A dynamic market price of coal real-time quotation and bargaining transaction method

Abstract

The application discloses a coal real-time pricing and bargaining transaction method for dynamic market quotations, and comprises the following steps: S1, collecting multi-source heterogeneous market data in real time, wherein the multi-source heterogeneous market data at least comprises macro market index data, micro coal index data, dynamic logistics network data and meteorological environmental factor data; through a constructed data cleaning and feature extraction model, the multi-source heterogeneous market data is mapped into a standardized multi-dimensional state vector; S2, constructing a dynamic benchmark price evaluation model based on a time-space attenuation mechanism; and inputting the multi-dimensional state vector into the dynamic benchmark price evaluation model.

Description

Technical Field

[0001] This invention relates to the fields of digital trading of bulk commodities, distributed ledger technology, and deep reinforcement learning in artificial intelligence, and particularly to a method for real-time pricing and negotiated trading of coal based on dynamic market conditions. Background Technology

[0002] Coal, as my country's core energy source and a major industrial raw material, has traditionally relied on periodic static guidance prices for its trading, accompanied by multiple rounds of offline negotiations via telephone and instant messaging software, often described as "back-to-back." However, with the evolution of modern supply chains and the dramatic fluctuations in the spot market, this traditional model has revealed serious limitations and inefficiencies.

[0003] Coal prices are often affected by macroeconomic indicators such as futures market activity, the Bohai Rim thermal coal index, and daily coal consumption at coastal power plants. At the same time, they are constrained by micro-logistic conditions such as the number of ships anchored at ports and railway congestion. Static guidance prices cannot capture these intraday fluctuations in real time.

[0004] Furthermore, coal is a bulk commodity highly susceptible to environmental impacts. During long-distance sea and rail transport or open-air storage, exposure to heavy rain can cause a surge in total moisture content and a nominal decrease in calorific value; exposure to strong winds and direct sunlight can lead to weathering, weight loss, and an absolute loss of calorific value. The existing pricing mechanism completely disrupts the causal chain from weather conditions to in-transit losses and then to real-time valuation, resulting in frequent quality disputes at the time of delivery.

[0005] Therefore, when faced with a massive number of inquiries, human sales representatives often cannot accurately calculate the company's current inventory pressure and capital occupation costs. The pace and degree of concessions made during the negotiation process rely entirely on experience, which can easily lead to a stalemate or cause profits to be excessively tilted towards the buyer or seller. Summary of the Invention

[0006] The main objective of this invention is to provide a method for real-time coal pricing and negotiation based on dynamic market conditions, which can effectively solve the problems in the background art.

[0007] To achieve the above objectives, the present invention adopts the following technical solution:

[0008] A method for real-time coal pricing and negotiated transactions based on dynamic market conditions, the method comprising the following steps:

[0009] S1: Real-time collection of multi-source heterogeneous market data, which includes at least macro market index data, micro coal index data, dynamic logistics network data, and meteorological environmental factor data; through the constructed data cleaning and feature extraction model, the above-mentioned multi-source heterogeneous market data is mapped into a standardized multi-dimensional state vector;

[0010] S2: Construct a dynamic benchmark price assessment model based on a spatiotemporal decay mechanism; input the multidimensional state vector into the dynamic benchmark price assessment model, and combine the carbon emission cost accounting mechanism and the coal quality decay penalty function in transit to calculate the time-sensitive initial dynamic benchmark price range for coal buying and selling at the current time node;

[0011] S3: Initialize the multi-agent bargaining game model under asymmetric information; based on the historical transaction profiles of buyers and sellers, capital turnover rate and inventory pressure indicators, set a dynamic reserve price floor and initial negotiation step size for the buyer's intelligent agent and the seller's intelligent agent, respectively.

[0012] S4: Automatically execute multi-round reinforcement learning negotiation strategy; within the initial dynamic benchmark price range, the intelligent agents of the buyer and seller interact with each other based on Markov decision process; in each round of interaction, the intelligent agent evaluates the other party's pricing intention and market urgency through its built-in nonlinear concession function, and outputs the preferred price for the next round until the price difference between the two parties is less than the preset convergence threshold or the maximum number of negotiation rounds is reached.

[0013] S5: Transaction matching and automatic generation of smart contracts; when the price difference between the buyer and seller is less than the convergence threshold, the final transaction price and the corresponding logistics performance conditions are extracted, the transaction metadata is put on the blockchain, and an immutable coal transaction smart contract is automatically generated using the blockchain.

[0014] Preferably, the multi-source heterogeneous market data in step S1 specifically includes:

[0015] The macro market index data includes: the Bohai Rim thermal coal price index, real-time order book data of the Zhengzhou Commodity Exchange thermal coal futures main contract, and daily coal consumption and available days of key power plants nationwide.

[0016] The microscopic coal index data include: calorific value, total moisture content, ash content, volatile matter content, and total sulfur content;

[0017] The dynamic logistics network data includes: railway dispatch plans and congestion index, real-time number of anchored vessels at major ports, inventory at ports, and real-time sea / road freight rates.

[0018] The meteorological environmental data includes: rainfall warnings along the route from the production area to the destination, the probability of extreme wind and snow weather, and the forecast of port closure duration.

[0019] Preferably, the dynamic benchmark price evaluation model based on the spatiotemporal decay mechanism in step S2 calculates the basic benchmark price. The improvement process is as follows:

[0020] Existing benchmark prices rely solely on simple index weighting. This invention introduces carbon tax adjustment and end-to-end cost mapping, the calculation formula of which is as follows: ;in, For the first Real-time values ​​of macro market indices. For the corresponding weights; This refers to the quality premium or discount calculated based on micro-level coal index data; The comprehensive door-to-door logistics cost is calculated in real time based on dynamic logistics network data. This refers to the implicit carbon emission cost calculated based on the total sulfur content of coal and the expected combustion conversion rate. This is the penalty function for the quality degradation of coal during transit.

[0021] Preferably, existing technologies neglect the quality loss of coal due to weather during long-distance transportation and open-air storage. This invention constructs a penalty function for the quality decay of coal during transit. The specific formula is as follows: ,in, The price of bare coal before deducting losses; For the estimated total transportation and storage time; for Rainfall intensity factor along the route and at the port at all times; for Wind intensity factor at any given time; and These are the sensitivity coefficients for the decrease in heat generation due to water penetration and the sensitivity coefficients for weathering loss, respectively. This represents the absolute loss of coal calorific value predicted based on meteorological and environmental factor data. Economic penalty multiplier for heat loss

[0022] Preferably, in step S3, the buyer's intelligent agent dynamically reserves a floor price. Dynamic price floor reserved with seller's smart agent The setting not only relies on static costs but also dynamically incorporates real-time inventory pressure: when the buyer's inventory is running low, the maximum price they are willing to pay will automatically increase, as represented by the formula: When a seller has excess inventory, the lowest price they are willing to accept (i.e., the seller's reserve price) will automatically decrease, as represented by the formula: ;in, The urgency coefficient is obtained by nonlinearly mapping the real-time available inventory days and cash flow health of an enterprise through a logistic regression model.

[0023] Preferably, in step S4, traditional negotiation methods use fixed-step concessions, which can easily lead to a stalemate or excessively rapid compromises. In this invention, the intelligent agents for both buyers and sellers use a non-linear concession function based on time pressure and remaining pricing space to generate the next round of optimal pricing. In the first round of negotiations, the current offer from one party's intelligent agent... The calculation logic is as follows:

[0024]

[0025] in, For the first The dynamic yield step size of the wheel is determined by the intelligent agent through reinforcement learning. Network output aims to maximize long-term expected returns.

[0026] Preferably, to ensure that the negotiation model can converge quickly and avoid getting stuck in an infinite loop, the dynamic concession step size is... In the reinforcement learning action space, there are explicit mathematical constraints, and the baseline constraint formula is:

[0027]

[0028] in, Reserve a price floor for the dynamics of this intelligent agent; The maximum number of negotiation rounds set for the system; To accommodate the convexity coefficient, when This strategy manifests as an initial hardline stance followed by significant concessions later on. This strategy manifests as initial tentative concessions followed by a firm adherence to bottom lines. This is a trading sentiment feedback factor calculated based on the concessions made by the opponent in the previous round.

[0029] Preferably, the transaction matching and smart contract automatic generation step S5 specifically includes: when hour, To preset the convergence threshold, the average of the final bids from both parties is taken as the final transaction price. The system calls the oracle node in the blockchain network to encapsulate the final transaction price, the real-time logistics network data obtained in step S1, and coal index data such as calorific value and total sulfur content into JSON format. The encapsulated data is encrypted using a hash algorithm and written into a smart contract deployed on the blockchain as a trigger condition. If the actual coal quality deviates from the contract agreement at the time of delivery, the smart contract will automatically execute the final payment deduction instruction based on the deviation value.

[0030] A real-time coal price quotation and negotiation trading system for implementing the aforementioned method, comprising:

[0031] Multi-dimensional environmental perception and data mapping module: used to connect with external meteorological bureau API, port scheduling system and futures exchange, to capture and clean multi-source heterogeneous market data in real time;

[0032] Spatiotemporal benchmark price projection engine: Built-in carbon accounting logic and coal quality decay algorithm, used to output a real-time benchmark price range that takes into account dynamic losses;

[0033] Multi-agent negotiation negotiation hub: Deploy a deep reinforcement learning model containing dual Actor-Critic networks for both buyers and sellers to replace manual execution of millisecond-level nonlinear concession negotiation;

[0034] Blockchain settlement and on-chain module: used to complete the matching and locking of the final price and deploy the performance terms as on-chain smart contracts.

[0035] The present invention also proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, enables the implementation of the aforementioned method for real-time coal price quoting and negotiated trading based on dynamic market conditions.

[0036] Compared with the prior art, the present invention has the following beneficial effects:

[0037] 1. Breaking through the time lag bottleneck of traditional static pricing, this invention achieves precise, dynamic pricing across all factors and timeframes. Current coal trading heavily relies on weekly or monthly long-term contract guidance prices, resulting in a delayed response to drastic intraday market fluctuations. This invention constructs a dynamic benchmark price assessment model based on a time-space decay mechanism by real-time collection of multi-source heterogeneous data, including macroeconomic market indices, micro-level coal indicators, dynamic logistics networks, and meteorological conditions. This mechanism can map sudden transient factors such as port congestion and soaring logistics rates into the benchmark price within milliseconds, completely eliminating the risk of pricing losses for enterprises due to information asymmetry.

[0038] 2. This invention quantifies the quality loss of coal during transit, significantly reducing the rate of contract disputes. Existing technologies generally overlook the objective fact that coal is highly susceptible to weather conditions during bulk transportation and open-air storage. This invention innovatively constructs a coal quality decay penalty function during transit. This mechanism converts the "increased total moisture and nominal decrease in calorific value" caused by rainfall along the route and the "weathering and absolute loss of calorific value" caused by strong winds into precise economic discounts for pre-emptive deduction. According to implementation examples, this mechanism, combined with implicit carbon emission cost accounting, has drastically reduced the quality dispute rate during delivery and fulfillment from 11.2% in the traditional model to 0.8%, significantly improving supply chain stability.

[0039] 3. Employing a multi-agent nonlinear reinforcement learning game theory approach significantly improves matching efficiency and transaction rate. Faced with massive inquiries, manual price negotiation is inefficient and easily influenced by subjective emotions; while traditional linear automatic price negotiation is too rigid. This invention introduces a deep reinforcement learning model based on an Actor-Critic network, which can automatically adjust the floor price based on the company's real-time inventory backlog severity or cash flow shortage, and employs a nonlinear dynamic concession step size. Experimental data demonstrates that this optimal game strategy enables the agent to possess advanced, anthropomorphic negotiation capabilities, such as "probing and protecting profits in the early stages and accelerating concessions to facilitate transactions in the later stages." This reduces the average negotiation interaction per order from 8.5 times to 4.2 times, boosts the automatic transaction rate to 94.8%, and reduces the matching time to only 185 milliseconds.

[0040] 4. Optimal Protection of Enterprise Net Profit: Through fully automated, high-dimensional environmental perception and dynamic bottom-line adjustment, the intelligent agent for both buyers and sellers in this invention can find the perfect Pareto optimal solution between avoiding escalating negotiation conflicts and avoiding excessive compromise. Comparative tests show that, compared with the manual static pricing model, the nonlinear negotiation model of this invention achieves a 6.7% real net profit optimization for the seller, maximizing the enterprise's business benefits while ensuring extremely high turnover.

[0041] 5. Transaction results are hashed and recorded on the blockchain, constructing an immutable trust loop. After the final match is reached, this invention encapsulates the final average transaction price of both sides, the corresponding logistics fulfillment conditions, and all coal indicators onto the blockchain by calling oracle nodes, generating a smart contract. If the actual coal quality verified during delivery deviates from the contract settings, the smart contract can automatically trigger a final payment deduction instruction, thereby completely eliminating offline disputes and defaults, and reshaping the digital trust mechanism for bulk commodity transactions. Attached Figure Description

[0042] Figure 1 This is a graph showing the real-time coal transaction price and the number of negotiations in a dynamic market-based real-time coal quotation and negotiation trading method proposed in this invention.

[0043] Figure 2 This is a flowchart illustrating a method for real-time coal price quoting and negotiated transactions based on dynamic market conditions, as proposed in this invention. Detailed Implementation

[0044] To make the technical means, creative features, objectives and effects of this invention easier to understand, the invention will be further described below in conjunction with specific embodiments.

[0045] Example 1

[0046] Data Collection and Benchmark Price: The system retrieves the Bohai Rim thermal coal price index and real-time freight rates from major ports every hour. The benchmark price is calculated using a simple index-weighted average. This does not include carbon tax accounting or penalties for quality degradation caused by weather.

[0047] Negotiation Mechanism: This embodiment does not activate the multi-agent adversarial central mechanism. When the buyer initiates an inquiry, the system recommends the aforementioned basic benchmark price to the seller's sales representative on the interface. The negotiation process adopts a semi-automatic method of "machine recommendation and manual confirmation." Both parties manually counter-offer in the system, and the system limits the concession step size to a fixed amount, with each concession being 1.0 yuan / ton.

[0048] Example 2

[0049] Spatiotemporal decay pricing: The system not only collects macroeconomic indices but also obtains rainfall warnings and strong wind probabilities for the next 5 days along the route from the production area to the destination and at ports by connecting to the meteorological bureau's API. The system invokes the coal transit quality decay penalty function described in claim 4. It automatically calculates economic losses such as increased moisture and decreased calorific value that may result from heavy rain, and deducts the corresponding discount from the initial quote.

[0050] Negotiation Mechanism: This embodiment achieves fully automatic negotiation, but its dynamic concession step size... A hard-coded linear function is used. That is, regardless of current inventory pressure or opponent sentiment, the agent adjusts its behavior with each round of negotiations. The negotiation proceeded with an equal concession of 0.5 yuan / ton each time, until the prices of both parties crossed or the maximum number of negotiation rounds of 15 were reached.

[0051] Example 3

[0052] Multi-dimensional environmental perception (S1-S2): The system accesses macro market indices (millisecond-level updates of futures market quotes, daily consumption of key power plants, etc.), micro coal indicators (calorific value, total sulfur, etc.), dynamic logistics networks (number of ships at anchor, railway congestion index), and end-to-end meteorological data in real time. These data are then mapped into multi-dimensional state vectors through a feature extraction model.

[0053] Carbon accounting and precise attenuation (S2-S3): The dynamic benchmark price is fully integrated into the carbon emission cost accounting mechanism, and the implicit carbon tax is automatically deducted based on the total sulfur content of coal and the expected combustion conversion rate. Meanwhile, the mass decay penalty function By employing the exponential compound attenuation formula in claim 4, the complex economic penalty of "rainwater infiltration, weathering loss, and absolute loss of calorific value" is accurately captured.

[0054] RL Nonlinear Adversarial Game (S3-S4): Intelligent agents for both buyers and sellers are trained based on an Actor-Critic network. The system reads the seller's real-time available inventory days and maps them to an inventory pressure urgency coefficient. When the seller has high inventory backlog, the bottom line is automatically lowered; conversely, when the buyer's inventory is critically low, the bottom line is raised. During negotiation, the intelligent agent uses the nonlinear concession step size constraint formula described in claim 7:

[0055]

[0056] Among them, the concession convexity coefficient Set to 1.5. The agent exhibits sophisticated dynamic game behavior: "initial aggressiveness to protect profits, followed by accelerated concessions to facilitate the transaction as time pressure increases in the mid-to-late stages."

[0057] Smart contract solidification (S5): Once the game converges within 4 rounds, the final transaction price and the accompanying quality data are automatically hashed, encrypted, and uploaded to the blockchain, and deployed as an immutable smart contract.

[0058] Example 4

[0059] System configuration: The basic data terminal, carbon accounting and weather attenuation model are completely consistent with those in Example 3.

[0060] Bargaining Mechanism: The game network parameters of the intelligent agent are manually intervened in. This involves adjusting the convexity coefficient of the seller's intelligent agent's concessions. The system was forcibly set to 5.0, while simultaneously increasing the urgency of the seller's inventory pressure. The AI ​​was forced to round down to zero, simulating a scenario where it mistakenly believed it had no inventory pressure and abundant funds. During negotiations, the seller's AI agent consistently adhered to its highest initial offer, making almost no concessions to the right.

[0061] Example 5

[0062] System configuration: The basic data terminal, carbon accounting and weather attenuation model are completely consistent with those in Example 3.

[0063] Negotiation Mechanism: The concession convexity coefficient of the intelligent agents of both buyers and sellers is manually adjusted. Setting it to 0.2 indicates extreme conservatism and a very short trial period, while also incorporating emotional feedback factors. Set to oversensitive. In negotiation, if the other party makes even a slight counter-offer, the agent will panic and make massive concessions in the first and second rounds in order to reach an unconditional deal within a very short period of time.

[0064] Experimental Data and Effect Verification: To rigorously verify the inventiveness of the core technical features of each claim of this invention, and the scientific validity of Example 3 as the optimal embodiment, the inventors utilized a full year's worth of real historical inquiry and settlement records (totaling...) from a large coal logistics trading platform in northern my country. (using a high-dimensional valid transaction sample), parallel backtesting experiments were conducted on the above five embodiments in a completely equivalent closed-loop sandbox simulation network.

[0065] 1. Definition of core evaluation indicators

[0066] Automatic Transaction Rate (%): The percentage of orders in the total number of inquiry orders in which the intelligent agents of both the buyer and seller successfully eliminate price differences and reach a price consensus within the preset maximum number of negotiation rounds.

[0067] Average negotiation rounds (times): The average total number of interactions between the buyer and seller during the process of a transaction from the triggering of the first offer to the final settlement or the complete breakdown of negotiations (reflecting matching efficiency and system computing overhead).

[0068] Profit optimization rate (%): Based on the net profit obtained from the traditional manual static guidance price transaction (0%), the percentage increase in the actual net profit locked by the seller after deducting the actual door-to-door logistics costs, environmental quality losses and expected carbon tax costs in each embodiment.

[0069] Contract dispute rate (%): During the delivery stage, the proportion of orders that trigger offline manual arbitration or legal disputes are caused by deviations in the calorific value, total moisture content, ash content, and other indicators of the coal actually delivered to the port / factory from the expected price at the time of contract signing due to adverse weather conditions along the way.

[0070] 2. Experimental Data Statistics Table The table below fully records the five implementation examples. Key performance indicators under a sample test of bulk coal inquiries:

[0071] Example 1 Basic dynamic price + semi-automatic manual quota concession 52.40% 8.5 +0.8% 11.20% 5.4 minutes Example 2 Includes weather attenuation + fully automatic hard-coded linear concession 76.10% 12.3 +3.5% 4.50% 120 milliseconds Example 3 Dynamic price of all factors + RL nonlinear game (optimal) 94.80% 4.2 +6.7% 0.80% 185 milliseconds Example 4 Dynamic pricing across all factors + aggressive agency (λ=5.0) 31.50% 2.1 +8.2% 1.10% 150 milliseconds Example 5 Dynamic price for all factors + conservative agency (λ=0.2) 98.20% 1.8 -4.10% 0.90% 145 milliseconds

[0072] 3. In-depth analysis of experimental data and technical conclusions: The control of dispute rate by multi-source data fusion and loss penalty:

[0073] Example 1, lacking a weather-induced mass decay penalty function, cannot offset calorific value loss in its pricing when encountering extreme weather events such as summer downpours or winter blizzards, resulting in a contract performance dispute rate as high as 11.2%. Examples 2 through 5, however, fully incorporate the exponential mass decay penalty function described in claim 4. The dispute rate plummeted. Example 3, by further integrating the mechanism of automatic deduction of final payments via blockchain smart contracts, suppressed the performance dispute rate to an extreme low of 0.8%, fully demonstrating the innovative and beneficial effects of the price-quality linkage accounting scheme. The improvement of matching efficiency through reinforcement learning nonlinear concession game theory:

[0074] While Comparative Example 2 achieves fully automated negotiation, its linear, fixed-step approach (a hard concession of 0.5 yuan each time) fails to detect the opponent's willingness to compromise, resulting in an extremely rigid and lengthy negotiation process with an average of 12.3 negotiation rounds. This makes it highly susceptible to missing the optimal transaction window in the rapidly changing market. In contrast, Example 3 employs a nonlinear convexity concession function dynamically adjusted based on an Actor-Critic network. The agent maintains its profit in the early stages of negotiation and converges rapidly in the later stages, requiring only an average of 4.2 interactions to perfectly lock in the price, improving matching efficiency by nearly 3 times.

[0075] Verification of the comprehensive coordination and superiority of optimal implementation example 3: Although example 4 achieved the highest single-transaction profit optimization rate of +8.2% among successfully completed orders, its overly aggressive negotiation strategy and refusal to compromise led to the direct failure of a large number of potential transactions, causing the automatic transaction rate to plummet to 31.5%. From the perspective of the company's overall macro revenue, this aggressive strategy will lead to serious inventory backlog and a broken capital chain. Although example 5 achieved a high transaction rate of 98.2% and extremely short negotiation rounds (1.8 times) through rapid compromise without any bottom line, it paid an extremely heavy economic price, with a negative profit optimization rate (-4.1%).

[0076] This means that companies are excessively cutting losses in transactions, severely damaging normal operating profit margins. Example 3 found the perfect Pareto optimal balance between a transaction rate of 94.8% and a profit optimization rate of +6.7%. It avoids missing out on the market due to blind arrogance like Example 4, and also avoids losing pricing power due to panic-driven compromises like Example 5.

[0077] Conclusion: The experimental data comprehensively and intuitively demonstrate that the composite technical solution constructed in the claims of this invention, consisting of a "multi-source heterogeneous spatiotemporal pricing center" and a "nonlinear multi-agent bargaining center based on reinforcement learning," exhibits significant technological advancements unmatched by traditional bulk trading and linear automated trading in multiple core business and technical indicators, such as ensuring optimal corporate profits, achieving millisecond-level efficiency, and minimizing contract disputes. This perfectly verifies that Example 3 is the optimal embodiment of this invention.

[0078] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely illustrative of the principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the present invention as claimed. The scope of protection of this invention is defined by the appended claims and their equivalents.

Claims

1. A method for real-time coal pricing and negotiation based on dynamic market conditions, characterized in that, The method includes the following steps: S1: Real-time collection of multi-source heterogeneous market data, which includes at least macro market index data, micro coal index data, dynamic logistics network data, and meteorological environmental factor data; through the constructed data cleaning and feature extraction model, the above-mentioned multi-source heterogeneous market data is mapped into a standardized multi-dimensional state vector; S2: Construct a dynamic benchmark price assessment model based on a spatiotemporal decay mechanism; input the multidimensional state vector into the dynamic benchmark price assessment model, and combine the carbon emission cost accounting mechanism and the coal quality decay penalty function in transit to calculate the time-sensitive initial dynamic benchmark price range for coal buying and selling at the current time node; S3: Initialize the multi-agent bargaining game model under asymmetric information; based on the historical transaction profiles of buyers and sellers, capital turnover rate and inventory pressure indicators, set a dynamic reserve price floor and initial negotiation step size for the buyer's intelligent agent and the seller's intelligent agent, respectively. S4: Automatically execute multi-round reinforcement learning negotiation strategy; within the initial dynamic benchmark price range, the intelligent agents of the buyer and seller interact with each other based on the Markov decision process; in each round of interaction, the intelligent agent evaluates the other party's pricing intention and market urgency through its built-in nonlinear concession function, and outputs the preferred price for the next round until the price difference between the two parties is less than the preset convergence threshold or the maximum number of negotiation rounds is reached. S5: Transaction matching and automatic generation of smart contracts; when the price difference between the buyer and seller is less than the convergence threshold, the final transaction price and the corresponding logistics performance conditions are extracted, the transaction metadata is put on the blockchain, and an immutable coal transaction smart contract is automatically generated using the blockchain.

2. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 1, characterized in that, The multi-source heterogeneous market data in step S1 specifically includes: The macro market index data includes: the Bohai Rim thermal coal price index, real-time order book data of the Zhengzhou Commodity Exchange thermal coal futures main contract, and daily coal consumption and available days of key power plants nationwide. The microscopic coal index data include: calorific value, total moisture content, ash content, volatile matter content, and total sulfur content; The dynamic logistics network data includes: railway dispatch plans and congestion index, real-time number of anchored vessels at major ports, inventory at ports, and real-time sea / road freight rates. The meteorological environmental data includes: rainfall warnings along the route from the production area to the destination, the probability of extreme wind and snow weather, and the forecast of port closure duration.

3. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 2, characterized in that, The dynamic benchmark price evaluation model based on the spatiotemporal decay mechanism in step S2 calculates the basic benchmark price. The improvement process is as follows: Existing benchmark prices rely solely on simple index weighting. This invention introduces carbon tax adjustment and end-to-end cost mapping, the calculation formula of which is as follows: ;in, For the first Real-time values ​​of macro market indices. For the corresponding weights; This refers to the quality premium or discount calculated based on micro-level coal index data; The comprehensive door-to-door logistics cost is calculated in real time based on dynamic logistics network data. This refers to the implicit carbon emission cost calculated based on the total sulfur content of coal and the expected combustion conversion rate. This is the penalty function for the quality degradation of coal during transit.

4. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 3, characterized in that, Existing technologies neglect the quality loss of coal due to weather during long-distance transportation and open-air storage. This invention constructs a penalty function for the quality decay of coal during transit. The specific formula is as follows: ,in, The price of bare coal before deducting losses; For the estimated total transportation and storage time; for Rainfall intensity factor along the route and at the port at all times; for Wind intensity factor at any given time; and These are the sensitivity coefficients for the decrease in heat generation due to water penetration and the sensitivity coefficients for weathering loss, respectively. This represents the absolute loss of coal calorific value predicted based on meteorological and environmental factor data. This is the economic penalty multiplier for heat loss.

5. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 4, characterized in that, In step S3, the buyer's intelligent agent dynamically reserves the floor price. Dynamic price floor reserved with seller's smart agent The setting not only relies on static costs but also dynamically incorporates real-time inventory pressure: when the buyer's inventory is running low, the highest price they are willing to pay (i.e., the buyer's reservation price) will automatically increase, as represented by the formula: When a seller has excess inventory, the lowest price they are willing to accept will automatically decrease, as represented by the formula: ;in, The urgency coefficient is obtained by nonlinearly mapping the real-time available inventory days and cash flow health of an enterprise through a logistic regression model.

6. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 5, characterized in that, In step S4, traditional negotiation methods use fixed-step concessions, which can easily lead to stalemates or premature compromises. In this invention, intelligent agents for both buyers and sellers use a non-linear concession function based on time pressure and remaining pricing space to generate the next round of optimal pricing. In the first round of negotiations, the current offer from one party's intelligent agent... The calculation logic is as follows: ,in, For the first The dynamic yield step size of the wheel is determined by the intelligent agent through reinforcement learning. Network output aims to maximize long-term expected returns.

7. The method for real-time coal pricing and negotiation based on dynamic market conditions according to claim 1, characterized in that, To ensure that the bargaining model can converge quickly and avoid getting stuck in an infinite loop, the dynamic concession step size is... In the reinforcement learning action space, there are explicit mathematical constraints, and the baseline constraint formula is: ,in, Reserve a price floor for the dynamics of this intelligent agent; The maximum number of negotiation rounds set for the system; To accommodate the convexity coefficient, when This strategy manifests as an initial hardline stance followed by significant concessions later on. This strategy manifests as initial tentative concessions followed by a firm adherence to bottom lines. This is a trading sentiment feedback factor calculated based on the concessions made by the opponent in the previous round.

8. The method for real-time coal price quotation and negotiation based on dynamic market conditions according to claim 1, characterized in that, The transaction matching and smart contract automatic generation step S5 specifically includes: when hour, To preset the convergence threshold, the average of the final bids from both parties is taken as the final transaction price. The system calls the oracle node in the blockchain network to encapsulate the final transaction price, the real-time logistics network data obtained in step S1, and coal index data such as calorific value and total sulfur content into JSON format. The encapsulated data is encrypted using a hash algorithm and written into a smart contract deployed on the blockchain as a trigger condition. If the actual coal quality deviates from the contract agreement at the time of delivery, the smart contract will automatically execute the final payment deduction instruction based on the deviation value.

9. A real-time coal price quotation and negotiation trading system for implementing the dynamic market conditions described in any one of claims 1 to 8, characterized in that, The system includes: Multi-dimensional environmental perception and data mapping module: used to connect with external meteorological bureau API, port scheduling system and futures exchange, to capture and clean multi-source heterogeneous market data in real time; Spatiotemporal benchmark price projection engine: Built-in carbon accounting logic and coal quality decay algorithm, used to output a real-time benchmark price range that takes into account dynamic losses; Multi-agent negotiation negotiation center: Deploy a deep reinforcement learning model containing dual Actor-Critic networks for both buyers and sellers to replace manual execution of millisecond-level nonlinear concession negotiation; Blockchain settlement and on-chain module: used to complete the matching and locking of the final price and deploy the performance terms as on-chain smart contracts.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program can implement a method for real-time coal price quotation and negotiation trading based on dynamic market conditions as described in any one of claims 1 to 8.

Images