All research datasets used in this paper are officially authenticated operation indicators exported from China State Railway Group freight dispatching big data platform, with statistical period from Jan. 1, 2020 to Dec. 31, 2025. Data coverage includes five empirical trunk lines (Datong-Qinhuangdao, Haoji, Shuohuang, Datang, Jingtong Railway) and national bulk freight macro statistical indicators. Original data records daily cargo throughput, wagon usage, locomotive energy consumption and train operation mode with daily time granularity, total effective raw sample data reaches 192,600 groups. In preprocessing phase, abnormal data caused by unexpected line maintenance, natural disaster shutdown and temporary policy adjustment is eliminated using the 3σ statistical discrimination rule; yearly and averaged monthly standardized panel data are formed for subsequent characteristic analysis and model parameter input to guarantee experimental reproducibility ([1]).

Bulk cargo is the main component of railway freight, and coal, iron ore and grain are the three dominant cargo types in railway bulk transportation. The freight volume changes and proportion distribution of main bulk cargoes from 2020 to 2025 are shown in Table 1, all data are derived from the official statistical annual report of China Railway and on-site operation settlement data of trunk lines ([5]; [7]).

It can be seen from the statistical data that coal has always been the core bulk cargo of railway transportation, accounting for nearly 70% of the total freight volume. The freight volume of grain maintains a stable growth, with an average annual growth rate of 7.2%, reflecting the important supporting role of railway transportation in ensuring national material security. In terms of structural changes, the proportion of single bulk cargo tends to be stable, and the overall freight structure presents the characteristics of concentration and specialization, which provides a basic condition for the popularization of unit train operation and low-carbon organizational optimization.

Table 1. Freight volume and proportion distribution of main bulk railway cargoes (2020-2025).

Combined with on-site railway operation management specifications, the current bulk cargo transportation organization modes are divided into unit train, mixed marshalling train and combined train. The operational efficiency and application proportion of the three modes in 2025 are shown in Table 2, which are all actual statistical data of railway trunk line operation ([11]; [3]).

Table 2. Operational indicators of different railway bulk cargo transportation organization modes.

The unit train adopts fixed origin-destination and fixed marshalling operation, with no intermediate marshalling operation, which has the highest loading rate and the lowest empty running rate, and is the most efficient low-carbon operation mode. However, affected by scattered cargo sources and flexible transportation demands, the proportion of mixed marshalling trains in actual operation is still as high as 38%. The frequent shunting and marshalling operations of mixed trains lead to prolonged transit time, low wagon utilization rate and serious invalid energy consumption, which are the main reasons for the high carbon emission of current bulk railway freight.

According to IPCC greenhouse gas inventory accounting standards and China’s railway transportation carbon emission management specifications, the carbon emissions of railway bulk cargo transportation are divided into direct emissions from diesel locomotive combustion, indirect emissions from electric locomotive power consumption, and auxiliary emissions from station and loading and unloading equipment energy consumption. The proportion of the three emission sources in actual operation is 35%, 55% and 10% respectively, and electric power consumption is the dominant emission source of current railway freight ([9]).

Carbon emission intensity is a core indicator to measure the low-carbon level of transportation, representing the carbon dioxide emissions per unit of freight turnover. The actual emission intensity of different cargo types and organizational modes in 2025 is shown in Table 3.

Table 3. Carbon emission intensity of different bulk cargoes and transportation modes.

It is obvious that the organizational mode is the decisive factor affecting carbon emission intensity. The carbon emission intensity of mixed trains is nearly twice that of unit trains. The fundamental reason is that mixed marshalling increases operation links, reduces equipment utilization efficiency, and generates a large amount of invalid carbon emissions. Therefore, optimizing the transportation organization structure and reducing mixed marshalling operation is the key to realize low-carbon upgrading of railway bulk freight ([13]).

Based on the emission factor method which is universally recognized by the industry and adapted to railway on-site measurement, the total carbon emission of bulk cargo transportation is the sum of direct emission, indirect emission and auxiliary emission. The accounting boundary covers the whole process from cargo departure station to destination station, including intermediate marshalling, loading and unloading, and station operation links, excluding the emission of infrastructure construction and equipment manufacturing ([12]).

where E_total is the total carbon emission of railway bulk transportation (t CO₂); E_direct is the direct carbon emission generated by diesel locomotive fuel combustion; E_indirect is the indirect carbon emission generated by electric locomotive power consumption; E_auxiliary is the auxiliary carbon emission of station operation and loading and unloading equipment.

Direct diesel emission is calculated by fuel consumption and diesel emission factor, and the default IPCC diesel emission factor is 3.16 t CO₂/t:

where F_i is the actual diesel consumption of the i-th diesel locomotive; EF_diesel is the diesel carbon emission factor.

Indirect power emission adopts the national average power grid carbon emission factor of 0.58 t CO₂/MWh in 2025:

where P_j is the actual power consumption of electric locomotives; EF_grid is the power grid carbon emission factor.

Auxiliary emission is calculated based on the power consumption of station and loading and unloading equipment:

Carbon emission intensity, the core evaluation index, is calculated by total emission divided by total freight turnover:

where Q is total freight volume and L is average transportation distance.

In order to verify the accuracy of the accounting model, this paper selects the typical Datong-Qinhuangdao coal dedicated line for verification. After recalculation with unified unit conversion and standard emission factor calibration, revised energy consumption and emission data are listed in Table 4 and Table 5; Datong-Qinhuangdao line is selected for verification owing to its full-specialized coal transportation attribute, stable operation organization and complete real-time energy consumption monitoring system, which can represent the operation characteristics of mainstream bulk dedicated railways nationwide ([6]).

Table 4. Basic operation parameters of unit train and mixed train for verification.

Table 5. Carbon emission verification results of different transportation organization modes.

The calculated MAPE (mean absolute percentage error) between model accounting result and field actual monitored emission data is 2.13%, which is taken as quantitative verification index to prove the reliability of the carbon accounting model.

Explicit definition of core decision variables in advance:

Ru: Operation proportion of unit train (decision variable 1);

Rm: Operation proportion of mixed train (decision variable 2);

Rc: Operation proportion of combined train (decision variable 3);

Wk: Allocated wagon quantity fork-th transport route;

Qk: Actual arranged cargo volume on route k.

Model optimization targets the adjustment of three train mode proportions, wagon quantity allocation and route cargo flow assignment for each trunk line.

Cost parameter supplementary description:

Fixed cost C_fixed,k = 8200 RMB/wagon·day (source: China Railway freight cost quota standard 2025);

Variable cost C_variable,k is composed of diesel and electricity energy cost, diesel unit price 7800 RMB/t, electricity price 0.52 RMB/kWh, parameters sourced from national railway published operation cost accounting manual.

Aiming at the three core goals of low carbon, low cost and high efficiency in railway bulk freight operation, this paper constructs a multi-objective constrained optimization model, which fully considers line capacity limit, wagon marshalling specification, minimum loading rate and other on-site hard constraints.

Objective 1: Minimize total carbon emission of transportation operation

Objective 2: Minimize comprehensive transportation cost, including fixed equipment cost and variable energy consumption cost

Objective 3: Maximize wagon loading rate and transportation operation efficiency

Constraint conditions include line capacity constraint, wagon quantity marshalling constraint, minimum loading rate constraint and non-negative constraint of decision variables, which are all formulated according to railway on-site operation management standards:

Add cargo demand full-satisfaction constraint: ∑ Q k = Q t o t a l _ d e m a n d , which ensures the total arranged cargo quantity after optimization is equal to original market total freight demand, emission reduction is realized only via organizational optimization rather than cutting transport volume.

The constructed multi-objective optimization model is a typical nonlinear NP-hard problem, which is difficult to solve by traditional linear programming methods. This paper adopts a hybrid genetic algorithm fused with particle swarm optimization to solve the model, which overcomes the defects of local optimal solution and slow convergence of the traditional genetic algorithm. The algorithm parameters are set as follows: population size 100, maximum iteration 200, crossover rate 0.85, mutation rate 0.15, PSO inertia weight 0.7, learning factor 2.0. The entropy weight method is used to determine the objective weight to avoid subjective weighting error, which ensures the objectivity and accuracy of the optimization results.

The core optimization idea is to adjust the structural proportion of transportation organization modes, increase the operation scale of high-efficiency and low-emission unit trains, reduce the proportion of mixed marshalling trains, and optimize wagon allocation and loading scheduling to improve loading rate and reduce empty running ratio. After optimization, the proportion of unit trains is increased from 45% to 68%, the proportion of mixed trains is reduced from 38% to 12%, and the proportion of combined trains is adjusted to 20%. Total freight volume after optimization keeps consistent with original market total cargo demand constrained by ∑ Q k = Q t o t a l _ d e m a n d , no shrinkage of overall transportation volume exists.

The comparison of core operation indicators before and after optimization is shown in Table 6, all data are authentic simulation results based on actual line parameters.

Table 6. Comparison of operational indicators before and after optimization.

The empirical results show that the optimized transportation organization scheme achieves significant comprehensive benefits in low-carbon emission reduction, cost control and efficiency improvement. From the perspective of low-carbon benefit, the overall carbon emission intensity of railway bulk freight is reduced by 15.8%, which effectively solves the problem of invalid carbon emission caused by unreasonable organization. From the perspective of economic benefit, the comprehensive transportation cost is reduced by 9.5%, which greatly reduces the social logistics cost of bulk materials. From the perspective of operational benefit, the wagon loading rate is significantly improved, the empty running waste of transportation capacity is reduced, and the freight transportation timeliness is greatly improved.

In order to verify the robustness of the optimization model, this paper carries out sensitivity analysis on key parameters such as power grid emission factor, minimum loading rate and unit train proportion. The results show that when the power grid carbon emission factor fluctuates by ±10%, the total carbon emission fluctuates by only ±8.5%, and the optimization scheme remains stable. When the minimum loading rate is increased from 80% to 85%, the carbon emission is further reduced by 3.2% with only 1.5% cost increase, showing excellent cost-emission reduction synergy. When the unit train proportion exceeds 68%, the marginal emission reduction benefit decreases and the cost increases significantly, which verifies that 68% is the optimal proportion of unit train operation for current bulk railway freight.