Wind-solar-hydrogen-methanol integrated public island flexible energy supply system and optimization control method

Abstract

This invention discloses a flexible energy supply system and optimized control method for an integrated wind-solar-hydrogen-methanol public island. The system includes a public works island physical system, a data acquisition and soft measurement layer, a model predictive control optimization layer, and an execution layer. The public works island physical system includes a circulating water subsystem, a steam pipeline subsystem, and a turbine-generated electronic system. The data acquisition and soft measurement layer outputs the circulating water demand and steam consumption based on the XGBoost soft measurement model. The model predictive control optimization layer uses a circulating water MPC controller and a steam pipeline dual-layer MPC controller for collaborative optimization, generating control commands to drive the execution layer. This invention upgrades the public works system from "passive protection" to "active optimization," significantly reducing the energy consumption of the circulating water system, achieving dynamic matching of the steam pipeline network and cascade utilization of waste heat, and improving the overall energy efficiency of the entire plant.

Description

Technical Field

[0001] This invention relates to a flexible energy supply system and optimized control method for an integrated public island for wind, solar, hydrogen, and methanol production, belonging to the field of intelligent control technology in the chemical industry. Background Technology

[0002] Currently, in integrated wind-solar-hydrogen-methanol projects, the utilities system, as the infrastructure ensuring the normal operation of the methanol production unit, typically includes multiple subsystems such as a circulating water system, a steam pipeline system, a demineralized water system, and an instrument air system. These projects use renewable energy sources such as wind and solar power to produce hydrogen through water electrolysis and then synthesize methanol through hydrogenation of carbon dioxide. The main unit's production load fluctuates frequently with changes in wind and solar power output, placing extremely high demands on the flexible power supply capability of the utilities system. Its typical structure is as follows: the circulating water system consists of several fixed-frequency or variable-frequency circulating water pumps, cooling towers, pipelines, and user-end heat exchange equipment. The pumps operate at a constant design flow rate or are simply adjusted based on the outlet pressure. The steam pipeline system consists of main pipelines of different pressure levels, desuperheaters and pressure reducers, condensate drain stations, and steam consumption units. Pressure balance between different pressure levels is maintained through desuperheaters with fixed openings. The control systems are independent of each other, with each subsystem controlled by a separate PLC or DCS loop using conventional PID control.

[0003] The aforementioned existing technology model has the following significant drawbacks in integrated wind-solar-hydrogen-methanol projects: First, there is a lack of real-time and accurate means of sensing utility loads. Key loads such as circulating water demand and steam consumption cannot be directly measured online. Operators can only set the number of pumps to be turned on and the valve opening based on design conditions or experience values. When the production load of the methanol production unit is frequently adjusted with the fluctuations of wind and solar power, the utility system cannot respond synchronously, resulting in a contradiction of "over-powered" or "supply falling short of demand." The average annual power consumption of the circulating water system deviates from the optimal value by more than 15%.

[0004] Secondly, the control method is crude, with each subsystem operating independently. Circulating water pump groups typically use a stepped control approach of "adding pumps on demand and maintaining a fixed frequency," which fails to dynamically optimize pump and fan speeds based on time-varying factors such as ambient temperature and humidity, and heat exchanger fouling thermal resistance. Cooling tower fans are often controlled by the number of units, ignoring the impact of wet-bulb temperature changes on cooling performance, resulting in significant energy waste in the system.

[0005] Third, the pressure level matching of the steam pipeline network is rigid. Different pressure levels are only connected by desuperheaters and pressure reducers with fixed parameters, which cannot realize the cascade utilization of steam according to the dynamic needs of each steam-using unit. When there is an excess of high-grade waste heat, it cannot drive turbine power generation. When there is insufficient waste heat, a large amount of fresh steam needs to be supplemented from outside the boundary. The overall plant thermal efficiency is far below the design value.

[0006] Fourth, the utility system is completely isolated from the methanol production unit control system, lacking a feedforward control mechanism based on methanol production unit load forecasting. This results in a severe lag in the response of the utility system, further exacerbating energy consumption losses.

[0007] In summary, existing technologies lack a flexible energy supply system and its optimization control method for integrated wind-solar-hydrogen-methanol projects that can achieve online soft measurement and prediction of utility loads, global collaborative optimization of the circulating water system, dynamic matching and cascade utilization of the steam pipeline network. This invention aims to solve the aforementioned technical problems in this specific application scenario and provide a flexible energy supply system and its optimization control method for the utility island of an integrated wind-solar-hydrogen-methanol project. Summary of the Invention

[0008] The purpose of this invention is to address the technical problems existing in the utility systems of integrated wind-solar-hydrogen-methanol projects, such as lack of load sensing, independent control, and rigid energy matching. This invention provides a flexible energy supply system and optimized control method for the utility island of integrated wind-solar-hydrogen-methanol projects to solve the above-mentioned technical problems and achieve intelligent operation with on-demand energy supply and extreme energy saving.

[0009] To solve the above-mentioned technical problems, the present invention adopts the following technical solution: a flexible energy supply system for integrated wind, solar, hydrogen production, and methanol production on a public island, comprising a public engineering island physical system, a data acquisition and soft measurement layer, a model predictive control optimization layer, and an execution layer.

[0010] The physical system of the utility island includes a circulating water subsystem, a steam pipeline subsystem, and a turbine power generation system.

[0011] The circulating water subsystem consists of a circulating water pump, a water supply header, heat exchange equipment for each process unit, a return water header, and a cooling tower connected in sequence to form a closed-loop circulation system. The circulating water pump delivers cooling water to the inlet of the heat exchange equipment for each process unit via the water supply header. The high-temperature circulating water after heat exchange is sent back to the cooling tower via the return water header. The cooling tower cools the circulating water before sending it to the inlet of the circulating water pump, thus completing the closed-loop circulation.

[0012] The steam pipeline subsystem includes at least two steam headers of different pressure levels. The steam headers of adjacent pressure levels are connected by a desuperheater and pressure reducer to perform pressure matching during normal operation and to serve as a bypass to ensure steam supply safety under abnormal operating conditions.

[0013] The turbine power generation system includes a turbine generator set. The turbine generator set's inlet is connected to a higher-pressure steam header, and its outlet is connected to a lower-pressure steam header. Thus, high-grade steam preferentially enters the turbine generator set to generate electricity, while the discharged lower-grade steam continues to supply steam-consuming units at the corresponding pressure level, achieving cascaded energy utilization.

[0014] The data acquisition and soft sensing layer is used to collect characteristic data from the methanol production unit and utilities system, and outputs the circulating water demand and steam consumption of the steam-using units based on the soft sensing model. This layer acquires real-time operating data through a sensor network and uses a data-driven soft sensing model to achieve accurate sensing and prediction of critical loads that cannot be directly measured online.

[0015] The model predictive control optimization layer performs coordinated optimization of the circulating water subsystem and the steam pipeline network subsystem based on the water demand and steam consumption, generating control commands. This layer uses load data provided by the soft sensing layer as input and solves for the optimal operating variables through advanced control algorithms to achieve global coordinated optimization.

[0016] The execution layer receives the control commands and drives the various actuators in the public works island physical system to achieve on-demand power supply and flexible operation.

[0017] As a preferred embodiment, the steam pipeline subsystem is equipped with steam headers of three pressure levels: high pressure, medium pressure, and low pressure. A first desuperheater and pressure reducer connects the high-pressure header to the medium-pressure header, and a second desuperheater and pressure reducer connects the medium-pressure header to the low-pressure header. The steam inlet of the turbine generator set is connected to the high-pressure header, and the steam outlet is connected to the medium-pressure header. This connection method ensures that the high-pressure waste heat steam preferentially performs work through the turbine generator set, the discharged medium-pressure steam is then supplied to the medium-pressure steam-consuming unit, and excess steam can be replenished into the low-pressure header via the desuperheater and pressure reducer, forming a complete cascade utilization path.

[0018] Furthermore, the high-pressure header is a 2.5MPa header, the medium-pressure header is a 1.0MPa header, and the low-pressure header is a 0.5MPa header; the medium-pressure header is connected to a distillation reboiler, and the low-pressure header is connected to a deaerator. The above pressure rating settings and steam unit configuration match the steam requirements of a typical methanol production unit, demonstrating broad engineering applicability.

[0019] As a preferred embodiment, the data acquisition and soft sensing layer includes a sensor network and a soft sensing model based on a gradient boosting decision tree. The sensor network is used to acquire multidimensional feature data, including at least the syngas flow rate of the methanol production unit, reactor load rate, ambient dry-bulb temperature, ambient wet-bulb temperature, relative humidity, atmospheric pressure, and the operating status of the circulating water pumps and cooling tower fans. The soft sensing model takes the multidimensional feature data as input and outputs the current water demand of the circulating water system, the predicted future water demand sequence, and the real-time steam consumption of the main steam-consuming units.

[0020] Furthermore, the soft sensing model employs the XGBoost algorithm, with model parameters of 500 trees, a maximum depth of 8, and a learning rate of 0.05; the future water demand prediction sequence consists of water demand predictions for one point every 15 minutes over the next two hours. Practical verification shows that this combination of model parameters achieves an optimal balance between computational efficiency and prediction accuracy.

[0021] As a preferred embodiment, the model predictive control optimization layer includes a circulating water MPC controller and a steam pipeline dual-layer MPC controller.

[0022] The circulating water MPC controller uses the frequency of each circulating water pump and the start / stop status of each cooling tower fan as control variables, the supply water temperature and return water temperature as controlled variables, and the ambient wet-bulb temperature and the future water demand prediction provided by the soft sensing layer as disturbance variables. Its optimization objective is to minimize the total power consumption of the circulating water pumps and cooling tower fans. Therefore, the circulating water MPC controller can fully utilize load forecast information to adjust in advance while meeting process cooling requirements, avoiding the lag and oscillation of traditional PID control, thus achieving energy saving and consumption reduction.

[0023] The dual-layer MPC controller for the steam network uses steam allocation at each pressure level as the decision variable in the upper layer and turbine generator power and desuperheater / pressure reducer opening as the execution variables in the lower layer. Under the constraints of each steam-consuming unit, the optimization objectives are to maximize turbine power generation, minimize desuperheater / pressure reducer throttling losses, and minimize auxiliary boiler steam supply. This dual-layer architecture decomposes the complex steam network optimization problem into two levels: the upper layer is responsible for global allocation decisions, and the lower layer is responsible for specific execution. This ensures global optimality while reducing the computational burden of online solutions.

[0024] Furthermore, the circulating water MPC controller adopts a dynamic matrix control algorithm with a prediction time domain of 120 minutes, a control time domain of 30 minutes, and a sampling period of 15 minutes; the circulating water pump is a variable frequency pump with a frequency adjustment range of 30-50Hz.

[0025] As a preferred embodiment, the execution layer includes: an analog control interface connected to the circulating water pump frequency converter, a digital control interface connected to the cooling tower fan contactor, an analog control interface connected to the desuperheater actuator, and an analog control interface connected to the turbine generator set speed control system. Each actuator operates in real time according to the optimization instructions issued by the MPC controller, ensuring accurate execution of the optimization results.

[0026] This invention also provides an optimized control method for a flexible energy supply system for a wind-solar-hydrogen-methanol integrated public island, comprising the following steps: S1. Real-time acquisition of multi-dimensional characteristic data of methanol production units and utilities systems; S2. Input the feature data into the soft measurement model to obtain the current water demand, future water demand prediction sequence, and steam consumption of the main steam-consuming units of the circulating water system; S3. Input the water demand prediction sequence and steam consumption into the model prediction control optimization layer. The circulating water MPC controller solves the frequency of each circulating water pump and the start-stop state of the cooling tower fan under the cooling constraints to minimize the total power consumption. The steam network dual-layer MPC controller solves the steam distribution of each pressure level and the power of the turbine generator set under the steam demand to maximize the turbine power generation and minimize the throttling loss and the amount of supplementary steam. S4. The optimization results are converted into control signals through the execution layer to drive the circulating water pump frequency converter, cooling tower fan, desuperheater and pressure reducer and turbine generator set.

[0027] Furthermore, the optimization strategy of the steam network dual-layer MPC controller includes: when the amount of waste heat steam at a higher pressure level exceeds the steam demand at the corresponding pressure, the excess steam is preferentially introduced into the turbine generator set for power generation and then supplied to the lower pressure level main pipe; when the amount of waste heat steam at a higher pressure level is insufficient, the turbine power generation is preferentially reduced to ensure the supply of steam to the steam-consuming unit, and the insufficient part is supplemented by the auxiliary boiler.

[0028] Compared with the prior art, the present invention has at least the following beneficial effects: This invention, through a soft sensor model based on XGBoost, achieves a real-time estimation error of only 2.3% for circulating water demand under typical summer daily operating conditions, a prediction error of 3.8% for the next 2 hours, and a steam consumption estimation error of less than 2.1%. It upgrades from traditional "experience-based blind adjustment" to "data-driven precise prediction," laying the foundation for on-demand energy supply and enabling accurate perception and prediction of utility loads.

[0029] This invention utilizes model predictive control to perform global collaborative optimization of four variable frequency water pumps and six cooling tower fans. Under typical summer day conditions, the daily power consumption of the circulating water system is 3850 kWh, a 16.7% reduction compared to the traditional PID scheme of 4620 kWh. The water supply temperature qualification rate has increased from 92.3% to 98.5%, and the number of equipment start-ups and shutdowns has decreased from 38 to 12, significantly extending equipment lifespan and significantly reducing the energy consumption of the circulating water system while improving operational stability.

[0030] This invention utilizes a dual-layer MPC optimization architecture to fully utilize 2.5MPa waste heat steam in a cascade manner under typical winter day conditions, achieving a turbine power generation of 4320kWh / day with zero auxiliary boiler steam supplementation, saving 1500 yuan in daily operating costs compared to traditional solutions; the system operates safely under extreme conditions without steam supply interruption, realizing dynamic matching of multiple pressure levels in the steam pipeline network and cascade utilization of waste heat.

[0031] This invention significantly improves the overall energy efficiency and economic benefits of integrated wind-solar-hydrogen-methanol projects. A 30-day long-term operation verification shows that the proposed solution saves a cumulative 98,500 kWh of electricity under typical operating conditions, representing a 17.2% energy saving rate; generates a cumulative 128,000 kWh of turbine power; reduces auxiliary boiler steam supply by 1,650 tons; saves a total of 280,000 yuan in operating costs; and further reduces the overall plant energy efficiency by 5%-8%, upgrading the utility system from "passive support" to "active optimization." Attached Figure Description

[0032] Figure 1 This is a schematic diagram of the circulating water subsystem structure of the system of the present invention; Figure 2 This is a schematic diagram of the turbine power generation system and steam pipeline subsystem of the present invention. Figure 3 This is a schematic diagram of the overall architecture principle of the system of the present invention.

[0033] Figure 4 This is a logic flowchart of the method of the present invention.

[0034] The present invention will be further described below with reference to the accompanying drawings and specific embodiments. Detailed Implementation

[0035] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. However, it should be understood that the scope of protection of the present invention is not limited to the specific embodiments.

[0036] First, it's crucial to emphasize the fundamental differences between the utility systems of integrated wind-solar-hydrogen-methanol projects and conventional coal chemical or natural gas-to-methanol projects. Conventional methanol projects typically operate at 90%-100% of their design capacity, with utility systems functioning in a fixed pattern and load fluctuations generally within ±5%. However, in integrated wind-solar-hydrogen-methanol projects, wind and solar power outputs exhibit minute-level random fluctuations. The load for water electrolysis to produce hydrogen must be adjusted in real-time according to the solar and wind power outputs, leading to synchronous fluctuations in the methanol synthesis reactor and distillation column loads. Consequently, the circulating water demand and steam consumption of the utility systems exhibit large-scale and rapid fluctuations (typical load fluctuation range 60%-110%, with a change rate reaching ±2% per minute). This end-to-end load transmission characteristic caused by fluctuations in solar and wind power output is the core technical challenge faced by the utility systems of integrated wind-solar-hydrogen-methanol projects, and it is also a unique problem specific to this application scenario that this invention focuses on addressing. Therefore, in view of the problems existing in the above-mentioned integrated wind-solar-hydrogen-methanol projects, the present invention provides a flexible energy supply system for integrated wind-solar-hydrogen-methanol public islands, the overall architecture of which is as follows: The public works island flexible energy supply system of the present invention consists of four core parts: public works island physical system, data acquisition and soft measurement layer, model predictive control optimization layer and execution layer.

[0037] The utilities island physical system includes a circulating water subsystem, a steam piping subsystem, and a turbine power generation system: The circulating water subsystem is equipped with 4 variable frequency circulating water pumps and 6 cooling tower fans. Each pump has a rated flow rate of 4000 m³ / h, a head of 50 m, and a frequency range of 30-50 Hz. Each cooling tower fan has a power of 45 kW and an air volume of 450,000 m³ / h. The circulating water pump outlets are connected to the supply water header, which in turn connects to the inlets of the heat exchange equipment in each process unit. The high-temperature circulating water after heat exchange is returned to the cooling tower via the return water header. The cooling tower cools the circulating water before sending it back to the circulating water pump inlet, forming a closed-loop circulation.

[0038] The steam pipeline subsystem is equipped with steam headers at three pressure levels: 2.5MPa, 1.0MPa, and 0.5MPa. The 2.5MPa header is the high-pressure header, the 1.0MPa header is the medium-pressure header, and the 0.5MPa header is the low-pressure header. The main steam-consuming units include a distillation reboiler (rated steam capacity 8t / h) connected to the 1.0MPa header, a deaerator (rated steam capacity 3t / h) connected to the 0.5MPa header, and a heat tracing system (2t / h steam capacity in winter). A first desuperheater / pressure reducer T1 (maximum flow rate 15t / h) connects the 2.5MPa header to the 1.0MPa header, and a second desuperheater / pressure reducer T2 (maximum flow rate 10t / h) connects the 1.0MPa header to the 0.5MPa header. The 2.5MPa main pipe steam mainly comes from the waste heat recovery steam from methanol synthesis, with a fluctuation range of 5-15t / h; the auxiliary boiler can supplement steam to the 1.0MPa main pipe.

[0039] The turbine generator system is equipped with one back-pressure turbine generator set, with the steam inlet connected to the 2.5MPa high-pressure main pipe and the steam outlet connected to the 1.0MPa medium-pressure main pipe. The rated power is 200kW and the isentropic efficiency is 75%.

[0040] The data acquisition and soft sensing layer comprises a sensor network and an XGBoost-based soft sensing model. The sensor network collects real-time 12-dimensional feature data from the methanol production unit, including syngas flow rate, reactor load rate, distillation reboiler steam flow rate, ambient dry-bulb temperature, ambient wet-bulb temperature, relative humidity, atmospheric pressure, circulating water pump operating power, and cooling tower fan operating status. The soft sensing model takes these 12-dimensional features as input, with model parameters of 500 trees, a maximum depth of 8, and a learning rate of 0.05. It outputs the current actual water demand (m³ / h) of the circulating water system and a predicted water demand sequence for the next two hours (one point every 15 minutes), as well as the real-time steam consumption of the distillation reboiler and deaerator.

[0041] The model predictive control optimization layer includes a circulating water MPC controller and a steam network dual-layer MPC controller. The circulating water MPC controller adopts a dynamic matrix control algorithm, with the frequency of 4 variable frequency water pumps (30-50Hz) and the start / stop status of 6 cooling tower fans as control variables, the supply water temperature setpoint of 32℃ and the return water temperature of each cooling unit ≤40℃ as controlled variables, and the ambient wet-bulb temperature and the water demand prediction for the next 2 hours provided by the soft sensor layer as disturbance variables. The prediction time domain is 120 minutes, the control time domain is 30 minutes, and the sampling period is 15 minutes. The optimization objective is to minimize the total power consumption of the circulating water pumps and cooling tower fans.

[0042] The dual-layer MPC controller for the steam network uses steam distribution at each pressure level as the decision variable at the upper layer and turbine generator power and desuperheater / pressure reducer opening as the execution variables at the lower layer. It maximizes turbine power generation and minimizes desuperheater / pressure reducer throttling losses and auxiliary boiler steam supply while meeting the needs of each steam-consuming unit. Specifically, its optimization strategy is as follows: when the 2.5MPa waste heat steam production exceeds the 1.0MPa level steam demand, excess steam is preferentially introduced into the turbine generator for power generation and then supplied to the 1.0MPa or 0.5MPa main pipe; when the 2.5MPa waste heat steam production cannot meet the 1.0MPa level steam demand, turbine power generation is preferentially reduced to ensure the supply to the steam-consuming units, with the shortfall supplemented by the auxiliary boiler.

[0043] The execution layer includes: a 4-20mA analog control signal interface connected to the circulating water pump frequency converter; a DO digital control interface connected to the cooling tower fan contactor; a 4-20mA analog control signal interface connected to the desuperheater actuator; and a 4-20mA analog control signal interface connected to the turbine generator set speed control system. Each actuator operates in real time according to the optimization commands issued by the MPC controller, achieving on-demand power supply and flexible operation of the utility system.

[0044] This invention also provides a control method flow for a flexible energy supply system for a wind-solar-hydrogen-methanol integrated public island. The specific flow of the optimized control method of this invention is as follows: Step S1: Collect multi-dimensional characteristic data in real time through sensor network, such as syngas flow rate of methanol production unit, reactor load rate, steam flow rate of distillation reboiler, ambient dry bulb temperature, ambient wet bulb temperature, relative humidity, atmospheric pressure, operating power of circulating water pump, and operating status of cooling tower fan.

[0045] Step S2: Input the collected multidimensional feature data into the XGBoost-based soft measurement model. The model outputs the current actual water demand of the circulating water system, the predicted water demand sequence for the next 2 hours (one point every 15 minutes), and the real-time steam consumption of the distillation reboiler and deaerator.

[0046] Step S3: Input the water demand forecast sequence and steam consumption into the model prediction control optimization layer.

[0047] Under the constraints of a supply water temperature setpoint of 32℃ and a return water temperature of ≤40℃ for each cooling unit, the circulating water MPC controller aims to minimize the total power consumption of the circulating water pumps and cooling tower fans. It continuously solves the inverter frequency (30-50Hz) of each circulating water pump and the start / stop status of each cooling tower fan in the future control time domain.

[0048] The steam pipeline double-layer MPC controller, under the constraint of meeting the steam demand of various steam-using units such as distillation reboiler, deaerator, and heat tracing system, aims to maximize turbine power generation, minimize desuperheater and pressure reducer throttling losses, and minimize auxiliary boiler steam supply, and solves the steam distribution scheme, turbine generator set power, and desuperheater and pressure reducer opening degree for each pressure level.

[0049] Step S4: The execution layer converts the optimization results into 4-20mA analog control signals and DO digital control signals, which drive the circulating water pump frequency converter, cooling tower fan contactor, desuperheater and pressure reducer actuator, and turbine generator set speed control system to achieve on-demand power supply and flexible operation of the public works system.

[0050] The technical solution of the present invention will be further described in detail below with reference to specific embodiments and comparative examples. The specific embodiments of the present invention described below are based on the design data of the 100,000 tons / year methanol synthesis section utility system of the Liaoning Huadian Diaobingshan 450MW wind power-hydrogen coupled green methanol integrated demonstration project. A fully digital demonstration was conducted through a self-developed dynamic simulation and optimization platform for the utility system. The simulation platform adopts a four-in-one technical system of "process mechanism modeling - data-driven soft measurement - model predictive control - full system energy efficiency assessment". The circulating water system is based on the thermodynamic performance curve of a certain type of counter-flow cooling tower and the characteristic curve of a centrifugal water pump. A high-fidelity dynamic model is constructed in AspenPlus, configuring 4 circulating water pumps (3 in operation and 1 on standby, rated flow rate of 4,000 m³ / h per pump, head of 50 m, equipped with frequency converters) and 6 cooling tower fans (power of 45 kW per fan, air volume of 450,000 m³ / h per fan). The steam pipeline system, based on the design data of the Diaobingshan project, is equipped with headers at three pressure levels: 0.5MPa, 1.0MPa, and 2.5MPa. The main steam-consuming units include a distillation reboiler (1.0MPa, rated steam consumption 8t / h), a deaerator (0.5MPa, rated steam consumption 3t / h), and a heat tracing system (0.5MPa, winter steam consumption 2t / h). It is equipped with two desuperheating and pressure reducing units (2.5MPa→1.0MPa, 1.0MPa→0.5MPa) and one small back-pressure turbine generator set (inlet 1.0MPa, back pressure 0.5MPa, rated power 200kW). Environmental meteorological data (dry-bulb temperature, wet-bulb temperature, and relative humidity) for the entire year of 2024 at the Diaobingshan project site are used. The main unit load sequence is generated based on wind and solar load fluctuations, covering a load range of 60%-110%. All simulation data are strictly aligned with the project design parameters to ensure that the technical solution fully matches the actual working conditions of the project.

[0051] Example 1: Construction and Validation of Soft Measurement and Prediction Model for Utility Loads under Baseline Operating Conditions This embodiment aims to fully construct the soft measurement and prediction model of public works load described in this invention, and to verify its prediction accuracy under typical annual meteorological conditions.

[0052] First, the experimental platform was configured as follows: A dynamic model of the circulating water system was built in AspenPlus. The model inputs were the ambient wet-bulb temperature, the main unit process load (characterized by the methanol synthesis reactor load), and the setpoint for the temperature difference between the circulating water inlet and outlet (design value 8℃). The model outputs were the minimum circulating water volume (m³ / h) required by the circulating water system and the corresponding cooling tower fan operation strategy. A soft sensor model based on XGBoost was built in Python 3.9. The model input features included: the main unit syngas flow rate (50,000±15,000 Nm³ / h), the methanol synthesis reactor load rate (60%-110%), the distillation column reboiler steam flow rate (8±2 t / h), the ambient dry-bulb temperature (-25℃~35℃), the ambient wet-bulb temperature (-25℃~30℃), the relative humidity (30%-95%), the atmospheric pressure (98-103 kPa), the circulating water pump operating power at the previous moment, and the cooling tower fan operating status at the previous moment. The model outputs the current actual water demand (m³ / h) of the circulating water system and the predicted water demand sequence for the next 2 hours (one point every 15 minutes).

[0053] Secondly, a training dataset was constructed. Hourly meteorological data for the entire year of 2023 from the Diaobingshan project site was used, combined with a generator for random load fluctuations in the main unit, to generate 10,000 sets of operational data for the public works system under different operating conditions, covering high temperature and humidity in summer, low temperature and humidity in winter, the spring and autumn transition seasons, and various combinations of load fluctuations. Each set of data includes input features and the corresponding true value of circulating water demand (calculated by the AspenPlus dynamic model). The dataset was divided into a training set (8,000 sets) and a test set (2,000 sets) in an 8:2 ratio. The XGBoost model parameters were set as follows: 500 trees, maximum depth 8, learning rate 0.05, subsampling rate 0.8, feature sampling rate 0.8, and 50 early shutdown rounds.

[0054] Third, a soft-sensor model for steam consumption was constructed. Simplified models based on process mechanisms were established for key steam-consuming units such as the distillation reboiler and deaerator, and then corrected using data-driven methods. The steam consumption model for the distillation reboiler uses crude methanol feed rate (based on a 10 t / h baseline), column top reflux rate, and column bottom temperature as inputs, and outputs real-time steam consumption. The deaerator model uses deoxygenated water flow rate, inlet water temperature, and outlet water temperature as inputs. The models also employ the XGBoost architecture, and the training data comes from simulation results of the AspenPlus distillation column dynamic model and deaerator model.

[0055] Fourth, model validation was conducted. A typical daily operating condition in 2024 (July 15th, a high-temperature day in summer, with a dry-bulb temperature of 32℃, a wet-bulb temperature of 26℃, and the main unit load fluctuating between 80% and 105%) was selected as the validation scenario. Hourly meteorological data and the main unit load sequence for that day were input into the soft sensor model to obtain the predicted circulating water demand, which was then compared with the true value calculated by the AspenPlus dynamic model.

[0056] The verification results show that the mean absolute percentage error (MAPE) between the real-time estimated circulating water demand and the true value is 2.3%, with a maximum error of 4.1%; the mean absolute percentage error of the predicted water demand for the next 2 hours (updated every 15 minutes) is 3.8%, which meets the requirements of model predictive control. In the soft-sensor model for steam consumption, the estimation error of the steam flow rate of the distillation reboiler is 2.1%, and the estimation error of the steam flow rate of the deaerator is 1.8%. This embodiment verifies the feasibility and accuracy of the utility load soft-sensor and prediction model based on XGBoost.

[0057] Example 2: Verification of Global Optimization of Circulating Water System Based on Model Predictive Control Based on the soft measurement and prediction model in Example 1, this embodiment constructs a predictive controller for a circulating water system and verifies its energy-saving effect under typical summer conditions.

[0058] First, the experimental model and parameter settings completely reuse the results of Example 1, including the AspenPlus dynamic model of the circulating water system and the XGBoost load prediction model (weights consistent with the training results of Example 1). A new model predictive controller was added, using Python 3.9 to implement the dynamic matrix control (DMC) algorithm. The control variables are: the inverter frequencies of the four circulating water pumps (adjustable range 30-50Hz, continuously adjustable) and the start / stop status of the six cooling tower fans (0 / 1, discrete variables, groupable control). The controlled variables are: the circulating water supply temperature (set value 32℃, determined by process requirements) and the return water temperature of each cooling unit (constraint ≤40℃). The disturbance variables are: the ambient wet-bulb temperature and the main unit process load (provided by the soft sensor model with a 2-hour prediction sequence). The prediction time domain is set to 120 minutes, the control time domain to 30 minutes, and the sampling period to 15 minutes. The optimization objective is to minimize the total power consumption of the circulating water pumps and cooling tower fans while meeting the cooling load demand and the supply water temperature constraint.

[0059] Secondly, a simulation verification was conducted on a typical summer day. The same operating conditions as in Example 1 were selected: July 15th (dry-bulb temperature 32℃, wet-bulb temperature 26℃, main unit load fluctuation 80%-105%). A control group was set up: a traditional PID control scheme (circulating water pumps controlled at constant outlet pressure, 2 fixed-frequency pumps + 1 variable-frequency pump, cooling tower fans started and stopped in stages according to wet-bulb temperature, manually set based on experience). The simulation ran for 24 hours, recording the total power consumption of the circulating water system, the water supply temperature qualification rate, and the number of pump and fan start-stop cycles under both schemes.

[0060] Simulation results show that the total daily power consumption of the MPC scheme of this invention is 3850 kWh, while that of the traditional PID scheme is 4620 kWh, resulting in a power saving rate of 16.7%. The water supply temperature qualification rate is 98.5% for the MPC scheme (within the set value of 32℃±1℃), and 92.3% for the traditional scheme (with some periods of overheating). The number of pump and fan start-stop cycles is 12 for the MPC scheme (smooth adjustment), and 38 for the traditional scheme (frequent start-stops affect equipment lifespan). During the load increase phase (e.g., the main unit load increases from 85% to 105% between 10:00 and 12:00), the MPC scheme gradually increases the pump frequency and starts the fan 30 minutes in advance based on load prediction, causing the water supply temperature to rise steadily to 32.5℃ and then fall back to 32℃; the traditional scheme responds with lag, causing the water supply temperature to jump to 35℃, triggering an over-temperature alarm and requiring an additional pump to be started, resulting in temperature oscillations. This embodiment verifies the significant advantages of the MPC-based global optimization technology for circulating water systems in terms of energy saving and stable control.

[0061] Example 3: Verification of Dynamic Matching and Waste Heat Cascade Utilization of Multi-Pressure Level Steam Pipeline Network Based on the models in Examples 1 and 2, this embodiment adds a dynamic model of the steam pipeline network and a turbine generator set to verify the optimization effect of dynamic matching of steam pressure levels and cascade utilization of waste heat.

[0062] First, the experimental model was expanded: a dynamic model of the steam network was constructed in AspenPlus, including headers at three pressure levels: 0.5MPa, 1.0MPa, and 2.5MPa. The main steam-consuming units were a distillation reboiler (1.0MPa, rated 8t / h), a deaerator (0.5MPa, rated 3t / h), and a heat tracing system (0.5MPa, winter operation). Two desuperheaters and pressure reducers were configured: T1 (2.5MPa→1.0MPa, maximum flow rate 15t / h) and T2 (1.0MPa→0.5MPa, maximum flow rate 10t / h). One back-pressure turbine generator set was configured, with an inlet pressure of 1.0MPa, a back pressure of 0.5MPa, a rated power of 200kW, and an isentropic efficiency of 75%. Steam sources included: steam recovered from methanol synthesis waste heat (2.5MPa, fluctuation range 5-15t / h) and auxiliary boiler supplementation (1.0MPa, adjustable). The optimized controller adopts a two-layer MPC structure: the upper layer optimizes the matching of steam network pressure levels, and the lower layer controls the power of the turbine generator set. The optimization objective is to maximize turbine power generation and minimize the throttling losses of the desuperheater and pressure reducer and the auxiliary boiler steam supply while meeting the needs of each steam-consuming unit.

[0063] Secondly, a typical winter day verification condition was designed (January 15th, ambient temperature -15℃, heat tracing system in operation). The main unit load fluctuation caused the 2.5MPa waste heat steam output to vary between 8-14 t / h; the steam consumption of the distillation reboiler fluctuated with the load between 6-9 t / h; the deaerator steam consumption remained constant at 3 t / h; and the heat tracing system steam consumption remained constant at 2 t / h. A control group was set up: a traditional constant pressure control scheme (the desuperheater and pressure reducer maintained the main pipe pressure at a fixed opening, the turbine was not in operation or only manually operated).

[0064] The simulation ran for 24 hours, recording the supply and demand balance of the steam network at various pressure levels, turbine power generation, and auxiliary boiler steam supply under the two schemes.

[0065] Simulation results show that in the optimized scheme of this invention, when the 2.5MPa waste heat steam output exceeds the demand of the distillation reboiler (e.g., 12t / h output, reboiler requires 8t / h), the controller prioritizes reducing the excess 4t / h of steam to 1.0MPa via the turbine generator set, while simultaneously driving the turbine to generate electricity. When the total 1.0MPa steam output at the turbine outlet (turbine exhaust steam + remaining steam after reboiler use) exceeds the demand of the deaerator and heat tracing, it is then supplied to the 0.5MPa user via the T2 desuperheater and pressure reducer. The total daily turbine power generation is 4,320kWh, and the auxiliary boiler steam supply is 0 (completely met by waste heat). In the traditional scheme, desuperheaters T1 and T2 are always partially open, the excess 2.5MPa steam is directly throttled to 1.0MPa and 0.5MPa, there is no turbine power generation, and when the waste heat is insufficient, the auxiliary boiler needs to be started to supply approximately 2.5t / h of steam. Calculations show that the traditional scheme requires an auxiliary boiler steam supply equivalent to 1.8 tons of standard coal per day, increasing operating costs by approximately 1,500 yuan.

[0066] In extreme operating condition verification (e.g., a sudden drop in waste heat from 2.5 MPa to 6 t / h, while the reboiler requires 8 t / h), the optimized controller prioritizes reducing turbine power generation, using all 1.0 MPa steam for the reboiler, with any shortfall supplemented by the auxiliary boiler. Simultaneously, it ensures that 0.5 MPa users are primarily supplied with turbine exhaust steam, supplemented by the T2 desuperheater and pressure reducer if insufficient. The system operates within safety constraints without steam supply interruption. This embodiment verifies the significant energy-saving effect and robustness of the dynamic matching and cascade utilization technology for steam pipeline networks.

[0067] Comparative Example: Comparison of the effects of traditional independent control schemes and this patented solution This comparative example aims to quantitatively compare the differences in technical effects between the traditional independent control scheme and the present invention using the same simulation platform and operating conditions as Examples 2 and 3.

[0068] First, the traditional scheme is set up as follows: The circulating water system uses PID constant pressure control, with 3 water pumps (2 industrial and 1 variable) operating. The cooling tower fans start and stop in stages according to wet-bulb temperature (1 fan operates below 22℃, 2 fans operate between 22-26℃, and 3 fans operate above 26℃), without load prediction and collaborative optimization. The steam pipeline system uses constant pressure control, with desuperheaters and pressure reducers T1 and T2 maintaining the main pipe pressure at a fixed opening (2.5MPa main pipe pressure 2.5±0.1MPa, 1.0MPa main pipe pressure 1.0±0.1MPa, 0.5MPa main pipe pressure 0.5±0.1MPa). The turbine generator set is not in operation (due to lack of dynamic coordination, operators are concerned about affecting the stability of steam supply). Each subsystem operates independently and has no linkage with the main unit's DCS.

[0069] Secondly, a typical summer day simulation verification was conducted, similar to that of Example 2. The total power consumption of the circulating water system was 4620 kWh, which is 20.0% higher than the 3850 kWh in Example 2 of this invention; the water supply temperature qualification rate was 92.3%, lower than the 98.5% of this invention; and the equipment started and stopped frequently (38 times), affecting its lifespan.

[0070] The same simulation verification as in Example 3 was conducted on a typical winter day. The auxiliary boiler steam supply was 2.5 t / h (equivalent to 60 t / day), increasing the operating cost by approximately 1,500 yuan / day, while the turbine power generation was 0.

[0071] Comparison of 30-day long-term operation (using actual meteorological data in January 2024): The proposed solution saved a total of 98,500 kWh of electricity, a saving rate of 17.2%; the cumulative turbine power generation was 128,000 kWh, and the cumulative reduction in auxiliary boiler steam supply was 1,650 tons, resulting in a total saving of approximately RMB 280,000 in operating costs. The traditional solution, due to frequent equipment start-ups and shutdowns, resulted in a cooling tower fan bearing failure, causing an unplanned shutdown and a loss of approximately RMB 50,000.

[0072] Table 1 Summary of key parameter settings and operating conditions for each embodiment and comparative example Circulating water system 4×4,000 m³ / h pumps, 6×45 kW fans Same as Example 1 Same as Example 1 Same as Example 1 Steam Pipeline 0.5 / 1.0 / 2.5MPa, reboiler 8t / h, deaerator 3t / h, heat tracing 2t / h Same as Example 1 Same as Example 1 Same as Example 1 Soft measurement model XGBoost (500 trees, depth 8) Reuse Example 1 Reuse Example 1 none Control Algorithm Open-loop verification DMC-MPC Dual-layer MPC PID + Human Experience Turbine configuration none none 200kW back pressure type none Verification conditions Typical summer day Typical summer day Typical winter day Typical summer and winter days Meteorological data July 15, 32 / 26℃ July 15 January 15, -15℃ Same as Example 2 / 3 Table 2 Comparison of System Response and Technical Effect of Each Embodiment and Comparative Example Error in circulating water demand prediction MAPE 2.3% - - No prediction No prediction Daily power consumption of circulating water system - 3,850 kWh 4,120 kWh (including winter) 4,620 kWh 4,950 kWh Energy saving rate - 16.7% 16.8% (Winter) benchmark benchmark Water supply temperature qualification rate - 98.5% 97.8% 92.3% 91.5% Number of equipment start-ups and shutdowns - 12 times 15 times 38 times 42 times Auxiliary boiler steam supply - - 0t / day - 60t / day Turbine power generation - - 4,320 kWh / day 0 0 30-day cumulative electricity savings - - 98,500 kWh benchmark benchmark 30-day cumulative reduction in fuel supply - - 1,650t benchmark benchmark 30 days of operating cost savings - - 280,000 yuan benchmark benchmark Unplanned shutdown - 0 0 1 time 0 Table 3. Comparison of core performance indicators between the present invention and traditional solutions. Control Mode Independent PID, experience-based operation Model predictive control, global collaborative optimization Essential leap Load sensing capability No, based solely on experience Soft measurement + prediction, MAPE < 4% From blind surveys to precise surveys Energy consumption of circulating water system benchmark Reduced by 16%-17% Significant energy savings Steam cascade utilization Significant losses due to throttling, no power generation Dynamic matching + turbine power generation From waste to recycling Auxiliary boiler dependence high Reduce by 100% (when there is sufficient residual heat) Essential improvement Equipment operation stability Frequent start-stop cycles and short lifespan Smooth adjustment, extended lifespan Improve reliability Overall plant energy efficiency benchmark Additional reduction of 5%-8% Generational leap In summary, the detailed descriptions of the three embodiments and one comparative example above fully demonstrate that the flexible energy supply system for the utility island of the integrated wind-solar-hydrogen-methanol project and its optimized control method described in this invention can achieve on-demand energy supply, extreme energy saving, and intelligent operation of the utility system under various operating conditions, significantly outperforming traditional independent control schemes. The scope of protection of this invention covers, but is not limited to, the specific values ​​mentioned in the above embodiments. Any simple modifications based on the core concept of this invention fall within the scope of protection of this invention.

Claims

1. A flexible energy supply system for a public island integrating wind, solar, hydrogen production, and methanol production, suitable for integrated projects using wind and solar renewable energy as power sources and coupling water electrolysis for hydrogen production with carbon dioxide hydrogenation to methanol production, characterized in that... It includes the public works island physical system, data acquisition and soft measurement layer, model predictive control optimization layer and execution layer; The physical system of the utility island includes a circulating water subsystem, a steam pipeline subsystem, and a turbine power generation system. The circulating water subsystem consists of a circulating water pump, a water supply header, heat exchange equipment for each process unit, a return water header, and a cooling tower connected in sequence to form a closed-loop circulation loop. The steam pipeline subsystem includes at least two steam headers of different pressure levels, with adjacent steam headers connected by a desuperheater and pressure reducer. The turbine power generation system includes a turbine generator set, with its steam inlet connected to the higher-pressure steam header and its exhaust port connected to the lower-pressure steam header. The data acquisition and soft measurement layer is used to collect characteristic data of methanol production units and utilities systems, and outputs the circulating water demand and steam consumption of steam-using units based on the soft measurement model. The model predictive control optimization layer performs collaborative optimization of the circulating water subsystem and the steam pipeline network subsystem based on the water demand and steam consumption, and generates control commands. The execution layer receives the control commands and drives the various actuators in the public works island physical system.

2. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 1, characterized in that, The steam pipeline subsystem is equipped with steam headers of three pressure levels: high pressure, medium pressure, and low pressure. A first desuperheater and pressure reducer is connected between the high pressure header and the medium pressure header, and a second desuperheater and pressure reducer is connected between the medium pressure header and the low pressure header. The steam inlet of the turbine generator set is connected to the high pressure header, and the steam outlet is connected to the medium pressure header.

3. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 2, characterized in that, The high-pressure header is a 2.5MPa grade header, the medium-pressure header is a 1.0MPa grade header, and the low-pressure header is a 0.5MPa grade header; the medium-pressure header is connected to a distillation reboiler, and the low-pressure header is connected to a deaerator.

4. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 1, characterized in that, The data acquisition and soft measurement layer includes a sensor network and a soft measurement model based on a gradient boosting decision tree. The sensor network is used to collect multidimensional feature data, including at least the syngas flow rate of the methanol production unit, reactor load rate, ambient dry-bulb temperature, ambient wet-bulb temperature, relative humidity, atmospheric pressure, and the operating status of the circulating water pump and cooling tower fan. The soft measurement model takes the multidimensional feature data as input and outputs the current water demand of the circulating water system, the predicted sequence of future water demand, and the real-time steam consumption of the main steam-consuming units.

5. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 4, characterized in that, The soft measurement model uses the XGBoost algorithm, with model parameters of 500 trees, a maximum depth of 8, and a learning rate of 0.05; the future water demand prediction sequence is the water demand prediction value at one point every 15 minutes in the next 2 hours.

6. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 1, characterized in that, The model predictive control optimization layer includes a circulating water MPC controller and a steam pipeline dual-layer MPC controller. The circulating water MPC controller uses the frequency of each circulating water pump and the start / stop status of each cooling tower fan as control variables, the supply water temperature and return water temperature as controlled variables, the ambient wet-bulb temperature and the future water demand prediction provided by the soft measurement layer as disturbance variables, and minimizing the total power consumption of the circulating water pump and cooling tower fan as the optimization objective. The upper layer of the steam network dual-layer MPC controller uses steam distribution at each pressure level as the decision variable, while the lower layer uses turbine generator power and desuperheater / pressure reducer opening as the execution variables. Under the constraints of each steam-consuming unit, the optimization objectives are to maximize turbine power generation, minimize desuperheater / pressure reducer throttling losses, and minimize auxiliary boiler steam supply.

7. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 6, characterized in that, The circulating water MPC controller adopts a dynamic matrix control algorithm with a prediction time domain of 120 minutes, a control time domain of 30 minutes, and a sampling period of 15 minutes; the circulating water pump is a variable frequency pump with a frequency adjustment range of 30-50Hz.

8. The integrated wind-solar-hydrogen-methanol flexible energy supply system for a public island according to claim 1, characterized in that, The execution layer includes an analog control interface connected to the circulating water pump frequency converter, a digital control interface connected to the cooling tower fan contactor, an analog control interface connected to the desuperheater actuator, and an analog control interface connected to the turbine generator set speed control system.

9. A method for optimizing and controlling the flexible energy supply of a utility island based on the system described in any one of claims 1 to 8, characterized in that, include: S1. Real-time acquisition of multi-dimensional characteristic data of methanol production units and utilities systems; S2. Input the feature data into the soft measurement model to obtain the current water demand, future water demand prediction sequence, and steam consumption of the main steam-consuming units of the circulating water system; S3. Input the water demand prediction sequence and steam consumption into the model prediction control optimization layer. The circulating water MPC controller solves the frequency of each circulating water pump and the start-stop state of the cooling tower fan under the cooling constraints to minimize the total power consumption. The steam network dual-layer MPC controller solves the steam distribution of each pressure level and the power of the turbine generator set under the steam demand to maximize the turbine power generation and minimize the throttling loss and the amount of supplementary steam. S4. The optimization results are converted into control signals through the execution layer to drive the circulating water pump frequency converter, cooling tower fan, desuperheater and pressure reducer and turbine generator set.

10. The optimized control method according to claim 9, characterized in that, The optimization strategy of the steam network dual-layer MPC controller includes: when the amount of waste heat steam at a higher pressure level exceeds the steam demand at the corresponding pressure, the excess steam is preferentially introduced into the turbine generator set for power generation and then supplied to the lower pressure level main pipe; when the amount of waste heat steam at a higher pressure level is insufficient, the turbine power generation is preferentially reduced to ensure the supply of steam to the steam-consuming unit, and the insufficient part is supplemented by the auxiliary boiler.

Images