP ~1.5-2 (least efficient)

This is why PyPSA-GB uses time-varying COP based on weather data.

[7]:

# Check if COP time series exists (stored as link efficiency)
# COP is on the TANK heat pump links
tank_hp_links = n.links[(n.links['carrier'] == 'heat pump') & (n.links.index.str.contains('tank'))]

if len(n.links_t.efficiency) > 0 and len(tank_hp_links) > 0:
    # Get COP for one heat pump link
    sample_link = tank_hp_links.index[0]

    if sample_link in n.links_t.efficiency.columns:
        cop_series = n.links_t.efficiency[sample_link]

        print(f"COP Statistics for {sample_link}:")
        print(f"  Mean: {cop_series.mean():.2f}")
        print(f"  Min: {cop_series.min():.2f}")
        print(f"  Max: {cop_series.max():.2f}")
        print(f"  Std: {cop_series.std():.2f}")

        # Plot COP over time
        fig, ax = plt.subplots(figsize=(14, 5))
        ax.plot(cop_series.index, cop_series.values, linewidth=1.5, color='green')
        ax.axhline(y=cop_series.mean(), color='red', linestyle='--',
                   label=f'Mean COP: {cop_series.mean():.2f}')
        ax.set_ylabel('COP')
        ax.set_xlabel('Time')
        ax.set_title('Heat Pump Coefficient of Performance Over Time')
        ax.legend()
        plt.tight_layout()
        plt.show()
else:
    print("Note: COP may be stored as static efficiency in links DataFrame")
    if len(tank_hp_links) > 0:
        print(f"Static efficiency values: {tank_hp_links['efficiency'].describe()}")

COP Statistics for CLAC1Q heat pump tank:
  Mean: 3.94
  Min: 3.80
  Max: 4.07
  Std: 0.07
../_images/tutorials_13-heat-flexibility_17_1.png

6. Analyzing Flexibility Behavior#

6.1 Heat Pump Power Consumption#

[8]:

# Get heat pump link power flows (electricity consumption)
# Include both TANK heat pump links and COSY thermal demand links
tank_hp_links = n.links[(n.links['carrier'] == 'heat pump') & (n.links.index.str.contains('tank'))]
cosy_thermal_links = n.links[n.links['carrier'] == 'thermal demand']

if len(n.links_t.p0) > 0 and (len(tank_hp_links) > 0 or len(cosy_thermal_links) > 0):
    tank_link_cols = [c for c in n.links_t.p0.columns if c in tank_hp_links.index]
    cosy_link_cols = [c for c in n.links_t.p0.columns if c in cosy_thermal_links.index]
    hp_link_cols = tank_link_cols + cosy_link_cols

    if hp_link_cols:
        # p0 is positive when power flows from bus0 (electrical) to bus1 (thermal)
        hp_power = n.links_t.p0[hp_link_cols].sum(axis=1)  # Total HP electricity consumption

        print("Heat Pump Electricity Consumption (MW):")
        print(f"  Mean: {hp_power.mean():.1f}")
        print(f"  Peak: {hp_power.max():.1f}")
        print(f"  Min: {hp_power.min():.1f}")

        # Total energy
        hours = len(n.snapshots)
        energy_gwh = hp_power.sum() / 1000
        print(f"\nTotal HP Energy: {energy_gwh:.1f} GWh over {hours} hours")

        # Plot
        fig, ax = plt.subplots(figsize=(14, 5))
        ax.plot(hp_power.index, hp_power.values / 1000, linewidth=1.5, color='orange')
        ax.fill_between(hp_power.index, hp_power.values / 1000, alpha=0.3, color='orange')
        ax.axhline(y=hp_power.mean() / 1000, color='red', linestyle='--',
                   label=f'Mean: {hp_power.mean()/1000:.2f} GW')
        ax.set_ylabel('Heat Pump Electricity (GW)')
        ax.set_xlabel('Time')
        ax.set_title('Total Heat Pump Electricity Consumption')
        ax.legend()
        plt.tight_layout()
        plt.show()

Heat Pump Electricity Consumption (MW):
  Mean: 3274.3
  Peak: 3622.0
  Min: 2747.9

Total HP Energy: 550.1 GWh over 168 hours
../_images/tutorials_13-heat-flexibility_20_1.png

6.2 Thermal Storage Behavior#

[9]:

# Check store state of charge (energy stored over time)
if len(n.stores_t.e) > 0:
    tank_store_cols = [c for c in n.stores_t.e.columns if n.stores.loc[c, 'carrier'] == 'hot water']
    cosy_store_cols = [c for c in n.stores_t.e.columns if n.stores.loc[c, 'carrier'] == 'thermal inertia']

    fig, axes = plt.subplots(2, 1, figsize=(14, 8), sharex=True)

    if tank_store_cols:
        tank_soc = n.stores_t.e[tank_store_cols].sum(axis=1) / 1000  # GWh
        axes[0].plot(tank_soc.index, tank_soc.values, linewidth=1.5, color='blue')
        axes[0].fill_between(tank_soc.index, tank_soc.values, alpha=0.3, color='blue')
        axes[0].set_ylabel('Energy Stored (GWh)')
        axes[0].set_title('TANK Mode: Hot Water Tank Storage')
        axes[0].axhline(y=tank_soc.mean(), color='red', linestyle='--',
                        label=f'Mean: {tank_soc.mean():.2f} GWh')
        axes[0].legend()

    if cosy_store_cols:
        cosy_soc = n.stores_t.e[cosy_store_cols].sum(axis=1) / 1000  # GWh
        axes[1].plot(cosy_soc.index, cosy_soc.values, linewidth=1.5, color='purple')
        axes[1].fill_between(cosy_soc.index, cosy_soc.values, alpha=0.3, color='purple')
        axes[1].set_ylabel('Energy Stored (GWh)')
        axes[1].set_title('COSY Mode: Building Thermal Mass Storage')
        axes[1].axhline(y=cosy_soc.mean(), color='red', linestyle='--',
                        label=f'Mean: {cosy_soc.mean():.2f} GWh')
        axes[1].legend()

    axes[1].set_xlabel('Time')
    plt.tight_layout()
    plt.show()
else:
    print("No store time series data found - check if network was solved with store tracking")
../_images/tutorials_13-heat-flexibility_22_0.png

6.3 Heat Demand vs Electricity Consumption#

[10]:

# Compare thermal demand with electrical consumption
# Thermal demand = heat delivered to space heating
# Electrical consumption = power drawn from grid
# Ratio should approximate mean COP

# Identify loads and links
tank_loads = n.loads[n.loads['carrier'] == 'hot water demand']
cosy_loads = n.loads[n.loads['carrier'] == 'space heating']
tank_hp_links = n.links[(n.links['carrier'] == 'heat pump') & (n.links.index.str.contains('tank'))]
cosy_thermal_links = n.links[n.links['carrier'] == 'thermal demand']

if len(n.loads_t.p_set) > 0 and (len(tank_hp_links) > 0 or len(cosy_thermal_links) > 0):
    # Get thermal demand from both TANK and COSY loads
    tank_load_cols = [c for c in n.loads_t.p_set.columns if c in tank_loads.index]
    cosy_load_cols = [c for c in n.loads_t.p_set.columns if c in cosy_loads.index]

    if tank_load_cols or cosy_load_cols:
        thermal_demand = n.loads_t.p_set[tank_load_cols + cosy_load_cols].sum(axis=1)

        # Get electrical consumption from HP links (TANK + COSY)
        tank_link_cols = [c for c in n.links_t.p0.columns if c in tank_hp_links.index]
        cosy_link_cols = [c for c in n.links_t.p0.columns if c in cosy_thermal_links.index]
        hp_link_cols = tank_link_cols + cosy_link_cols

        if hp_link_cols:
            elec_consumption = n.links_t.p0[hp_link_cols].sum(axis=1)

            # Calculate effective COP
            effective_cop = thermal_demand / elec_consumption.replace(0, np.nan)

            print(f"Thermal Demand: {thermal_demand.sum()/1000:.1f} GWh")
            print(f"Electrical Consumption: {elec_consumption.sum()/1000:.1f} GWh")
            print(f"Effective System COP: {effective_cop.mean():.2f}")

            # Plot comparison
            fig, axes = plt.subplots(2, 1, figsize=(14, 8), sharex=True)

            axes[0].plot(thermal_demand.index, thermal_demand.values / 1000,
                        label='Thermal Demand', color='red', linewidth=1.5)
            axes[0].plot(elec_consumption.index, elec_consumption.values / 1000,
                        label='Electrical Consumption', color='blue', linewidth=1.5)
            axes[0].set_ylabel('Power (GW)')
            axes[0].set_title('Heat Pump Energy Balance')
            axes[0].legend()

            axes[1].plot(effective_cop.index, effective_cop.values, color='green', linewidth=1.5)
            axes[1].axhline(y=effective_cop.mean(), color='red', linestyle='--',
                           label=f'Mean COP: {effective_cop.mean():.2f}')
            axes[1].set_ylabel('Effective COP')
            axes[1].set_xlabel('Time')
            axes[1].set_title('Effective Coefficient of Performance')
            axes[1].legend()

            plt.tight_layout()
            plt.show()
        else:
            print("No heat pump links found")
    else:
        print("No thermal loads found")
else:
    print("Load or link time series not available")

Thermal Demand: 718.9 GWh
Electrical Consumption: 550.1 GWh
Effective System COP: 1.32
../_images/tutorials_13-heat-flexibility_24_1.png

7. Configuration Options#

Heat pump flexibility is configured in config/defaults.yaml:

heat_pumps:
  enabled: true
  flexibility:
    enabled: true
    tank_share: 0.5    # 50% of HP demand uses TANK mechanism
    cosy_share: 0.5    # 50% of HP demand uses COSY mechanism
    tank_storage_hours: 4.0  # Hours of storage at rated capacity
    cosy_storage_hours: 2.0  # Hours of thermal inertia
    tank_standing_loss: 0.01  # 1% per hour heat loss
    cosy_standing_loss: 0.05  # 5% per hour (less insulated)

Key Parameters:#

Parameter

Description

Typical Value

tank_share

Fraction of HP demand with hot water tank flexibility

0.3-0.5

cosy_share

Fraction of HP demand with building thermal inertia

0.3-0.5

tank_storage_hours

Tank storage duration at rated power

2-6 hours

cosy_storage_hours

Building thermal storage duration

1-3 hours

tank_standing_loss

Hourly heat loss from tank

0.5-3%

cosy_standing_loss

Hourly heat loss from building

3-10%

8. Summary#

Key Takeaways:#

  1. Both TANK and COSY are space heating mechanisms - they differ in where heat is stored:

    • TANK: Hot water cylinder (more storage, less loss)

    • COSY: Building fabric (less storage, more loss)

  2. COP is crucial - Heat pumps produce 2-5x more thermal energy than electrical input

    • Higher COP in mild weather, lower in cold weather

    • Time-varying COP from weather data

  3. Flexibility helps the grid by:

    • Shifting demand away from peak hours

    • Better utilizing renewable generation

    • Reducing need for expensive peaking plants

  4. PyPSA components:

    • Links (heat pumps): Electricity → Thermal conversion at COP efficiency

    • Stores (tanks/buildings): Thermal energy storage

    • Loads (space heating): Final thermal demand