A Two-Stage Decision-Making Method Based on WebGIS for Bulk Material Transportation of Hydropower Construction

Abstract

Bulk materials are necessary for hydropower construction. The bulk materials transportation (BMT) scheme is a guiding document for material supply, and its selection has a significant influence on hydropower construction. Since the BMT problem includes transportation planning and scheme selection issues simultaneously, only a small number of studies have focused on it. This paper presents a theoretical two-stage decision-making method (TDM), which innovatively combines the path optimization method and the multi-criteria decision-making (MCDM) method to solve the BMT problem. In the first stage, a multi-source path optimization model is established to optimize the transportation network and obtain a set of alternatives from each supply point to the construction site. In the second stage, considering the factors of economy, risk and construction progress, the MCDM method is adopted to select the optimal scheme from the alternatives. In addition, web crawler technology is used to obtain the transportation network data from the public WebGIS automatically. Case results show that the TDM can effectively solve this problem, and its result keeps consistent with engineering practice; with the help of the web crawler, it can reduce the design task time from months to days. Therefore, the TDM based on WebGIS can benefit hydropower construction design efficiency.

1. Introduction

Renewable energy is currently the hotspot of international energy development and will gradually become the world’s primary energy source in the future [1,2]. Hydropower projects are the most concentrated way to obtain renewable energy and effectively stimulate the economic growth of surrounding areas [3,4]. However, hydropower construction generally consumes large amounts of building materials. Those construction materials can be logically grouped into two major categories: (1) materials fabricated on the construction site, such as sand–gravel aggregate; (2) materials fabricated or further processed off-site and that need to be transported from outside the construction site, such as cement, fly ash and steel—this category is named the bulk materials in this paper [5]. The bulk materials transportation (BMT) scheme selection involves the selection of a key guidance transportation scheme from supply points to the construction site. Furthermore, the BMT scheme is closely related to project investment, construction progress and engineering quality. A reasonable BMT scheme is the critical foundation and guarantee of hydropower construction. Currently, there is no systematic and reliable theoretical model for the BMT problem. The planning of the BMT relies heavily on an expert’s experience or stakeholder’s will [6], which are susceptible to personal preferences and cognitive differences. Hence, it is essential to carry out relevant research on the BMT problem in hydropower projects.

The scheme selection of the BMT for hydropower construction needs to comprehensively consider the characteristics of the transportation network and various influencing factors, including the economy, risk and construction progress. From the features of global water power development, those hydropower projects are mostly located in remote mountainous areas, while the bulk materials suppliers are primarily located in the industrial centers and far away from the construction site. Figure 1 provides an overview of the bulk materials transportation of a hydropower project, and the features of the transportation network are as follows:

(1)

A variety of bulk materials are involved—each one has multiple suppliers who meet all of the project’s requirements for productivity and material quality conditions; i.e., there are multiple alternative starting points in the transportation network.

(2)

Multiple optional transportation routes from the material suppliers to the construction site can be chosen, and each route has its own combination of transportation modes, including highway, railway or waterway.

(3)

Since most of the hydropower projects are located in remote mountainous areas, the original transportation conditions of most routes are poor and cannot meet the transportation requirements. Refurbishment, expansion and even new transportation facilities are required, including roads, bridges and tunnels for transportation. Therefore, the BMT of the hydropower projects considers both the existing transportation network and the reconstructed and newly built transport facilities mentioned above.

Concerning the above features, the planning of the BMT needs to consider the material suppliers, the transportation network, the transportation modes and the reconstructed and newly built transport facilities. Hence, it can be taken as a complex multi-source path optimization (MSPO) problem from the perspective of transportation route planning.

However, besides the characteristics of the above-mentioned transportation network, a good transportation scheme also needs to meet the conditions of the economy, safety and construction progress. The economy factor is generally the most important objective for decision-makers, the safety factor reflects the transportation risk related to life and property, and the construction progress is mainly a restriction on the transportation duration, which affects the project construction progress. Hence, to select the optimal scheme, the investment, the transportation risk and the transportation duration should all be considered to form the decision-making basis. Hence, from the perspective of the scheme selection, the BMT problem can be a multi-criteria decision-making (MCDM) problem as well [7].

In summary, bulk materials transportation planning is essentially a transportation problem involving both path optimization and multi-criteria decision-making. Currently, there is no systematic research on the BMT problem in hydropower projects [8]. In similar research fields, such as dangerous goods transportation [9], logistics [10] and urban traffic planning [11], the research is generally carried out by path optimization. These studies transform the transportation network into a path–node topology relationship network, use the path’s edge weight to reflect the optimization objectives comprehensively and adopt the optimal path algorithm to find the optimal transportation scheme [12,13,14]. For the scheme decision-making problem, the commonly used research method is to establish a decision-making model by selecting the main influencing factors as evaluation indicators and then adopt an appropriate MCDM solution algorithm to select the optimal scheme [15,16,17]. Although the existing methods could provide certain references, they can only solve one aspect of the BMT problem [18], are poor in terms of engineering practicability and are challenging to apply to the large-scale transportation planning problem. Hence, the current research gap is revealed; i.e., systematic and practical research specifically for the BMT problem from the whole perspective is needed to solve the path optimization and decision-making problems.

This paper addresses the BMT problem of hydropower projects and presents a feasible and effective theoretical method to select the optimal scheme, named the two-stage decision-making method (TDM), which solves the path optimization problem and the decision-making problem in two stages [19,20]. In the first stage, an MSPO model is established to solve the path optimization problem and obtain alternatives instead of a manual solution. In the second stage, an MCDM method is adopted to select the optimal scheme from the alternatives. The results of the MSPO are the basis of the MCDM. Meanwhile, to improve the efficiency of information collection and data calculation, with the help of Python, a web crawler technology is used to obtain public transportation network information from the public WebGIS platform and realize the complex data calculation.

The rest of the paper includes a literature review, which provides an overview of relevant methodologies and literature, summarizes the TDM and web crawler technology based on WebGIS and proves the rationality and feasibility of the TDM by comparing it with traditional methods. The methodology explains the TDM in detail. A case study demonstrates the process of the TDM based on WebGIS regarding the calculation and scheme selection, followed by a discussion, and finally, conclusions are presented.

2. Literature Review

This section provides a comprehensive literature review about existing studies on the transportation problem of engineering materials from two main aspects: path optimization methods and MCDM methods. In addition, this section also presents the research gap between existing methods and practical engineering applications and proves the rationality and feasibility of the TDM by comparing it with traditional methods.

Among existing literature on the transportation problem of engineering materials, studies about the path optimization problem mainly focus on the optimization algorithm [21,22]. The commonly-applied path optimization methods include the shortest path search algorithm [23], the ant colony intelligent algorithm [24], the Dijkstra algorithm [25], the Floyd algorithm [26], etc. For example, Wu et al. [27] designed a new fuzzy scheduling optimization system based on the ant colony algorithm for multi-objective transportation paths. Woerz et al. [28] applied Dijkstra’s algorithm to find a transport route with minimal fuel consumption. Zhang et al. [29] applied the genetic algorithm (GA) to search for the optimal path of sheep transportation and designed the model solving algorithm based on the basic GA. Moreover, the objective is also critical for path optimization, which should reflect the characteristics of each path section in the transportation network. The typical optimization objectives include transportation costs, time, risks, economic benefits, etc. [30,31,32]. However, for the BMT problem in hydropower projects, the time and the risks represent the characteristics of a whole transportation scheme; in addition, they are global characteristics and are difficult to be characterized through the edge weights, which need to be considered from a global perspective. Therefore, adopting the path optimization to solve the BMT problem of hydropower projects is insufficient due to in the failure to reflect these global characteristics.

The transportation scheme selection for the engineering materials is a compositive decision-making problem, which is often solved by multi-criteria decision-making (MCDM) methods [33,34,35]. The commonly-applied MCDM methods include the Analytic Hierarchy Process [36], grey evaluation [37], the fuzzy set method, and the Technique for Order Preference by Similarity to an Ideal Solution (TOPSIS) [38]. Garg and Kumar [39] extended the traditional TOPSIS method into an interval-valued intuitionistic fuzzy (IVIF) set environment to select the best company for a project. Bandeira et al. [40] proposed a fuzzy MCDM model to select alternative configurations for sustainable urban freight transportation. Büyüközkan et al. [41] presented a group decision-making technique based on an integrated intuitionistic fuzzy Choquet integral for selecting sustainable urban transportation alternatives. Although the existing methods could provide certain references, they cannot perfectly fit the application of the evaluation and selection of the BMT alternatives from two main aspects: (1) the method only can evaluate the existing alternatives that have been planned by artificiality, neglecting the research on how to obtain the alternatives, and cannot be applied to the large-scale transportation planning problem; (2) the method may lead to centralized optimization for the similar schemes because the attributes of the alternatives that have the same transportation starting point are less different in terms of the investment, the transportation duration or the transportation risk; (3) the existing research did not consider the infrastructure investment that characterizes the investment into refurbishing, expanding and building new transport facilities to transport bulk materials, and the infrastructure investment is an indispensable part of the BMT of hydropower projects.

From the literature review, the current research gap is revealed: i.e., systematic solution methods specifically for the BMT problem from the whole perspective are needed but lacking. Only adopting the path optimization method or the decision-making method cannot solve the BMT problem because the problem involves both path optimization and decision-making. This paper presents a TDM from a holistic view of the BMT problem based on the consideration of the transportation planning and the scheme selection [42]. This TDM solves the BMT problem in two stages. The MSPO model is established in the first stage to solve the transportation planning problem. The investment that contains the infrastructure investment is selected as the optimization objective because it can be characterized through the edge weights. The Floyd algorithm is adopted to solve the MSPO model and find the optimal transportation routes arriving at the construction site for each transportation starting point [43]. These transportation routes are considered as alternatives that are the basis of the second stage. Then, the MCDM method is adopted in the second stage to solve the decision-making problem. The investment, the transportation duration and the transportation risk are selected as the decision objectives, all of which are the BMT scheme’s fundamental properties. Among the commonly applied MCDM methods, TOPSIS is selected because it suits decision scenarios with mostly tangible attributes as a high degree of numerical calculation is involved.

Table 1. Comparative information between the TDM and the traditional methods.

	Methods	Two-Stage Decision-Making Method (TDM)	Decision-Making	Path Optimization
Features
Addressed problems		The bulk materials transportation problem of hydropower projects.	The scheme evaluation problem for alternatives that have been planned previously.	The transportation path optimization problem.
Research content		The research content is BMT planning for hydropower projects, which involves both the MSPO problem from the perspective of the transportation route planning and the MCDM problem from the perspective of the scheme selection.	The multi-criteria decision is the main research method; the research content includes selecting the decision objectives and selecting the decision methods.	The method uses the edge weight of the path to comprehensively reflect the optimization objectives and selects the optimal path algorithm to find the optimal transportation route.
Research objects		Transportation network and scheme properties of the BMT as a whole.	Alternatives planned previously.	Transportation network.
Research perspective		A global perspective from the transportation route planning to the scheme selection.	Optimal scheme selection.	Transportation route optimization.
Methods evaluation		The method analyzes the transportation network and the scheme properties of the BMT systematically and can reasonably and ideally solve the BMT by solving the path optimization problem and the scheme selection problem in stages; the result is consistent with actual engineering.	The method cannot be applied to large-scale transportation planning problems, may lead to centralized optimization for similar schemes and cannot characterize the infrastructure investment.	The method cannot reflect holistic indicators, such as transportation duration and risks, so the optimization result could not meet the construction progress requirements and transportation safety.

shows the comparative information between the TDM and the traditional research methods, including the addressed problems, the research content, the research objects, the research perspectives and the evaluation for the research methods. The comparison shows that the TDM has apparent advantages over the traditional methods and can solve the practical transportation problems for the bulk materials well.

In addition, the web crawler technology based on the public WebGIS platform should be highlighted because it provides great convenience for data collection and calculation and makes it possible to realize large-scale transportation planning [44]. The WebGIS is the combined outcome of the web technology and the GIS technology, making it possible to conduct the integration, query, analysis and release of the GIS spatial data on the internet [45]. With the rapid development of internet and GIS technology, the WebGIS platforms, such as Baidu Maps and Google Maps, already have a great deal of geographic information and provide the Application Programming Interface (API) for extended applications. Hence, the web crawler technology is adopted to obtain public transportation network information from the public WebGIS platform.

3. Methodology

3.1. Decision-Making Framework

Based on the above analysis, there is no systematic and reliable research method to solve the BMT problem of hydropower projects. In engineering design, the planning of the BMT mainly relies on expert experience, which is too subjective. In this paper, by studying the characteristics of material transportation, it is found that the bulk materials transportation planning in hydropower projects is essentially a transportation problem involving both path optimization and decision-making. The two-stage decision-making method based on WebGIS is developed to solve the problem from the whole perspective. The TDM contains two stages: in the first stage, the MSPO model is established with the investment as the optimization objective, and the Floyd algorithm is used to solve the optimization model and get a set of alternatives instead of a manual solution; in the second stage, the investment, the transportation risks and transportation duration are selected as the decision objectives, the MCDM model is established for the three objectives and the weighted TOPSIS method is adopted to rank alternatives and select the optimal scheme. In addition, with the help of web crawler technology, the public transportation network information about the BMT is collected from the public WebGIS platform and allows the complex data calculation to be realized.

The structure and procedures of this method are shown in Figure 2. The specific information about the TDM and the technical details of the web crawler technology are described in detail in Section 3.2–3.4, respectively.

3.2. Collection Data by Web Crawler Technology

As in Figure 2, the transportation network and its information are the basis of the TDM, which needs to collect relevant information around the construction site, such as the material supplies, transport rates, the geological and meteorological conditions and the highway, railway and waterway transportation conditions. Currently, such a large amount of complex data mainly relies on manual investigation and measurement, which suffers from low efficiency and a long cycle and is labor-intensive. Usually, the information acquisition process would take several months. Hence, it is necessary to improve the data collection efficiency and reduce the labor cost. This paper adopts web crawler technology to make computers assist designers in completing the data collection and calculation. The technical process is shown in Figure 3.

According to collection methods, the data of the BMT can be grouped into two categories: (1) information included in the public WebGIS platform can be automatically obtained using the Maps API, such as the transportation distance, the highway distribution information and the driving time of vehicles; (2) information that is not included in the public WebGIS platform needs to be collected manually, such as the transport rates, the transport vehicles and the geological and meteorological conditions. Hence, the information about the BMT of hydropower projects should be collected by combining the manual investigation and the web crawler technology.

The detailed technical process of web crawler technology is shown in Figure 3. The client and the local server are written in the Python language. The client is mainly used to determine the information to be collected and accept the relevant information from the local server. The local server is mainly used to edit and send the request parameters and accept and parse the returned parameters. Usually, the back parameters from the Maps API service are in the JSON format and require data analysis to convert them into a dictionary format allowed by Python, and then the valid data are extracted, including the transportation distance, the highway distribution and the driving time of vehicles.

After completing the information collection, the transportation network can be constructed and transformed into the path–node topology relationship networks diagram using the relevant graph theory. To facilitate the construction of the mathematical models and the subsequent calculations, in this paper, the parameters defined as follows:

• E: the transportation network under study, E= (B, Y, V);
• B: the set of all city nodes in the transportation network;
• Y: the set of all path sections in the transportation network;
• V: the set of the bulk materials, including cement, fly ash and steel;
• i,j: the city nodes in the transportation network, i.e., i,j∈B;
• (i,j): the path section in the transportation network, (i,j)∈Y;
• k: one of the bulk materials, i.e., k∈V;
• q^k: total demand of the bulk material k for the project construction.

3.3. Multi-Source Path Optimization

The MSPO method is adopted to obtain alternatives from the transportation network. Its input parameters are transportation network information, including transportation starting points, destination and city nodes. Its output results are alternatives.

3.3.1. Establishing the Objective Function of Path Optimization

As stated above, the investment, the transportation risk and the transportation duration are all vital decision properties for the BMT of hydropower projects. Through a comparative analysis of the characteristics of different properties, the investment is selected as the optimization objective. However, in addition to the regular transportation costs, the calculation of the investment also needs to consider both the infrastructure investment and transshipment expenses: (1) the infrastructure investment refers to the investment spent on the new construction, reconstruction and expansion of roads, bridges, tunnels and transfer stations that do not meet the transportation conditions for the BMT, which can be obtained through estimation; (2) transshipment expenses refers to the investment spent on changing the transportation modes at a designated transfer station, where the transportation modes are changed from the railway to the highway or from the waterway to land transportation. Because the supplies of the bulk materials are usually far from the construction site, it is not easy to transport materials directly by a single transportation mode. Therefore, it is necessary to change the transportation modes at the transfer station. The transshipment expenses can be estimated through surveys and experience.

In summary, in addition to the regular transportation costs, which are mainly determined by the transportation distance and the transportation volume, the BMT must also consider the infrastructure investment and the transshipment expenses; i.e., the investment consists of the transportation costs, the infrastructure investment and the transshipment expenses. The calculation formula of the investment is as follows:

$${\tilde{c}}_{ij}^k=c_{ij}^{1k}+c_{ij}^{2k}+c_{ij}^{3k}$$

(1)

where ${\tilde{c}}_{ij}^k$ is the investment to transport bulk material k on path section (i,j); $c_{ij}^{1k}$, $c_{ij}^{2k}$ and $c_{ij}^{3k}$ are the transportation costs, the infrastructure investment and the transshipment expenses to transport bulk material k on path section (i,j), respectively.

3.3.2. Obtaining Alternatives Based on the Floyd Algorithm

As mentioned above, each bulk material has multiple suppliers, and each supplier has multiple optional paths to the construction site. That is to say that the transportation network understudy may have multiple transportation starting points and paths that can reach the construction site. The Floyd–Warshall algorithm is an algorithm that uses the idea of dynamic programming to compare all possible paths through the graph between each pair of vertices from weighted graphs with positive or negative edge weights [46,47]. The algorithm has high reliability and can efficiently calculate the optimal solution between two nodes. Hence, the Floyd algorithm is adopted to find the optimal transportation route from the starting point to the construction site for each supplier, with a minimum investment.

Assume that N suppliers supply the bulk material k that meets the project’s demand. That is to say that there are N transportation starting points and one transportation endpoint in the transportation network. Then, calculate the investment of each path section (i,j) and use the Floyd algorithm to find the optimal transportation route for each transportation starting point, with the goal of the minimum investment. The result is that N starting points can obtain N transportation routes that all are optimal, which are regarded as a set of alternatives. Equation (2) represents the set of alternatives for the bulk material k:

$$F^k=\left\{f_n^k|n\in N\right\}$$

(2)

where F^k is the set of alternatives corresponding to the bulk material k, N is the number of alternatives, n represents the n-th transportation scheme in the set of alternatives, and f_n^k is the n-th transportation scheme in the set of alternatives for the bulk material k.

3.4. Multi-Criteria Decision-Making

As in Figure 2, after obtaining the set of alternatives, the MCDM method ranks alternatives and selects the optimal scheme. Details about the method are described below. Its input parameters are information on alternatives. Its output result is the optimal scheme.

3.4.1. Establishing Decision Objectives Functions

In the second stage, the minimum investment, the shortest transportation duration and the least transportation risk are selected as the decision objectives. The hierarchical structure of the decision objectives for the MCDM is shown in Figure 4.

1.

The minimum investment

The investment has been calculated using the Floyd algorithm in the first stage. The formula for calculating the total investment of each alternative is as follows:

$$C_n^k={\displaystyle \sum _{(i,j)\in Y}{\tilde{c}}_{ij}^k·y_{ij}^n}$$

(3)

where $C_n^k$ is the total investment to transport the bulk material k using the n-th alternative, and y_ijⁿ is a decision variable—if the n-th alternative passes the path section (i,j), its value would be 1, otherwise 0, and n∈N.

2.

The shortest transportation duration

The transportation duration refers to the time consumed in the entire transportation process from the material suppliers to the construction site, which is composed of the transportation time and the transshipment time: (1) the transportation time refers to the time consumed by transporting bulk materials on the road, which is mainly determined by the distance and the transport vehicles. The higher the average speed of the vehicle and the shorter the transportation distance, the shorter the transportation time; (2) the transshipment time refers to the time consumed by changing the transportation modes at the transfer station because the bulk materials will be placed at the designated transfer station for a period of time to change the transportation modes. The transshipment time can be obtained through investigation. The calculation formula of the transportation duration is as follows:

$$T_n^k={\overline{t}}_n^k+{\displaystyle \sum _{(i,j)\in Y}{t}_{ij}^k·y_{ij}^n}$$

(4)

where T_n^k is the total transportation duration to transport the bulk material k using the n-th alternative, and ${\overline{t}}_n^k$ is the transshipment time to transport the bulk material k using the n-th alternative. If there is no change in the transportation modes during transport, then ${\overline{t}}_n^k$ = 0; t^k_ij is the transportation time of the bulk material k on path sections (i,j).

3.

The least transportation risk

By analyzing the related literature and the industry norms about the BMT, it can be found that the risk factors of transporting the bulk materials are mainly considered from geological disasters and meteorological disasters. Geological disasters include landslides, debris flows, land subsidence, etc. Meteorological disasters include rainstorms, floods, etc. Among existing literature, the commonly applied risk assessment metrics include the occurrence probability and the risk losses. However, because each transportation route could contain multiple potential risk sources and risk factors in the transportation network, it is of no practical significance to estimate the losses of each risk. Hence, in this paper, the occurrence probability is adopted to describe the transportation risk, and P is used to indicate the occurrence probability. The geological disaster can be estimated through the survey and expert experience. A meteorological disaster can be inferred based on the meteorological data from past years. So, the transportation risk of each alternative can be estimated by inviting experts. The occurrence probability, including the probabilities of geological disaster and the meteorological disaster, is evaluated using the method of expert scoring. The risk evaluation grades are divided into five grades: no risk, low risk, average risk, high risk and higher risk, corresponding to 1, 3, 5, 7 and 9, respectively. The median values of two adjacent judgments of the risk grades are 2, 4, 6 and 8, respectively.

3.4.2. Decision-Making Based on the TOPSIS

TOPSIS is a classic method for solving multi-criteria decision-making problems, which is a method of sorting according to the closeness of a limited number of evaluation objects to the idealized target [48,49]. This paper uses the TOPSIS method to rank the alternatives based on the MCDM model. The specific solution steps of this method are as follows:

Step 1: Determining the judgment matrix. For each alternative, the attribute values of the investment, the transportation duration and the transportation risk are calculated, according to Section 3.4.1 of this paper, respectively. The decision matrix A is as follows:

$$A=\left(\begin{array}{ccc}\begin{array}{c}{C}_{11}^k\\ ⋮\end{array}& \begin{array}{c}{T}_{12}^k\\ ⋮\end{array}& \begin{array}{c}{P}_{13}^k\\ ⋮\end{array}\\ \begin{array}{c}{C}_{n1}^k\\ ⋮\end{array}& \begin{array}{c}{T}_{n2}^k\\ ⋮\end{array}& \begin{array}{c}{P}_{n3}^k\\ ⋮\end{array}\\ C_{N1}^k& T_{N2}^k& P_{N3}^k\end{array}\right)$$

(5)

where $C_{n1}^k$, $T_{n2}^k$ and $P_{n3}^k$ represent the attribute values of the investment, the transportation duration and the transportation risk using the n-th alternative to transport the bulk material k, respectively.

Step 2: Standardize the decision objectives. The investment, the transportation duration and the transportation risk selected in this paper are all cost-based (reverse direction) indicators. The smaller the attribute values, the better the transport scheme. Because the TOPSIS ranks the alternatives according to the attribute values, converting the cost-based indicators into benefit-based indicators (positive direction) is necessary. The “inverse reciprocal transformation method” is used to change the decision indicators into the benefit-based indicators, which is the reciprocal of the attribute value of each decision indicator. Then, “the mathematical vector normalization method” is used to standardize the decision objectives to eliminate the dimensional effects of different indicators [50,51]. We record the processed decision matrix as Z.

$$Z=\left(\begin{array}{ccc}\begin{array}{c}{z}_{11}^k\\ ⋮\end{array}& \begin{array}{c}{z}_{12}^k\\ ⋮\end{array}& \begin{array}{c}{z}_{13}^k\\ ⋮\end{array}\\ \begin{array}{c}{z}_{n1}^k\\ ⋮\end{array}& \begin{array}{c}{z}_{n2}^k\\ ⋮\end{array}& \begin{array}{c}{z}_{n3}^k\\ ⋮\end{array}\\ z_{N1}^k& z_{N2}^k& z_{N3}^k\end{array}\right)$$

(6)

Step 3: Determining the weight of decision objectives. From the perspective of decision-makers, under the conditions of meeting the engineering requirements and regulations, it is always expected that the transportation scheme of the bulk materials can achieve the minimum investment, the shortest transportation duration and the least transportation risk at the same time. However, there are contradictions and incommensurability in actual engineering among the decision objectives. Generally, all decision objectives cannot be achieved at the same time. Hence, the weighting method is adopted to balance the decision objectives. Considering that the MCDM model contains multiple qualitative indicators, and because of the complex conditions of the BMT, many indicators are difficult to quantify. This paper adopts the AHP method to determine the weights of the decision objectives. The weights of the objectives are recorded as $w=\left\{w_1,w_2,w_3\right\}$.

Step 4: Ranking the alternatives. Based on the decision matrix Z, the ideal solution $Z^{+}$ and the negative-ideal solution $Z^{-}$ are calculated using Equations (7) and (8), respectively. Then, we calculate the weighted distance of each alternative in the set of alternatives F^k according to Equations (9) and (10). Finally, we calculate the relative closeness to the ideal solution using Equation (11) and rank the alternatives based on the score.

$$Z^{+}=\left(Z_1^{+},Z_2^{+},\cdots ,Z_m^{+}\right)=\left(\max \left\{z_{11}^k,z_{21}^k,\cdots ,z_{N1}^k\right\},\max \left\{z_{12}^k,z_{22}^k,\cdots ,z_{N2}^k\right\},\max \left\{z_{13}^k,z_{23}^k,\cdots ,z_{N3}^k\right\}\right)$$

(7)

$$Z^{-}=\left(Z_1^{-},Z_2^{-},\cdots ,Z_m^{-}\right)=\left(\min \left\{z_{11}^k,z_{21}^k,\cdots ,z_{N1}^k\right\},\min \left\{z_{12}^k,z_{22}^k,\cdots ,z_{N2}^k\right\},\min \left\{z_{13}^k,z_{23}^k,\cdots ,z_{N3}^k\right\}\right)$$

(8)

$$D_n^{+}=\sqrt{{\displaystyle \sum _{j=1}^m{w}_j{\left(Z_j^{+}-z_{ij}^k\right)}^2}}$$

(9)

$$D_n^{-}=\sqrt{{\displaystyle \sum _{j=1}^m{w}_j{\left(Z_j^{-}-z_{ij}^k\right)}^2}}$$

(10)

$$S_n=\frac{D_n^{-}}{D_n^{+}+D_n^{-}}$$

(11)

where $D_n^{+}$ is the weighted distance of the n-th alternative from the ideal solution, $D_n^{-}$ is the weighted distance of the n-th alternative from the negative-ideal solution, and $S_n$ is the relative closeness of the n-th alternative, $0\le S_n\le 1$.

In summary, the weighted TOPSIS method processes the three objectives according to steps 1 to 4 and obtains the score $S_n$ of each alternative. Then, according to the value of $S_n$, the alternatives are ranked, and the optimal BMT scheme of hydropower projects can be given.

4. Case Study

A case study is presented below, which demonstrates the theoretical framework and methodology proposed in this paper.

4.1. Case Profile

The cement transportation of a large-scale hydropower station project in China is set as the case background. The transportation network around the construction site is shown in Figure 5. The basic data to be provided are as follows:

(1)

Material volume: This project requires a total of 901,000 t (tonnes) of cement.

(2)

Cement manufacturers: There were three cement manufacturers that met the project needs and were selected as the project suppliers in terms of the production capacity and material quality conditions, which are on node 2, node 10 and node 18.

(3)

Base price: The transport rate of the highway is 1 yuan/t · km, and the loading and unloading fee is 6 yuan/t. The railway freight base price is 0.103 yuan/t · km, the arrival base price is 18.6 yuan/t, and the loading and unloading fee is 15.1 yuan/t. The waterway freight base prices are 20~25 yuan/t, and the loading and unloading fee is 18 yuan/t. The transit cost from the railway to the highway is 25 yuan/t, and the transit cost from the waterway to the highway is 20 yuan/t.

In addition, some other data that have an impact on the BMT planning need to be included. The complementary data include node 5, node 7, node 10, node 17 and node 18, all of which have the potential to perform as the transit station where the transportation modes can be changed; through investigation, the waterway of the path section (10, 17) and (17, 13) fails to transport in the dry season due to the channel depth.

4.2. Definition of the Transportation Network

As in Figure 2, when building the transportation network of the cement, it is necessary to collect the relevant information from the public WebGIS platform using the web crawler technology. Considering that the project is located in China, Baidu Maps was selected as the WebGIS platform to retrieve the public transportation information. The Python programming language was used to collect and calculate relevant data automatically. The API interface obtained the transportation distance, the transportation modes and driving time consumption from Baidu Maps. Then, the Python program processed and calculated the transportation costs. The infrastructure investment was obtained through investigation.

Table 2. The information of each path in the transportation network of the cement.

Path (i,j)	Modes	Distance (km)	Infrastructure Investment *	Transportation Costs *	Path (i,j)	Modes	Distance (km)	Infrastructure Investment *	Transportation Costs *
(1,2)	Highway	34.5	197.38	3108.45	(15,18)	Highway	145.9	0	13,145.59
(1,3)	Highway	222.6	1273.53	20,056.26	(18,17)	Highway	197.1	0	17,758.71
(1,6)	Highway	205.7	0	18,533.57	(17,14)	Highway	107.5	0	9685.75
(2,4)	Highway	48.7	278.62	4387.87	(14,13)	Highway	75.4	431.37	6793.54
(4,5)	Highway	92.1	0	8298.21	(13,site)	Highway	113.6	649.92	10,235.36
(3,8)	Highway	268.4	1535.56	24,182.84	(2,7)	Railway	420.0	0	3897.73
(6,7)	Highway	81.5	0	7343.15	(7,10)	Railway	192.0	1537.85	1781.82
(7,10)	Highway	184.9	0	16,659.49	(10,15)	Railway	198.0	0	1837.50
(8,11)	Highway	90.3	516.62	8136.03	(15,18)	Railway	174.5	0	1618.95
(11,12)	Highway	199.2	1139.65	17,947.92	(18,17)	Railway	183.0	1465.76	1698.29
(10,16)	Highway	112.5	643.63	10,136.25	(5,9)	Waterway	323.7	0	1937.15
(12,13)	Highway	63.6	0	5730.36	(9,10)	Waterway	285.4	0	2207.45
(16,17)	Highway	81.3	465.13	7325.13	(10,17)	Waterway	342.9	1373.25	1847.05
(10,15)	Highway	227.3	0	20,479.73	(17,13)	Waterway	213.5	1710.05	2252.50

* The unit of the infrastructure investment and the transportation costs is 10 thousand Yuan.

shows the transportation distance, the transportation modes, the infrastructure investment and the transportation costs of each path (i,j) in the transportation network. With the help of web crawler technology, the information collection process only took about one week. Compared with the traditional method that takes several months, web crawler technology dramatically improves work efficiency.

The transportation network can be transformed into a path–node topology relationship networks diagram with prepared data. The key nodes in the transportation network have been identified in Figure 5.

4.3. Application of the TDM

According to the theoretical framework of the TDM, the relationship between the MSPO model and the MCDM model is that the results of path optimization are the basis of multi-criteria decision-making. The detailed steps are as follows.

4.3.1. Results of the MSPO

According to Figure 2, in the first stage, the MSPO model is established with the investment as the optimization objective, and the Floyd algorithm is adopted to solve the optimization model and obtain a set of alternatives.

Firstly, the investment of each path (i,j) is calculated using Formula (1), which is used as the edge weight of the path–node topology relationship network. Then, the Floyd algorithm is used to optimize the transportation network. According to the case profile, there are three cement suppliers in the transportation network, and every supplier corresponds to a transportation scheme that can transport the cement to the construction site with their minimum investment. Hence, the results of the MSPO include three transportation schemes for the three cement suppliers. The corresponding transportation routes of each scheme are as follows:

• Route 1: 1~3~8~11~12~13~ the construction site;
• Route 2: 10~17~14~13~ the construction site;
• Route 3: 18~17~14~13~ the construction site.

The calculation process of the MSPO took 17.6 s in total. However, the process of manually drafting alternatives usually takes several days. Obviously, the MSPO method can significantly reduce the time needed to obtain alternatives. The detailed information of each scheme is shown in

Table 3. The detailed information of each scheme.

Schemes	Routes	Transportation Modes	Transfer Node	Transportation Costs (10 Thousand Yuan)	Infrastructure Investment (10 Thousand Yuan)	Transshipment Expenses (10 Thousand Yuan)	Total Investment (10 Thousand Yuan)
f1	Route 1	Highway	/	66,288.77	5115.28	0	71,404.05
f2	Route 2	Waterway to Highway	17	28,561.70	2454.55	1802.00	37,503.45
f3	Route 3	Railway to Highway	17	28,412.94	2547.06	2252.50	38,690.58

, including the transportation modes, the transfer node, the transportation costs, the infrastructure investment, the transshipment expenses and the total investment. The infrastructure investment and the transshipment expenses are obtained through investigation.

4.3.2. Results of the MCDM

As demonstrated before, the three schemes obtained in the first stage form a set of alternatives. According to Figure 2, the MCDM model is established in the second stage with the minimum investment, the shortest transportation duration and the lowest transportation risk as the decision objectives, and the weighted TOPSIS method is adopted to rank the alternatives and select the optimal scheme [33,52]. The detailed steps are as follows.

Step 1: Calculate the attribute values of the decision objectives for each alternative; the investment has been calculated in the first stage, the transportation duration can be calculated by the formula (4), and the transportation risk can be estimated using the method of expert scoring. The attribute values of the three decision objectives for each alternative are listed in

Table 4. The attribute values of the three decision objectives.

Alternatives	Investment C (10 Thousand Yuan)	Transportation Duration T (day)	Transportation Risk P
f1	71,404.05	3	5
f2	37,503.45	8	4
f3	38,690.58	7	3

.

Step 2: The decision matrix A can be established according to the corresponding attribute values of the decision objectives in Table 4, as shown in Equation (12). Then, according to step 2 in Section 3.4.2, Equations (13) and (14) are used to process the decision matrix A to eliminate the dimensional effects of indicators. The results are shown in Equation (15).

$$A=\left(\begin{array}{ccc}{a}_{11}& a_{12}& a_{13}\\ a_{21}& a_{22}& a_{23}\\ a_{31}& a_{32}& a_{33}\end{array}\right)=\left(\begin{array}{ccc}71404.05& 3& 5\\ 37503.45& 8& 4\\ 38690.58& 7& 3\end{array}\right)$$

(12)

$${a^{\prime }}_{ij}=\frac{1}{a_{ij}},i=1,2,3;j=1,2,3.$$

(13)

$$z_{ij}=\frac{{a^{\prime }}_{ij}}{\sqrt{{\displaystyle \sum _j{\left({a^{\prime }}_{ij}\right)}^2}}},i=1,2,3;j=1,2,3.$$

(14)

$$\mathit{Z}=\left(\begin{array}{lll}0.3529\hfill & 0.8690\hfill & 0.4327\hfill \\ 0.6718\hfill & 0.3259\hfill & 0.5409\hfill \\ 0.6512\hfill & 0.3724\hfill & 0.7212\hfill \end{array}\right)$$

(15)

Step 3: Determine the ideal and negative-ideal solutions through Equations (7) and (8). The ideal solution is shown in Equation (16). The negative-ideal solution is shown in Equation (17).

$${\mathit{Z}}^{+}=\left\{0.6718,0.8690,0.7212\right\}$$

(16)

$${\mathit{Z}}^{-}=\left\{0.3529,0.3259,0.4327\right\}$$

(17)

Step 4: The AHP is adapted to determine the investment weights, the transportation duration and the transportation risk, respectively. The weights of the three decision objectives are denoted as w = {0.5390, 0.1638, 0.2972}. The detailed process of this step can be found in Appendix A.

Step 5: The distance of each alternative from $D_n^{+}$ and $D_n^{-}$ can be currently calculated using Equations (9) and (10). Finally, TOPSIS solves the similarities to an ideal solution by Equation (11). In order to clarify what has been mentioned, an example is presented as follows:

$$D_1^{+}=\sqrt{0.5390\times {\left(0.3529-0.6718\right)}^2+0.1638\times {\left(0.8690-0.8690\right)}^2+0.2972\times {\left(0.4327-0.7212\right)}^2}=0.2821$$

(18)

$$D_1^{-}=\sqrt{0.5390\times {\left(0.3529-0.3529\right)}^2+0.1638\times {\left(0.8690-0.3259\right)}^2+0.2972\times {\left(0.4327-0.4327\right)}^2}=0.2198$$

(19)

As a result,

$$S_1=\frac{D_1^{-}}{D_1^{+}+D_1^{-}}=\frac{0.2198}{0.2821+0.2198}=0.4379$$

(20)

Similar calculations are done for the other alternatives, and the results of TOPSIS analyses are summarized in

Table 5. Closeness coefficients and ranking of alternatives.

Alternatives	$D_n^{+}$	$D_n^{-}$	$S_n$	Rank
f1	0.2821	0.2198	0.4379	3
f2	0.2408	0.2415	0.5007	2
f3	0.2015	0.2703	0.5729	1

.

Finally, according to the closeness coefficients of S_n, the results of ranking the alternatives are as follows: (i) f₃ is the optimal method among the alternatives; (ii) f₃ precedes f₂ precedes f₁. Hence, alternative f₃ is recommended as the optimal scheme for cement transportation. The details of the alternative f₃ include the following: (i) the route is 18~17~14~13~ the construction site, and the cement’s supplier is located at node 18; (ii) among the alternatives, the total investment is relatively small, the transportation duration is relatively long, and the transportation risk is the lowest; (iii) the transportation mode is from the railway to the highway, and the transfer station is set at node 17.

5. Discussion

The engineering materials transportation problem in hydropower projects is a new research field that has not attracted widespread attention from scholars. This paper proposes a systematic and reliable TDM framework which can effectively solve the material transportation problem. This method provides a theoretical reference for researchers in the field and helps to understand the nature of the BMT problem in hydropower projects. In addition, considering the difficulty of transportation network data collection, this paper adopts web crawling technology to collect data quickly and efficiently based on the public WebGIS platform. Combined with the case study part, the calculation method and process of the TDM method are explained, and the application method of the web crawling technology is given. By analyzing the case-solving process and results, constructive information is discussed to support the findings of the study.

(1) Case result analysis. According to the path optimization results (Table 3), the total investment of scheme f₂ is the lowest, at about 37,503.45 (10 thousand Yuan). On the contrary, the total investment of scheme f₁ is the highest, at about 71,404.05 (10 thousand Yuan). There are two reasons for this result: (i) adopting the highway transportation, scheme f₁ involves a great deal of reconstruction and expansion of bridges, tunnels and other infrastructure, which results in high infrastructure investment. Scheme f₂ adopts the waterway to highway transportation mode, with less reconstruction and expansion; (ii) railway and waterway transportation rates are lower than the highway, and transportation costs are less. Hence, the path optimization results are realistic, which proves the reliability and applicability of the MSPO model established in this paper. According to the MCDM results, scheme f₃ is the optimal approach among the alternatives. According to the data in Table 4, it can be seen that (i) the attribute values of the investment and the transportation duration of scheme f₃ are in the middle of scheme f₁ and scheme f₂; (ii) the transportation risk of scheme f₃ is minimal; (iii) comparing f₂ and f₃, the attribute values of the decision objectives have little difference, so it is difficult for the single-criteria decision-making method to distinguish the differences between the two schemes and to reflect the superiority of f₃. The MCDM method can comprehensively consider the investment, the transportation duration and the transportation risk to compare the two schemes. Therefore, it is necessary and reliable to adopt the MCDM method to select the optimal scheme from the alternatives.

(2) Method comparison analysis. Currently, there is little literature that studies the BMT problem systematically. In engineering design, BMT planning relies heavily on the experience of experts, which lacks objectivity. Combined with the case study, the necessity and applicability of the TDM method can be demonstrated from four aspects: (i) the BMT planning of hydropower projects is not just a path optimization problem or a decision-making problem. It is necessary to simultaneously analyze the transportation network and the influencing factors of bulk materials transportation. That is to say, the planning of the BMT of hydropower projects involves both the path optimization problem and the decision-making problem. Meanwhile, the TDM can achieve a one-stop solution for the BMT problem of hydropower projects; (ii) the investment and the transportation duration of scheme f3 all are not optimal among the three alternatives, but it is the overall optimal approach, which illustrates the necessity of multi-criteria decision-making; (iii) the MSPO method can characterize the reconstructed, expansive and newly-built facilities through the edge weights, which makes the path optimization model conform to the actual situation of the BMT for hydropower projects; (iv) this study adopts the MSPO method to solve the transportation planning and build a mathematical model, which makes up for the gap of existing research—i.e., research is lacking on obtaining the alternatives; and with the help of Python, the MSPO model can replace the process of drafting alternatives manually to obtain the alternatives efficiently. In the case study, the calculation process of the MSPO took only 17.6 s. However, the process of manually drafting one alternative usually takes several days. Therefore, it is reasonable and reliable to use the TDM method to solve the BMT problem.

(3) Advantages of adopting web crawler technology. (i) As stated above, the transportation network of cement has a complex spatial distribution and includes a large amount of information. Using the traditional method to collect and summarize the relevant information manually would take several months for multiple suppliers. However, in the case study, with the help of web crawler technology, the process only takes about one week. Obviously, adopting web crawler technology can improve work efficiency by dozens of times. (ii) As in Figure 4, 18 nodes, 25 sections and 3 starting points in the transportation network can be combined into dozens of cement transportation schemes. This is a large-scale planning problem, and it is impractical to plan and compare all these possible transportation schemes manually. The application of web crawler technology makes it possible to solve the large-scale planning problem. (iii) Using web crawler technology to collect and summarize the relevant information instead of performing this manually can avoid errors caused by measurement and empirical estimation and improve the data accuracy.

6. Conclusions

This paper develops a theoretical framework for solving the bulk materials transportation problem in hydropower projects. After analyzing the BMT problem from a global perspective, we show that BMT planning involves both the path optimization problem from the perspective of the transportation planning and the multi-criteria decision problem from the perspective of the scheme selection. Hence, this study proposes a TDM that solves the path optimization problem and the scheme selection problem in stages. The specific method of the TDM is as follows: an MSPO model is established in the first stage to solve the transportation planning problem and obtain alternatives instead of manual preparation; in the second stage, the MCDM method is adopted to select the optimal scheme. The case study results show that the TDM can plan multiple alternatives of the BMT and can select the optimal scheme among alternatives. The discussion proves the practicality and reliability of this method. In addition, web crawler technology is used to automatically obtain the transportation network information from the public WebGIS platform and calculate related parameters. Contributions of the study include the following:

(1)

This paper innovatively proposes a two-stage decision-making method: the TDM, which solves the transportation planning problem and the decision-making problem in stages and is effective in the engineering transportation problem, which helps to narrow the current research gap of lacking systematic and practical research specifically for the BMT problem of hydropower projects. The TDM realizes the systematic decision and global optimization of the entire scheme planning process from transportation planning to scheme selection.

(2)

Combination of the MSPO method and MCDM method. Within the TDM, the MSPO model replaces the process of manually planning the potential draft alternatives from the transportation network under study and considers the infrastructure investment, which makes up for the lack of study of engineering transportation problems, and the MCDM model can characterize the key features of the BMT for hydropower projects and conforms to the preferences of decision-makers.

(3)

Application of web crawler technology. This technology realizes the automatic acquisition of the transportation network information from the public WebGIS platform and the calculation of the related data, improving the efficiency of information collection and processing and making it possible to compare large-scale alternatives.

The TDM based on WebGIS can play a positive role in materials transportation in the engineering field and provide a reference for combining the traditional industry and intelligence.