Variable load scheduling system and method based on load prediction and economic model prediction

Abstract

The present application relates to the technical field of air separation device, and particularly relates to a variable load scheduling system and method based on load prediction and economic model prediction. The system comprises a data acquisition module, a load prediction module and an economic MPC control module. The data acquisition module acquires device operation, customer gas consumption, tank liquid level and time-of-use electricity price data and stores historical data sets; the load prediction module predicts future gas consumption based on multi-dimensional historical data as a feedforward signal to the economic MPC control module. The present application cancels the traditional RTO hierarchical architecture, reconstructs the MPC control matrix, and sets the real-time electricity cost as a controlled variable; relying on a dynamic cost weighting mechanism, the load is reduced to reduce liquid production during high electricity price periods, and the load is increased to increase liquid product energy storage during low electricity price periods. The present application has fast response speed and does not require manual intervention, can fully utilize the peak-valley electricity price to reduce the electricity cost of the air separation device, and takes into account gas supply stability and product quality.

Description

Technical Field

[0001] This invention relates to the field of air separation unit technology, and in particular to a variable load dispatching system and method based on load forecasting and economic model forecasting. Background Technology

[0002] Air separation units (ASUs) are core energy-consuming units in process industries such as metallurgy and chemicals, with electricity costs typically accounting for over 70% of their operating costs. Currently, industrial power grids generally implement peak-valley time-of-use pricing policies, resulting in significant price differences at different times of day, providing substantial optimization opportunities for energy conservation and consumption reduction in air separation units.

[0003] Traditional air separation unit optimization control typically employs a two-tier architecture of Real-Time Optimization (RTO) and Advanced Process Control (APC). However, this architecture has the following inherent drawbacks in practical applications: First, the RTO layer is based on a steady-state mechanism model, while the APC layer uses a dynamic response model. The inconsistency between the two models makes it difficult to accurately execute the theoretically optimal setpoint calculated by the RTO layer at the APC layer, and may even cause system oscillations. Second, the optimization cycle of RTO is typically 1-2 hours, making it impossible to track minute-level fluctuations in electricity price signals in real time, and also difficult to quickly respond to demand-side dispatch commands from the power grid. Third, the two-tier architecture increases system investment costs, maintenance complexity, and the risk of hardware-software coupling.

[0004] To address the aforementioned issues, there is an urgent need to develop a single-layer optimized control system that can directly embed economic indicators into the control layer and eliminate intermediate RTO links, thereby achieving the technical goal of control as optimization. This would improve the air separation unit's response speed to electricity price signals and its adaptive scheduling capability, and further tap into its energy-saving potential. Summary of the Invention

[0005] To address the technical problems in the existing two-layer architecture of air separation unit optimization control, such as system oscillation caused by inconsistency between the two-layer models, low response efficiency due to long optimization cycles, and high cost due to complex architecture, this invention provides a variable load scheduling system and method based on load forecasting and economic model forecasting.

[0006] Therefore, the present invention provides the following technical solution: A variable load dispatching system based on load forecasting and economic model forecasting includes a data acquisition module, a load forecasting module, and an economical MPC control module. The data acquisition module is used to collect in real time the operating parameters of the air separation unit, the real-time gas consumption of downstream customers, the liquid level data of the liquid storage tank, and the real-time time-of-use electricity price data of the power grid, and to store the collected data information as a historical dataset. The load forecasting module retrieves historical air separation unit operating parameters, historical downstream customer gas consumption, historical liquid storage tank level, and historical grid time-of-use electricity price from the historical dataset as input features for the forecasting model, forecasts the downstream customer gas demand in the future control time domain, and sends the forecast result as a feedforward interference signal to the economical MPC control module. The economical MPC control module employs an economic model predictive control algorithm. It constructs an MPC dynamic model matrix using the collected operating parameters of the air separation unit, and introduces real-time electricity cost or total power consumption of the air separation unit as new controlled variables into the MPC dynamic model matrix. The economical MPC control module receives the predicted future downstream customer gas demand from the load forecasting module as a measurable feedforward constraint boundary. During the rolling optimization process, it constructs an optimization objective function with the goal of minimizing electricity cost and ensuring product purity is within the constraints. It automatically calculates the optimal manipulator trajectory for adjusting the operation of the air separation unit, thereby achieving adaptive economic dispatch of the air separation unit to adapt to the peak and valley electricity prices of the power grid.

[0007] Furthermore, the economical MPC control module has a built-in state observer, which is used to update the real-time electricity cost of the MPC dynamic model matrix based on the data collected in real time by the data acquisition module.

[0008] Furthermore, the economical MPC control module is configured with a dynamic cost-weighted mechanism: During periods of high electricity prices, the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix is ​​increased. Under the premise of meeting the lower limit of downstream gas supply, product purity and equipment safety constraints, the controller automatically searches for the lowest energy consumption condition of the air separation unit, thereby reducing the operating load and liquid output of the air separation unit. During periods of low electricity prices, the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix is ​​reduced, and soft constraints on liquid storage tank levels that tend to be high are added. The economical MPC control module uses low-priced electricity during off-peak hours to increase the operating load of the air separation unit and increase the production of liquid products until the equipment operating constraints or the high limit of the liquid storage tank level are reached.

[0009] Furthermore, the MPC dynamic model matrix of the economical MPC control module includes the controlled variable CV, the manipulated variable MV, and the disturbance variable DV; The controlled variable CV includes real-time electricity cost, product oxygen purity, product nitrogen purity, pressure at the air separation unit, and liquid storage tank level; the real-time electricity cost is the minimization optimization objective, the product oxygen purity, product nitrogen purity, and pressure at the air separation unit are hard constraint indicators, and the liquid storage tank level is an interval control energy storage buffer indicator. The manipulated variables include the opening degree of the guide vane of the raw material air compressor, the opening degree of the expander nozzle, the opening degree of the liquid oxygen take-off valve, and the opening degree of the distillation reflux ratio regulating valve, which are used to regulate the energy consumption of the unit's air intake, the cooling capacity of the system, the liquid output and the distillation conditions, respectively. The interference variables include the real-time electricity price of the power grid and the gas demand of downstream customers. The real-time electricity price of the power grid is an interference variable that is updated and optimized in real time, and the gas demand of downstream customers is a feedforward interference signal output by the load forecasting module.

[0010] Furthermore, the constraints on the controlled variables are specifically as follows: oxygen purity is constrained to ≥ 99.6%, nitrogen purity is constrained to oxygen content less than 5 ppm, upper tower pressure is constrained to less than 50 kPa, and liquid storage tank level is constrained to a range of 20%-90%. The specific adjustment range of the manipulated variables is as follows: guide vane opening of the raw material air compressor 0-100%, nozzle opening of the expander 0-100%, liquid oxygen take-off valve opening 0-100%, and reflux ratio regulating valve opening 0-100%.

[0011] Furthermore, the system adopts a single-layer MPC control architecture that eliminates the RTO layer, and is equipped with a comprehensive control server; the comprehensive control server carries and independently runs the economical MPC control module, the server is configured with a DCS interface, and communicates with the air separation unit DCS system through the OPC DA protocol.

[0012] Furthermore, the system is configured with a solver to support the operation of the economical MPC control module. The solver is selected from IPOPT or Gurobi and supports quadratic programming and nonlinear programming solutions.

[0013] The method for implementing the variable load dispatch optimization system based on load forecasting and economic model prediction includes the following steps: S1. The data acquisition module collects real-time operating data of the air separation unit, real-time gas consumption of downstream customers, liquid storage tank level data, and real-time time-of-use electricity price data of the power grid and stores them to form a historical dataset. The load forecasting module retrieves multi-dimensional historical feature data from the historical dataset, predicts the gas demand of downstream customers in the future control time domain, and sends the predicted value as a feedforward interference signal to the economical MPC control module. S2, the economical MPC control module calls the built-in state observer to update the real-time electricity price of the power grid in the MPC dynamic model matrix based on real-time acquired data; S3. Construct an optimization objective function using the MPC dynamic model matrix in the economical MPC control module. Based on the dynamic cost weighting mechanism, dynamically adjust the optimization weight of the electricity cost item in the objective function according to the real-time electricity price of the power grid. S4. Use the configured IPOPT or Gurobi solver to perform single-layer rolling optimization on the objective function. The solution process is constrained by the customer's gas demand forecast, product purity, tower pressure, and storage tank level range. During periods of high electricity price, output manipulated variable schemes to reduce air compressor load and reduce liquid production. During periods of low electricity price, output manipulated variable schemes to increase air separation unit load and increase liquid storage production when the storage tank level does not reach the upper limit constraint. S5. Send the optimal manipulated variable increment obtained from the solution to the air separation unit DCS system to adjust the raw material air compressor guide vanes, expander nozzles, liquid oxygen extraction valve, and reflux ratio regulating valve to complete the adaptive adjustment of the air separation unit's operating conditions.

[0014] Furthermore, the economical MPC control module is configured with a dynamic cost weighting mechanism, and the objective function is calculated as follows:

[0015] In the formula: Let k be the value of the overall objective function for MPC rolling optimization. The real-time power consumption cost of the space separation unit at time k; This is the weighting coefficient for electricity cost items. It can be set as a constant and combined with the electricity price multiplier to achieve peak-valley differentiated adjustment, or dynamically adjusted in real time according to the dispatch strategy. These are controlled variables for product quality, including oxygen purity, nitrogen purity, and pressure at the top of the tower. Reference values ​​are set for the controlled variables of product quality. This refers to the weighting coefficient for quality tracking error. To control the penalty weighting coefficient for actions; To manipulate the change in the variable; These are traditional quality tracking errors and control action penalty terms, used to constrain parameter fluctuations and ensure stable device operation.

[0016] Advantages and positive effects of the present invention: The upper-level RTO setting value distribution link is eliminated, and a single-layer economical MPC architecture is adopted. The hardware only requires a high-performance integrated control server to complete all optimization calculations, resulting in lower equipment investment costs and a reduction of investment and maintenance costs by about 30%. It also reduces the risk of latency and communication failures caused by multi-platform data interaction.

[0017] MPC directly calculates the optimal valve operation quantity in real time based on electricity price and gas demand and sends it to DCS, eliminating the need for RTO to calculate and transmit set values ​​layer by layer. This significantly improves the response speed of operating condition adjustment and adapts to scenarios where the power grid's peak and valley electricity prices switch rapidly.

[0018] It eliminates the problem of hierarchical data matching bias, avoiding the defects of traditional RTO and MPC two-layer architectures such as hierarchical target conflict, setting lag, and model mismatch.

[0019] By directly using real-time electricity costs as a controlled variable in rolling optimization, it is no longer limited to simply tracking fixed values ​​for flow and purity, but achieves autonomous optimization with the lowest electricity cost as the core, which aligns with the cost reduction requirements of peak-valley electricity pricing policies.

[0020] During peak electricity prices, the load on the air separation unit and liquid production are automatically reduced to maximize the reduction of peak power consumption while ensuring gas supply and product quality. During off-peak electricity prices, energy consumption constraints are relaxed, and liquid oxygen and liquid nitrogen are actively produced and stored in storage tanks. Low-priced electricity is used for energy storage to achieve peak shifting and valley filling, significantly reducing overall electricity costs. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram of the overall architecture of the variable load scheduling system based on load forecasting and economic model forecasting provided by the present invention (single-layer MPC architecture).

[0023] Figure 2 Define a diagram for the MPC control matrix that includes economic indicators.

[0024] Figure 3 The flowchart illustrates the optimization method for implementing a variable load dispatching system based on load forecasting and economic model prediction provided by this invention.

[0025] Figure 4 This is a flowchart illustrating the load adjustment logic in an embodiment of the present invention. Detailed Implementation

[0026] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0027] This invention provides a variable load dispatching system based on load forecasting and economic model prediction, such as... Figure 1 As shown, it includes a data acquisition module, a load forecasting module, and an economical MPC control module.

[0028] The data acquisition module is used to collect real-time operating parameters of the air separation unit, real-time gas consumption of downstream customers, liquid storage tank level data, and real-time time-of-use electricity price data of the power grid, and to store the collected data as a historical dataset.

[0029] The load forecasting module retrieves historical air separation unit operating parameters, historical downstream customer gas consumption, historical liquid storage tank levels, and historical grid time-of-use electricity prices from historical datasets as input features for the forecasting model. It then forecasts the downstream customer gas demand in the future control time domain and sends the forecast result as a feedforward interference signal to the economical MPC control module.

[0030] The economic MPC control module employs an Economic Model Predictive Control (EMC) algorithm. It constructs an MPC dynamic model matrix using collected air separation unit operating parameters, introducing real-time electricity costs or the total power consumption of the air separation unit as new controlled variables. The economic MPC control module receives future downstream customer gas demand forecasts from the load forecasting module as measurable feedforward constraints. During rolling optimization, the control module no longer tracks fixed upper-level values ​​but directly constructs an optimization objective function with the goal of minimizing electricity costs and ensuring product purity remains within constraints. It automatically calculates the optimal manipulated variable trajectory for adjusting the air separation unit's operation, achieving adaptive economic dispatch of the air separation unit to adapt to peak-valley electricity prices. Model predictive control ensures that fluctuations in key parameters (purity, pressure) during load changes are less than ±0.5%.

[0031] like Figure 2 As shown, the MPC dynamic model matrix of the economical MPC control module includes the controlled variable CV, the manipulated variable MV, and the disturbance variable DV.

[0032] The controlled variable CV includes real-time electricity cost, product oxygen purity, product nitrogen purity, pressure at the air separation unit, and liquid storage tank level. Real-time electricity cost is the optimization objective to minimize, product oxygen purity, product nitrogen purity, and pressure at the air separation unit are hard constraints, and liquid storage tank level is the energy storage buffer index for interval control.

[0033] The manipulated variables include the opening degree of the guide vane of the raw material air compressor, the opening degree of the expander nozzle, the opening degree of the liquid oxygen extraction valve, and the opening degree of the distillation reflux ratio regulating valve, which are used to regulate the energy consumption of the unit's air intake, the cooling capacity of the system, the liquid output, and the distillation conditions, respectively.

[0034] The interference variables include the real-time electricity price of the power grid and the gas demand of downstream customers. The real-time electricity price of the power grid is an interference variable that is updated and optimized in real time, and the gas demand of downstream customers is a feedforward interference signal output by the load forecasting module.

[0035] The specific constraints for the controlled variables are as follows: oxygen purity is ≥ 99.6%, nitrogen purity is less than 5 ppm, pressure at the top of the tower is less than 50 kPa, and liquid level in the storage tank is 20%-90%.

[0036] The adjustment ranges for the manipulated variables are as follows: raw material air compressor guide vane opening 0-100%, expander nozzle opening 0-100%, liquid oxygen extraction valve opening 0-100%, and reflux ratio regulating valve opening 0-100%. See the table below for each variable: Table 1. Variables of the MPC Dynamic Model Matrix

[0037] The economical MPC control module has a built-in state observer, which is used to update the real-time electricity cost of the MPC dynamic model matrix based on the data collected in real time by the data acquisition module.

[0038] The economical MPC control module is configured with a dynamic cost-weighted mechanism, which dynamically adjusts the weight of electricity costs in the objective function based on the current electricity price. During periods of high electricity prices, the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix is ​​increased. Under the premise of meeting the lower limit of downstream gas supply, product purity and equipment safety constraints, the controller automatically searches for the lowest energy consumption condition of the air separation unit and reduces the operating load and liquid output of the air separation unit, such as reducing the compressor guide vanes and reducing liquid output.

[0039] During periods of low electricity prices, the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix is ​​reduced, and a soft constraint favoring higher liquid levels in the liquid storage tank is added. This soft constraint serves only as an optimization guide and does not impose mandatory boundaries. During off-peak electricity price periods, a penalty term is added to the optimization objective to encourage the liquid level to approach higher values. For example, if a hard constraint for the upper limit of the liquid level in the storage tank is set to 90% of the tank height, the controller will proactively increase liquid production and raise the tank level when there is sufficient margin between the tank level and the 90% upper limit. Only when the liquid level approaches the 90% upper limit hard constraint will the increase in liquid product production be stopped. This soft constraint only changes the optimization direction and will not exceed the safe range of the upper and lower limits of the liquid level, balancing energy storage needs and equipment operation safety. The economical MPC control module utilizes the low-priced electricity during off-peak hours to increase the operating load of the air separation unit and increase liquid product production until the equipment operation constraints or the upper limit of the liquid storage tank level are reached.

[0040] The system adopts a single-layer MPC control architecture that eliminates the RTO layer, and is equipped with a comprehensive control server. The comprehensive control server carries and independently runs an economical MPC control module. The server is configured with a DCS interface and communicates with the air separation unit's DCS system via the OPC DA protocol to achieve millisecond-level data interaction. The comprehensive control server has a CPU with ≥8 cores and ≥32GB of memory.

[0041] The system is configured with a solver to support the operation of the economical MPC control module. The solver is either IPOPT or Gurobi and supports quadratic programming and nonlinear programming to handle complex optimization problems with economic objectives.

[0042] Methods for implementing variable load dispatching systems based on load forecasting and economic model prediction, such as Figure 3 As shown, it includes the following steps: S1. The data acquisition module collects real-time operating data of the air separation unit, real-time gas consumption of downstream customers, liquid storage tank level data, and real-time time-of-use electricity price data of the power grid and stores them to form a historical dataset. The load forecasting module retrieves multi-dimensional historical feature data from the historical dataset, predicts the gas demand of downstream customers in the future control time domain, and sends the predicted value as a feedforward interference signal to the economical MPC control module.

[0043] S2, the economical MPC control module calls the built-in state observer to update the real-time electricity price of the power grid in the MPC dynamic model matrix based on real-time acquired data.

[0044] S3. Construct an optimization objective function using the MPC dynamic model matrix within the economical MPC control module. Based on the dynamic cost weighting mechanism, dynamically adjust the optimization weight of the electricity cost item within the objective function according to the real-time electricity price of the power grid.

[0045] The economical MPC control module is configured with a dynamic cost-weighted mechanism, and the objective function is calculated as follows:

[0046] In the formula: Let k be the value of the overall objective function for MPC rolling optimization. The real-time power consumption cost of the space separation unit at time k; This is the weighting coefficient for electricity cost items. It can be set as a constant and combined with the electricity price multiplier to achieve peak-valley differentiated adjustment, or dynamically adjusted in real time according to the dispatch strategy. These are controlled variables for product quality, including oxygen purity, nitrogen purity, and pressure at the top of the tower. These are the reference values ​​set for the controlled variables of product quality. This refers to the weighting coefficient for quality tracking error. To control the penalty weighting coefficient for actions; To manipulate the change in the variable; These are traditional quality tracking errors and control action penalty terms, used to constrain parameter fluctuations and ensure stable device operation.

[0047] S4. Use the configured IPOPT or Gurobi solver to perform single-layer rolling optimization on the objective function. The solution process is constrained by the customer's gas demand forecast, product purity, tower pressure, and storage tank level range. During periods of high electricity price, output manipulated variable schemes to reduce air compressor load and reduce liquid production. During periods of low electricity price, output manipulated variable schemes to increase air separation unit load and increase liquid storage production when the storage tank level does not reach the upper limit constraint.

[0048] S5. Send the optimal manipulated variable increment obtained from the solution to the air separation unit DCS system to adjust the raw material air compressor guide vanes, expander nozzles, liquid oxygen extraction valve, and reflux ratio regulating valve to complete the adaptive adjustment of the air separation unit's operating conditions.

[0049] Example The MPC control matrix in this application has been redesigned to directly include economic indicators and eliminate the dependence on the upper-level RTO settings.

[0050] According to Table 1, the formula for calculating the real-time electricity cost of the controlled variable CV1:Elec_Cost is as follows:

[0051] In the formula: Elec_Cost is the real-time electricity cost, and Power(MV) is the real-time power consumption model calculated based on MV manipulated variables (such as compressor guide vane opening). Price refers to the real-time electricity price.

[0052] like Figure 4 As shown, removing the RTO layer makes the MPC algorithm flow more direct and compact: A1. Collect real-time data and update the internal model through the state observer. At the same time, update the current real-time electricity price (DV1).

[0053] A2. The MPC control module constructs the optimization objective function for the current moment. : .

[0054] A3. Single-layer rolling optimization solution: The solver directly solves the above objective function.

[0055] When electricity prices rise, the cost of Elec_Cost in the objective function increases. The solver automatically seeks solutions that reduce MAC_GV (air compressor guide vane opening), and simultaneously closes LOX_Valve (liquid extraction valve) to reduce load and ensure air supply in order to meet O2_Purity / N2_Purity (product purity) and Cust_Flow (customer demand).

[0056] When the electricity price decreases, the cost weight of Elec_Cost in the objective function decreases. If Tank_Level (liquid storage tank level) does not reach the upper limit at this time, the solver can appropriately increase the opening of MAC_GV and LOX_Valve on the premise of satisfying the constraints of purity, pressure and customer demand, so as to increase the liquid production during off-peak hours and replenish the storage tank level, thereby automatically increasing the unit load.

[0057] A4. Directly send the calculated optimal MV increment to DCS.

[0058] As Figure 4 shows, when the electricity price changes from flat period to peak period (14:00): Signal trigger: the value of DV1 (electricity price) jumps, changing from 0.6 yuan to 1.0 yuan.

[0059] MPC calculation: In the objective function, the value of the term increases instantaneously.

[0060] To minimize the total objective , the optimizer tends to reduce power consumption.

[0061] Constraint check: must satisfy (purity) and (customer flow rate) constraints.

[0062] Action execution: MPC issues instructions simultaneously: slowly close (liquid valve) to the minimum flow rate, and synchronously close (air compressor guide vane).

[0063] Due to the model prediction mechanism, MPC can accurately calculate the rate matching for closing the guide vane and the liquid valve to prevent purity fluctuation.

[0064] Result: the unit automatically slides to the minimum energy consumption operating state, without manual intervention, and no need for the upper-level RTO to calculate the setpoint.

[0065] Implementation Example Take a 30000Nm 3 / h oxygen production air separation unit supporting a chemical industrial park as an example. The unit comprises a feed air compressor, a pre-cooling system, a molecular sieve purification system, an expander refrigeration system, a distillation column system, as well as a liquid oxygen storage tank and a liquid nitrogen storage tank. The original DCS system of the unit has collected data including air compressor guide vane opening, expander nozzle opening, liquid oxygen extraction valve opening, reflux ratio regulating valve opening, oxygen purity, nitrogen purity, upper column pressure, customer oxygen flow rate, customer nitrogen flow rate, liquid oxygen storage tank level, liquid nitrogen storage tank level and time-of-use electricity price.

[0066] In this implementation case, the data acquisition module reads real-time operating data through the OPC DA interface at a 1-minute sampling cycle, and stores the historical operating data of the most recent 90 days, customer gas consumption, and time-of-use electricity price data as a historical dataset. The load forecasting module uses historical customer gas consumption, time period, weekday / restday identifier, previous period tank level, and electricity price type as input features to predict the downstream customer's oxygen and nitrogen demand within the next 6 hours, and updates the forecast results every 15 minutes.

[0067] The control cycle of the economical MPC control module is set to 5 minutes, with 24 control steps in the prediction time domain and 8 control steps in the control time domain. Controlled variables include real-time electricity cost, oxygen purity, nitrogen purity, tower pressure, and liquid tank level. Manipulated variables include the opening of the feed air compressor guide vanes, the opening of the expander nozzles, the opening of the liquid oxygen extraction valve, and the opening of the reflux ratio regulating valve. Disturbance variables include real-time electricity price and predicted customer demand. Constraints are set as follows: oxygen purity ≥ 99.6%, oxygen content in nitrogen < 5 ppm, tower pressure < 50 kPa, and liquid tank level maintained between 20% and 90%.

[0068] When the system identifies the period from 23:00 to 07:00 as off-peak electricity pricing, the economical MPC control module reduces the weight of electricity cost items and sets soft constraints on the liquid storage tank level that favor higher levels. When the storage tank level is below 85% and does not trigger purity, pressure, or equipment load constraints, the controller gradually increases the guide vane opening of the raw material air compressor by 2%-6% and appropriately increases the opening of the liquid outlet valve, raising the liquid oxygen and liquid nitrogen storage tank levels from approximately 48% to approximately 76%.

[0069] When the system identifies the period from 14:00 to 17:00 as the peak electricity price period, the economical MPC control module increases the weight of the electricity cost item and uses the customer demand output from the load forecasting module as the feedforward constraint boundary. Under the premise of meeting the customer's lower gas consumption limit and product purity constraints, the controller gradually reduces the opening of the liquid outlet valve to the minimum opening required to maintain safe cooling capacity, and reduces the guide vane opening of the raw material air compressor by approximately 5%, utilizing the liquid reserves formed during the previous off-peak period to buffer peak demand.

[0070] After seven days of continuous operation, the oxygen purity of the unit remained above 99.6%, the oxygen content in the nitrogen was consistently less than 4 ppm, the pressure at the top of the tower did not exceed 46 kPa, and the liquid level in the liquid storage tank remained consistently between 32% and 84%. Compared with manual fixed-load operation, the average power consumption during peak hours decreased by approximately 6.8%, the liquid production during off-peak hours increased by approximately 9.5%, and the average daily electricity cost decreased by approximately 7.6%, without any issues of insufficient gas supply to customers or product quality exceeding limits. Therefore, this invention can achieve economical variable-load scheduling of air separation units by utilizing load forecasting and economical MPC single-layer rolling optimization, even without the upper-level RTO.

[0071] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A variable load dispatching system based on load forecasting and economic model prediction, characterized in that, It includes a data acquisition module, a load forecasting module, and an economical MPC control module; The data acquisition module is used to collect in real time the operating parameters of the air separation unit, the real-time gas consumption of downstream customers, the liquid level data of the liquid storage tank, and the real-time time-of-use electricity price data of the power grid, and to store the collected data information as a historical dataset. The load forecasting module retrieves historical air separation unit operating parameters, historical downstream customer gas consumption, historical liquid storage tank level, and historical grid time-of-use electricity price from the historical dataset as input features for the forecasting model, forecasts the downstream customer gas demand in the future control time domain, and sends the forecast result as a feedforward interference signal to the economical MPC control module. The economical MPC control module employs an economic model predictive control algorithm. It constructs an MPC dynamic model matrix using the collected operating parameters of the air separation unit, and introduces real-time electricity cost or total power consumption of the air separation unit as new controlled variables into the MPC dynamic model matrix. The economical MPC control module receives the predicted future downstream customer gas demand from the load forecasting module as a measurable feedforward constraint boundary. During the rolling optimization process, it constructs an optimization objective function with the goal of minimizing electricity cost and ensuring product purity is within the constraints. It automatically calculates the optimal manipulator trajectory for adjusting the operation of the air separation unit, thereby achieving adaptive economic dispatch of the air separation unit to adapt to the peak and valley electricity prices of the power grid.

2. The variable load dispatching system based on load forecasting and economic model prediction according to claim 1, characterized in that, The economical MPC control module has a built-in state observer, which is used to update the real-time electricity cost of the MPC dynamic model matrix based on the data collected in real time by the data acquisition module.

3. The variable load dispatching system based on load forecasting and economic model prediction according to claim 1, characterized in that, The economical MPC control module is configured with a dynamic cost-weighted mechanism: During periods of high electricity prices, the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix is ​​increased. Under the premise of meeting the lower limit of downstream gas supply, product purity and equipment safety constraints, the controller automatically searches for the lowest energy consumption condition of the air separation unit, thereby reducing the operating load and liquid output of the air separation unit. During periods of low electricity prices, reduce the weight of electricity cost or total power consumption of the air separation unit in the MPC dynamic model matrix, and add soft constraints on liquid storage tank levels that favor high liquid levels. The economical MPC control module utilizes low-cost electricity during off-peak hours to increase the operating load of the air separation unit and increase the production of liquid products until the equipment operating constraints or the liquid storage tank level limit are reached.

4. The variable load dispatching system based on load forecasting and economic model forecasting according to claim 1, characterized in that, The MPC dynamic model matrix of the economical MPC control module includes the controlled variable CV, the manipulated variable MV, and the disturbance variable DV. The controlled variable CV includes real-time electricity cost, product oxygen purity, product nitrogen purity, pressure at the air separation unit, and liquid storage tank level; the real-time electricity cost is the minimization optimization objective, the product oxygen purity, product nitrogen purity, and pressure at the air separation unit are hard constraint indicators, and the liquid storage tank level is an interval control energy storage buffer indicator. The manipulated variables include the opening degree of the guide vane of the raw material air compressor, the opening degree of the expander nozzle, the opening degree of the liquid oxygen take-off valve, and the opening degree of the distillation reflux ratio regulating valve, which are used to regulate the energy consumption of the unit's air intake, the cooling capacity of the system, the liquid output and the distillation conditions, respectively. The interference variables include the real-time electricity price of the power grid and the gas demand of downstream customers. The real-time electricity price of the power grid is an interference variable that is updated and optimized in real time, and the gas demand of downstream customers is a feedforward interference signal output by the load forecasting module.

5. The variable load dispatching system based on load forecasting and economic model prediction according to claim 4, characterized in that, The specific constraints on the controlled variables are as follows: oxygen purity is ≥ 99.6%, nitrogen purity is less than 5 ppm, the pressure at the top of the tower is less than 50 kPa, and the liquid level in the liquid storage tank is 20%-90%. The specific adjustment range of the manipulated variables is as follows: guide vane opening of the raw material air compressor 0-100%, nozzle opening of the expander 0-100%, liquid oxygen take-off valve opening 0-100%, and reflux ratio regulating valve opening 0-100%.

6. The variable load dispatching system based on load forecasting and economic model forecasting according to claim 1, characterized in that, The system adopts a single-layer MPC control architecture that eliminates the RTO layer and is equipped with a comprehensive control server. The comprehensive control server carries and independently runs the economical MPC control module. The server is configured with a DCS interface and communicates with the air separation unit's DCS system through the OPC DA protocol.

7. The variable load dispatching system based on load forecasting and economic model prediction according to claim 1, characterized in that, The system is configured with a solver to support the operation of the economical MPC control module. The solver is selected from IPOPT or Gurobi and supports quadratic programming and nonlinear programming solutions.

8. A method for implementing a variable load dispatching system based on load forecasting and economic model forecasting as described in any one of claims 1-7, characterized in that, Includes the following steps: S1. The data acquisition module collects real-time operating data of the air separation unit, real-time gas consumption of downstream customers, liquid storage tank level data, and real-time time-of-use electricity price data of the power grid and stores them to form a historical dataset. The load forecasting module retrieves multi-dimensional historical feature data from the historical dataset, predicts the gas demand of downstream customers in the future control time domain, and sends the predicted value as a feedforward interference signal to the economical MPC control module. S2, the economical MPC control module calls the built-in state observer to update the real-time electricity price of the power grid in the MPC dynamic model matrix based on real-time acquired data; S3. Construct an optimization objective function using the MPC dynamic model matrix in the economical MPC control module. Based on the dynamic cost weighting mechanism, dynamically adjust the optimization weight of the electricity cost item in the objective function according to the real-time electricity price of the power grid. S4. Use the configured IPOPT or Gurobi solver to perform single-layer rolling optimization on the objective function. The solution process is constrained by the customer's gas demand forecast, product purity, tower pressure, and storage tank level range. During periods of high electricity price, output manipulated variable schemes to reduce air compressor load and reduce liquid production. During periods of low electricity price, output manipulated variable schemes to increase air separation unit load and increase liquid storage production when the storage tank level does not reach the upper limit constraint. S5. Send the optimal manipulated variable increment obtained from the solution to the air separation unit DCS system to adjust the raw material air compressor guide vanes, expander nozzles, liquid oxygen extraction valve, and reflux ratio regulating valve to complete the adaptive adjustment of the air separation unit's operating conditions.

9. The variable load dispatching method based on load forecasting and economic model forecasting according to claim 8, characterized in that, The economical MPC control module is configured with a dynamic cost-weighted mechanism, and the objective function is calculated as follows: In the formula: Let k be the value of the overall objective function for MPC rolling optimization. The real-time power consumption cost of the space separation unit at time k; This is the weighting coefficient for electricity cost items. It can be set as a constant and combined with the electricity price multiplier to achieve peak-valley differentiated adjustment, or dynamically adjusted in real time according to the dispatch strategy. These are controlled variables for product quality, including oxygen purity, nitrogen purity, and pressure at the top of the tower. These are the reference values ​​set for the controlled variables of product quality. This refers to the weighting coefficient for quality tracking error. To control the penalty weighting coefficient for actions; To manipulate the change in the variable; These are traditional quality tracking errors and control action penalty terms, used to constrain parameter fluctuations and ensure stable device operation.

Images