Cost increase in the electricity supply to achieve carbon neutrality in China

The Chinese government has set long-term carbon neutrality and renewable energy (RE) development goals for the power sector. Despite a precipitous decline in the costs of RE technologies, the external costs of renewable intermittency and the massive investments in new RE capacities would increase electricity costs. Here, we develop a power system expansion model to comprehensively evaluate changes in the electricity supply costs over a 30-year transition to carbon neutrality. RE supply curves, operating security constraints, and the characteristics of various generation units are modelled in detail to assess the cost variations accurately. According to our results, approximately 5.8 TW of wind and solar photovoltaic capacity would be required to achieve carbon neutrality in the power system by 2050. The electricity supply costs would increase by 9.6 CNY¢/kWh. The major cost shift would result from the substantial investments in RE capacities, flexible generation resources, and network expansion.


List of Supplementary Figures
Supplementary Figure 1 Illustration

List of Supplementary Tables
Supplementary Table 1 Total electricity energy demands under different scenarios(TWh  Table 7 The R-squared value of the transmission line length fitting results for each region . . . 37 Supplementary Table 8 The R-squared value of the transformer capacities fitting results for each region . . . 37 Supplementary Table 9 Projection results of within-province transformer capacities in CN2050 (GW during the manufacturing process of the units are not taken into account. 50 • The base year of market prices is set to 2020, which means the prices are normalized based on Chinese Yuan in 2020. In 51 the main article, the exchange rate of CNY to USD is set to 0.144982, which is the average rate in 2020. The operating costs include the penalty costs of load shedding in each region, the power generation costs, and the start-up 70 costs for each unit. The CO 2 capture costs are also included when carbon capture and storage (CCS) units are involved. impacts. This constraint is formulated as: where U GP m,r is the upper limit of generation plants for technology m in region r.

88
For different units, the installed capacity at a certain planning stage depends on the installed capacity at the previous stage, the expansion in capacity and the retired capacity at the current stage. The capacity increments in each region are constrained due to the limits on construction ability and environmental issues. The following continuity constraints are met: The supply of natural gas resources in China is tight and heavily dependent on imports. Hence, generation unit (GU) 99 expansion planning must consider the natural gas resource constraints faced by each province. The amount of natural gas used 100 for power generation cannot exceed the available natural gas power generation resources in the region. The gas networks are 101 also modelled: where η m denotes the gas consumption per MWh for the technology m. Γ r,y denotes the natural gas supply for power generation 103 in the region r including the local production and imports from foreign countries. The gas import Γ For each provincial region, the total power generation water consumption at each planning stage cannot exceed the given 108 power generation water consumption limit, that is: The total CO 2 emitted by the electrical sector is supposed to meet the given carbon emission reduction goals, that is: • Power Reserve Requirements The unit capacity in each region is supposed to be larger than the local peak load to guarantee security under unexpected 121 accidents. The proportion of the excess part is called the reserve rate, which needs to meet the requirements in each provincial 122 power system. The expressions is as follows: where rs r is the required reserve rate in region r (around 13%-15%) and r m is the credit capacity rate of generation technology m. Since the outputs of renewable energy units, such as wind and PV, are intermittent and not dispatchable, their credit capacity 125 rates are lower than those of conventional units. The item on the left of the inequality sign represents the power reserve 126 provided by the local generators and the transmission grid, and the item on the right represents the local reserve demand in the 127 region r (Note that a region corresponds to a province here). r m U m,r,y denotes the power reserve capacities provided by the local the subset of the whole DC transmission lines whose sending end is region r. rs r is the required reserve rate in region r. In  between provinces when considering the national transmission power system planning with high voltage levels. Currently, 152 many planning studies for national or regional power systems have adopted the pipeline model 3, 9-13 . In our model, each bus 153 corresponds to an aggregation of a provincial power grid, not a real bus in the power system as stated in Supplementary Note 154 1.1. Thus, the free control of power flow on the AC transmission lines can be achieved by the line switch operation or the coordinated dispatch of reactive and active power within the province grid. For DC transmission lines, free control is their inherent advantage thanks to the power-electronic control technologies. Hence, the pipeline model is reasonable for our case.

157
The modelling of power losses is also smoother when applying the pipeline model. Line losses are assumed to be proportional 158 to the online power flow. The expressions for AC transmission lines are as follows: where ρ m denotes the maximum capture rate, with a typical value between 80 % and 95 %: it is taken as 90 % in this paper. For intermittent generation units such as wind power and PV, the actual output during operation cannot exceed the maximum 186 generation output, that is: where ω IRE m,r,y,d,t denotes the maximum generation output at time period t.

189
Pumped hydro storage and battery storage are considered in the model. The following constraints need to be met during 190 operation: S ESS m,r,y,d,t=0 = S ESS m,r,y,d,t=T , ∀m, r, y, d where η ESS operation is as follows: S TES m,r,y,d,t=0 = S TES m,r,y,d,t=T , ∀m ∈ Ω CSP , r, y, d 0 ≤ P m,r,y,d,t ≤ U m,r,y , ∀m ∈ Ω CSP r, y, d,t where η PB m denotes the thermoelectric conversion efficiency and ω SF m,r,y,d,t denotes the capacity factor at time period t. The actual output of hydropower units should not exceed the total installed capacity during operation. Notably, the annual 205 energy generation constraint of hydropower is reflected in (22).
• Spinning Reserve Requirements during Operation

207
Spinning reserve constraints represent the flexibility requirements during power system operation. Due to the uncertainties 208 in the load and VRE output, the actual value may fluctuate greatly in a short time. Hence, the system must have sufficient 209 spinning capacity to supplement potential power shortages. Here, we define the spinning capacity as the output increments 210 generators and ESSs can provide within 10 minutes. The expression of spinning requirements is presented in (54).
where P hot m,r,y,d,t denotes the spinning reserve capacities unit m in bus r can provide in time period t of representative day d at 212 stage y. The spinning reserves are set to address fluctuations in load and variable renewable energy. The three terms on the right 213 side present the spinning demand caused by load, wind and PV power. hr r is the spinning factor, which is set to 5% according 214 to planning criteria and grid codes in China 15 . The spinning capacities must cover potential shortages. The spinning reserves 215 provided by various units are as follows: 0 ≤ P hot m,r,y,d,t ≤ min U on m,r,y,d,t − P net,CCS m,r,y,d,t , rp m ·U on m,r,y,d,t , ∀m ∈ Ω CCS , r,t, d, y 0 ≤ P hot m,r,y,d,t ≤ U GP m,r,y − P m,r,y,d,t , ∀m ∈ Ω HU , r,t, d, y 0 ≤ P hot m,r,y,d,t ≤ min U m,r,y − P m,r,y,d,t , 6 · S TES m,r,y,d,t · η PB m , ∀m ∈ Ω CSP , r,t, d, y where rp m denotes the maximum ramp rate within 10 minutes for unit m. The term, 6 · S ESS m,r,y,d,t · η ESS where the coefficient h m denotes the inertia constant of unit m. The total inertia that can be provided by thermal units and The LCOE of VRE increases remarkably due to the decrease in capacity factors and the increase in the difficulty of construction 260 and grid integration with the growth of installed capacity. This variation must be considered in the model. Without considering 261 the spatial distribution of LCOE, the cost will be underestimated by 2.2 CNY¢/kWh, as discussed in the main paper. We 262 assess the LCOE for every 500m×500m plot in each province based on the GREAN dataset and the method described in 263 Fig. 6. Consequently, the provincial VRE supply curves can be obtained, as shown in Supplementary Figs. 7 and 8. Supply 264 curves characterize the quantity, quality, and cost of renewable resources. The reason for the change in LCOE is due mainly 265 to the differences in capacity factors and grid-connection costs. According to our assumptions, only one aggregated unit is 266 considered in each province to simplify the model. Naturally, one unit can correspond to only one capacity factor, which is set 267 as the average value over the province in this paper. Hence, the supply curves cannot be integrated into the GTEP model, an 268 optimization problem directly where the total generation costs consist of operating costs, maintenance costs and capital costs.

269
Here, we convert the supply curves of LCOE into supply curves of capital cost considering no fuel costs and low proportion of 270 maintenance costs for wind and PV power. The calculation is as follows: function package implemented in python to fit the curve. The differential evolution optimization algorithm, a popular heuristic 280 algorithm, is used in this package to find the best location for the user-defined number of line segments, which is set to seven.

281
The width of the segments U The electricity supply cost is the average cost the power system must pay to supply per kWh of electricity load demand. For a single stage, this value is the ratio of the total cost caused by power generation and transmission to the load demand in the stage, as follows: where c y denotes the electricity supply cost at stage y. c total y denotes the total cost caused by power generation and transmission in year y. ∑ where U type m,r,y,y i denotes the remaining capacity of unit m that is invested during stage y i of stage y in province r. When y i is equal to zero, U type m,r,y,0 denotes the remaining capacity of unit m that originally exists in the system at stage y. The planning period in GTEP model is from 2020 to 2050, and the variation in capital costs before 2020 is not considered in the calculation. We assume that the capital costs of existing devices are equal to the value in 2020. For wind and PV units, whose developing potential is split into several segments, the expression is modified as follows:  However, the objective function is the present value at the end of 2020. The original shadow prices are supposed to be converted to a future value in the corresponding stage. The expressions are as follows: where λ raw y denotes the original shadow price provided by the solver for carbon emission limits at stage y. i is the discount rate, 305 and λ y is the converted margin prices in future values.

307
Average carbon mitigation costs are the additional costs per tonne of carbon emission between two scenarios. This value is numerically equal to the ratio of the difference between the total cost of the two scenarios and the difference between the carbon emission budget. The expression is as follows: where λ n,m denotes the average carbon mitigation costs between scenarios n and m. C n and E n are the total costs and carbon 308 emission budget during the planning period for scenarios n, respectively.

309
Note that there is no direct relationship between marginal carbon prices and average carbon mitigation costs since they are   Supplementary Fig. 3. The manufacturing capability of pumped hydro storage (PHS) is also not considered in the sensitivity 336 analysis because of its mature technical level.

338
We assume that load demand would fluctuate 5% compared with the baseline scenario, i.e., the CN2050 scenario. The high 339 scenario assumes that demand increases linearly 5% until 2050. The low scenario assumes that demand decreases linearly 5% 340 until 2050. The total electricity energy demands under different scenarios are presented in Table 1.

342
Three kinds of secure reserve constraints are considered in the GTEP model in both the planning and operation periods. The 343 three constraints are the power reserve requirements (24), hot reserve requirements(54), and minimum system inertia limits(61).

344
To analyse the impacts of security requirements on electricity supply costs, we set different secure requirement levels with 345 different parameter settings on rs, hr, and α, as shown in Table 2. The parameter setting of the low scenario is equivalent to  Table 3.   across the whole country must be considered when optimizing RE investments and regional network connections.

368
The RE potential varies greatly not only between provinces but also in different areas within a single province. Supplemen-369 tary Fig. 7 and Supplementary Fig. 8 present the spatial VRE capacity factor distribution and regional LCOE in each province  shown in Supplementary Fig. 7 and Supplementary Fig. 8. Details on the piecewise calculation method are presented in        Land cover data covers the land range from 80°degrees north latitude to 80°south latitude released by National Geomatics Center of China, with a resolution of 30m×30m. (Raster data).

Global distribution of major reservoirs
Reservoirs data is from the global water system projects in Bonn, Germany, including more than 6,500 artificial reservoirs with a cumulative storage capacity of about 6.2 trillion m3 (Raster data).

Global distribution of lakes and wetlands
Lakes and wetlands data is jointly developed by the World Wide