Unit multi-objective real-time optimization decision method for spot market

Abstract

This invention discloses a multi-objective real-time optimization decision-making method and system for coal-fired power units oriented towards the electricity spot market. The method includes: real-time acquisition of multi-source data; construction of a refined cost-benefit model encompassing multi-fuel coupled costs, full-condition environmental costs, and net fly ash revenue coupled with fly ash quality qualification rate prediction; establishment of a multi-objective optimization model with the objective of maximizing total net revenue in the rolling time domain, considering constraints such as unit output, ramp-up, biomass co-firing ratio, environmental emissions, and fly ash quality; solution of the model using a dynamic feasible region shrinking particle swarm optimization hybrid algorithm, achieving second-level solution by pre-emptively narrowing the search space through hard constraints and handling soft constraints with an adaptive penalty function; and issuance of commands according to the rolling time domain optimization mechanism to form closed-loop control. This invention incorporates fly ash quality and biomass co-firing constraints into the total cost optimization, balancing safety, environmental protection, and economy, significantly improving the profitability and response speed of the unit in the spot market.

Description

Technical Field

[0001] This invention belongs to the field of power system automation and power market trading technology, specifically relating to a multi-objective real-time optimization decision-making method and system for coal-fired power units oriented towards the power spot market, and particularly to a real-time optimization decision-making technology for coal-fired power units that considers fly ash quality constraints and biomass co-firing characteristics. Background Technology

[0002] In the electricity spot market model, the power generation plan and settlement price of generating units are determined by the market clearing results. This fundamentally changes the profit model of coal-fired units, unlike the traditional planned dispatch model. Units need to quickly make decisions on power output and generation costs within a short trading cycle of 15 minutes, in response to frequently fluctuating market electricity prices.

[0003] The existing technology has the following three shortcomings in this application scenario: 1. Delayed decision-making response, unable to adapt to high-frequency market fluctuations. Traditional planning and dispatching systems generate power generation plans based on day-ahead or rolling pre-dispatch mechanisms, which cannot respond to the 15-minute-level electricity price fluctuations in the spot market at a high frequency; on-site operators rely on manual experience to make decisions, resulting in slow response speed, high workload, and difficulty in capturing short-term market arbitrage opportunities.

[0004] 2. The problem of strong coupling of multiple variables cannot achieve real-time and accurate trade-offs. Human decision-making cannot simultaneously and accurately quantify and trade off multiple strong coupling and nonlinear variables such as electricity price signals, unit efficiency, environmental protection costs, coal inventory, and by-product revenue. Especially under the deep peak shaving conditions of the unit, the coal consumption characteristics of the unit show a nonlinear and steep increase trend, the commissioning cost of environmental protection facilities increases dramatically, and manual estimation is prone to causing the unit to operate at a loss.

[0005] 3. The power generation cost-benefit model lacks refinement, resulting in distorted calculation results. Existing power generation optimization models typically treat coal as a single homogeneous fuel in their cost models, neglecting several key influencing factors, leading to distorted cost calculations and one-sided benefit assessments. The constraints of fly ash quality were not considered. Fly ash, as an important by-product of coal-fired power units, can be sold for profit, but its quality (such as water demand ratio and loss on ignition) is directly affected by boiler operating conditions (combustion temperature, flue gas oxygen content) and coal quality parameters. If low-load operation is pursued unilaterally or low-cost, low-quality coal is blended in, the quality of fly ash may deteriorate, turning it from a profitable by-product into solid waste that requires disposal costs, posing a significant economic risk.

[0006] The complex impacts of biomass co-firing are not quantified. Coal-fired units can obtain additional subsidy benefits and carbon emission rights benefits by co-firing biomass. However, the alkali metal content and calorific value of different types of biomass vary significantly. The biomass co-firing ratio is not only limited by fuel supply capacity, but also by boiler coking, the risk of corrosion of heating surfaces, and environmental constraints on flue gas emissions. Existing models cannot quantify the coupling relationship between these constraints and benefits.

[0007] The existing models fail to reflect the piecewise nonlinear characteristics of environmental protection costs. The material and electrical costs of desulfurization and denitrification systems exhibit strong nonlinear characteristics at different load levels and are strictly constrained by emission concentration limits. The existing models cannot accurately match the changing patterns of environmental protection costs under all operating conditions.

[0008] Therefore, there is an urgent need for a high-frequency real-time optimization decision-making method that can comprehensively consider market signals, the full operating characteristics of the unit, environmental costs, and the quality constraints of fly ash by-products, so as to maximize the revenue and ensure safe and stable operation of coal-fired units in the electricity spot market environment. Summary of the Invention

[0009] 1. Technical problems to be solved In view of the above-mentioned deficiencies in the prior art, the technical problem to be solved by the present invention is as follows: (1) To solve the problem that the existing technology has a slow decision response and cannot meet the real-time decision-making needs of 15-minute electricity price fluctuations in the spot electricity market, and to improve the unit’s response speed to market signals.

[0010] (2) To solve the problem that existing technologies cannot accurately quantify and weigh strongly coupled nonlinear variables such as electricity price, unit efficiency, environmental protection costs, and by-product revenue in real time, and that loss-making operation is likely to occur under deep peak shaving conditions.

[0011] (3) To address the problem that the existing power generation optimization model is not refined enough, and does not take into account the constraints of fly ash quality, the multi-dimensional impact of biomass co-firing, and the piecewise nonlinear characteristics of environmental protection costs, which leads to distorted cost calculation and one-sided benefit assessment.

[0012] 2. Technical Solution This invention adopts a three-layer architecture: perception layer, decision layer, and execution layer. The perception layer, corresponding to step S1, is responsible for the real-time acquisition of multi-source heterogeneous data. The decision layer, corresponding to steps S2 to S4, incorporates a refined full-cost-benefit model and a multi-objective rolling optimization solver, serving as the core of the system. The execution layer, corresponding to step S5, automatically sends the optimization results to the unit control system to form a closed-loop control. Through a rolling time-domain optimization mechanism, the system repeatedly executes data acquisition, model updates, optimization solutions, and command issuance in each decision cycle, achieving multi-objective real-time optimization decision-making for coal-fired power units oriented towards the electricity spot market.

[0013] The core technology transmission logic is as follows: The two decision variables, unit output and biomass co-firing ratio, jointly determine the boiler combustion conditions (furnace temperature, flue gas oxygen content, pollutant generation, and ash production), which in turn affect fuel costs, environmental costs, fly ash revenue, and various safety / environmental / quality constraints. This invention integrates multi-dimensional constraints and revenue items into a rolling optimization framework and achieves second-level solutions through improved algorithms.

[0014] (1) Multi-objective real-time optimization decision-making method The method specifically includes the following steps S1 to S5.

[0015] Step S1: Real-time acquisition and status awareness of multi-source data By collecting multi-source heterogeneous data in real time through multiple interfaces, a comprehensive input foundation is provided for optimization decisions. The duration of a single decision cycle is defined as... (For example, 15 minutes), the rolling time domain duration is denoted as (For example, 4 hours), then the total number of time periods within the rolling time domain. ; The time period number, The data collected specifically includes: S1.1 Market and Policy Data: Obtaining real-time electricity price forecast sequences for the future rolling time domain from power trading centers and carbon trading markets. Deep peak shaving compensation price Ancillary service winning bid price and corresponding winning bid capacity Biomass power generation subsidy unit price Carbon trading prices Simultaneously, sulfur dioxide emission concentration limits are obtained from environmental data platforms or unit monitoring systems. and nitrogen oxide emission concentration limits .

[0016] S1.2 Unit operating data: Real-time acquisition of unit power generation from the unit's distributed control system (DCS). Main steam parameters used to correct the thermal efficiency of steam turbines Boiler furnace temperature Oxygen content in flue gas Inlet and outlet concentrations of the desulfurization system and Inlet and outlet concentrations of the denitrification system and Simultaneously, obtain the unit's rated maximum output from the DCS or equipment specifications. Minimum technical output Maximum load increase rate within a single decision cycle and maximum load reduction rate .

[0017] S1.3 Fuel Inventory and Characteristic Data: Obtain industrial analysis data on the current coal type from the plant-level management information system, including calorific value. Ash content , sulfur content Unit calorific value coal cost (Cost per ton of coal delivered to the plant divided by) (Calculated); simultaneously obtain biomass fuel reserves and biomass calorific value. Alkali metal content in biomass that affects ash melting characteristics and Unit calorific value biomass cost (Cost per ton of biomass delivered to the plant divided by) (Calculated).

[0018] S1.4 By-product Market Data: Real-time collection of fly ash market prices With solid waste disposal costs .

[0019] Step S2: Construct a refined full-cost-benefit model. This involves building a cost-benefit function for the unit under all operating conditions to accurately calculate the real-time net operating benefit under any combination of decision variables. The following sub-models all use the decision cycle... Output the cumulative cost or revenue for each time period, using a time step; unit output. Proportion of biomass co-firing The core input variables are as follows: ① The multi-fuel coupling cost model quantifies the total fuel cost under different outputs and different biomass blending ratios. The calculation formula is as follows:

[0020] In the formula: for Total fuel cost for the period; Boiler efficiency is a nonlinear function of unit load rate; The thermal efficiency of a steam turbine generator set is inversely proportional to the steam turbine heat rate and can be determined by real-time acquisition of main steam parameters. Online correction; The biomass blending ratio refers to the proportion of biomass heat to the total heat of the fuel fed into the furnace. and These are the coal cost and biomass cost converted to unit calorific value, respectively, and can be directly weighted by calorific value.

[0021] ② All-condition environmental protection cost model A mechanism-data hybrid modeling approach is used to quantify the operating costs of desulfurization and denitrification facilities under different operating conditions in order to meet emission limits. The calculation formula is as follows:

[0022] In the formula: for Total environmental protection cost for all operating conditions during the same period; The desulfurization cost function takes the unit output, biomass co-firing ratio, real-time collected desulfurization inlet concentration, and sulfur dioxide emission limit as inputs, and outputs the total material and electricity consumption cost of the desulfurization system. The desulfurization inlet concentration is determined by the sulfur content of the coal. The biomass co-firing ratio and unit load are calculated through a mechanistic model, and the inlet and outlet concentrations collected in real time are used for online parameter calibration of the mixing model. The denitrification cost function takes the unit output, biomass co-firing ratio, real-time collected denitrification inlet concentration, and nitrogen oxide emission limit as inputs, and outputs the total material and power consumption cost of the denitrification system.

[0023] ③ Net Profit Model of Fly Ash The calculation of fly ash production is coupled with the prediction of fly ash quality qualification rate, quantifying the sales revenue of qualified fly ash and the disposal cost of unqualified fly ash. The calculation formula is as follows:

[0024] In the formula: for Net revenue from fly ash during the period; for The time-period fly ash production model takes the unit load, biomass co-firing ratio, and coal ash content as inputs. The output is the total ash mass produced in a single time period (includes time integral). The fly ash qualification rate model is one of the core innovations of this invention. It is used to quantitatively describe the dynamic changes in fly ash quality (loss on ignition, etc.) under combustion conditions and fuel characteristics. The calculation formula is as follows:

[0025] Among them, furnace temperature and oxygen content in flue gas Power output from the generator unit Proportion of biomass co-firing The alkali metal content in biomass was calculated based on the boiler combustion mechanism. , By altering the ash fusion characteristics to affect the loss on ignition of fly ash, the results are incorporated into the qualification rate calculation. When the predicted fly ash quality is substandard, the corresponding portion of the revenue turns negative, i.e., the cost of solid waste disposal. This refers to the market price of fly ash. Costs associated with solid waste disposal.

[0026] Step S3: Construct a multi-objective rolling optimization mathematical model Based on the full cost-benefit model constructed in step S2, in each decision cycle In the future, a rolling time domain (common To maximize the total net system revenue (across several time periods), a multi-objective rolling optimization mathematical model is constructed, incorporating multi-dimensional constraints. The decision variable is the unit output in each time period. Biomass co-firing ratio .

[0027] ① Objective function The optimization objective is to maximize the total net system revenue over the rolling time domain. The objective function is as follows:

[0028] In the formula: The total net system revenue over the rolling time domain; This represents the total number of time periods within the rolling time domain. ; for The formula for calculating the revenue from power generation and grid connection during a given period is as follows:

[0029] in for Real-time electricity price forecast for the specified time period; for The revenue from ancillary services and subsidies during the specified period includes revenue from winning bids for ancillary services, revenue from deep peak shaving compensation, revenue from biomass subsidies, and revenue from carbon emission reduction. The calculation formula is as follows:

[0030] in: for The winning bid price for ancillary services during the specified time period. The corresponding ancillary service capacity won in the bid; For the nonlinear compensation function of deep peak shaving, based on the deep peak shaving compensation price The compensation amount is calculated based on the peak shaving depth of the generating unit, and increases as the peak shaving depth of the generating unit increases. Subsidize the unit price of biomass power generation; For carbon trading prices; The carbon emission reduction per unit time generated by co-firing biomass to replace coal is related to the biomass co-firing ratio. Positive correlation, calculated based on the amount of coal replaced by biomass.

[0031] ②Constraints The optimization model is divided into two types of boundaries: hard constraints and soft constraints. Hard constraints are the physical boundaries of the unit and the constraints on equipment capacity, which must be strictly satisfied. Soft constraints are operational compliance constraints, which are guided to convergence through an algorithmic penalty function. Details are as follows: Unit output constraints (C1, hard constraints): ; Unit ramp-up constraint (C2, hard constraint): ; Biomass co-firing ratio constraints (C3, hard constraint): ,in The real-time upper limit of the biomass co-firing ratio is dynamically limited by the fuel delivery equipment capacity, boiler safety assessment results, and fuel inventory levels. The boiler safety assessment is based on the boiler coking trend and the heat transfer surface corrosion rate model. Sulfur dioxide emission concentration constraints (C4, soft constraint): ; Nitrogen oxide emission concentration constraints (C5, soft constraint): ; Fly ash quality constraints (C6, soft constraints): ,in This is the minimum acceptable threshold for fly ash qualification rate, used to avoid generating a large amount of substandard solid waste.

[0032] Step S4: Solving the model based on the hybrid algorithm of dynamic feasible region shrinkage and particle swarm optimization. To address the aforementioned complex optimization problem involving multiple time periods, nonlinearity, and strong coupling of continuous variables, a hybrid solution strategy of "dynamic feasible region shrinkage-particle swarm optimization" is adopted: first, all hard constraints are preprocessed through feasible region preprocessing to reduce the search space; then, soft constraints are handled through an adaptive penalty function to guide algorithm convergence, ultimately achieving a solution in seconds. The specific steps are as follows: S4.1 Initialization Settings Set particle swarm size Maximum number of iterations ; For particle number ( ), the position of each particle Represents a decision-making scheme, generated within the rolling time domain. The decision variables for each time period consist of:

[0033] Randomly initialize the positions and velocities of all particles in the search space. .

[0034] S4.2 Feasibility Region Preprocessing (Hard Constraint Preprocessing) For each time period Based on the output of the previous period With the slope constraint C2, dynamically calculate the absolute feasible region of the output in the current time period. ; Based on real-time fuel data, inventory levels, and boiler safety models, the real-time upper limit of biomass blending ratio is dynamically updated. ; By strictly limiting the solution space of particles to the aforementioned dynamically shrinking feasible region, and performing boundary absorption or random reset processing on the positions of particles that exceed the boundary, the optimization efficiency is greatly improved.

[0035] S4.3 Constructing the fitness function Objective function By combining soft constraints (C4-C6) with the penalty function method, an unconstrained fitness function is constructed. :

[0036] In the formula: For particles The fitness value indicates that the larger the value, the better the solution. This is a function for calculating the out-of-bounds amount. When the constraint is not exceeded, the out-of-bounds amount is 0. When the constraint exceeds the bounds, the out-of-bounds magnitude is output. , , The adaptive penalty factor increases exponentially with the number of iterations, guiding the algorithm to gradually converge to the optimal solution within the feasible region.

[0037] S4.4 Particle Position and Velocity Update The particles are updated according to the standard update formula of the particle swarm optimization algorithm, and a strategy of linearly decreasing inertia weight and asynchronous adjustment of learning factor is introduced to balance the algorithm's global search and local exploitation capabilities; the update formula is as follows:

[0038] In the formula: This represents the current iteration number; For the first The inertial weight of the generation; , For learning factors; , A random number whose value is between 0 and 1; For particles The individual's historical optimal position; This represents the globally optimal position for the population.

[0039] S4.5 Iteration and Termination Calculate the fitness of each particle and update its historical best. and global optimal Repeat the particle update step until the maximum number of iterations is reached or the global optimal fitness value shows no significant improvement for several consecutive generations, then output the optimal solution sequence. .

[0040] Step S5: Command Issuance and Rolling Closed-Loop Control A typical mechanism for rolling temporal optimization is that each optimization covers the future. Of the time periods, only the first time period obtained from optimization is taken. ) decision value , As the currently executed command; load commands are transmitted via the plant-level Automatic Generation Control (AGC) interface. The signal for biomass co-firing ratio is sent to the unit's DCS. Issued to the fuel delivery control system.

[0041] In the next decision cycle, the system rolls forward one time period, updates all real-time data, and repeats steps S1 to S5 to achieve rolling closed-loop optimization and continuously respond to fluctuations in spot market electricity prices.

[0042] (2) Multi-objective real-time optimization decision system Corresponding to the above method, the system adopts a three-layer architecture of "perception layer - decision layer - execution layer", including: Perception layer unit: used to perform multi-source data real-time acquisition and status perception function in step S1, and to acquire market and policy data, unit operation data, fuel inventory and characteristic data, and by-product market data in real time; Decision layer unit: used to perform the functions of steps S2 to S4, with built-in refined full cost-benefit model and multi-objective rolling optimization solver, to complete sub-model construction, optimization model building and intelligent algorithm solution, and output the optimal decision sequence; Execution layer unit: Used to execute the instruction issuance and rolling closed-loop control function of step S5, issue the optimal decision instruction to the unit control system, and trigger the optimization process of the next decision cycle to form a rolling closed loop.

[0043] 3. Beneficial effects Compared with the prior art, the present invention has the following beneficial effects: (1) Significantly improve the profitability of the unit in the spot market. Through a refined full cost-benefit model, especially the dynamic prediction of fly ash qualification rate, the risk of "profit from power generation but loss from ash disposal" is avoided, and the optimal operating window in the high / low electricity price range can be accurately captured, which is expected to improve the overall economic benefits by 2%-5%.

[0044] (2) Achieving the optimal balance between environmental compliance and economic operation. The hybrid model of environmental cost mechanism data under all operating conditions transforms the environmental protection strategy from "extensive compliance" to "economic precision compliance", significantly reducing environmental operation costs while strictly adhering to emission red lines.

[0045] (3) Resolve the contradiction between safety and economy in biomass co-firing. Quantify the risks of coking and corrosion and inventory constraints into a dynamic upper limit for the proportion of biomass co-firing, so as to maximize the benefits of subsidies and carbon emission reduction while ensuring safety.

[0046] (4) Achieve second-level market response. The hybrid solution strategy of dynamic feasible region shrinkage and particle swarm optimization reduces the search space by hard constraints and guides convergence with an adaptive penalty function, thereby compressing the solution time to the second level, enabling the unit to have a real-time response capability to the spot market at the 15-minute level. Attached Figure Description

[0047] Figure 1 This is a flowchart of the method of the present invention. Figure 2 This is a system flowchart of the present invention. Detailed Implementation

[0048] The technical solution of the present invention will be clearly and completely described below with reference to specific embodiments. The described embodiments are typical implementations of the present invention. Based on the technical solution disclosed in the present invention, those skilled in the art can reproduce the entire technical process without creative effort. Therefore, the present invention has practicality.

[0049] This embodiment uses a 300MW subcritical coal-fired power generating unit participating in provincial electricity spot market transactions as the application object. This unit is equipped with a limestone-gypsum wet desulfurization system and a selective catalytic reduction denitrification system, and has the capability for biomass co-firing. Fly ash is sold as a by-product. This invention employs a three-layer architecture of "perception layer - decision layer - execution layer" to conduct real-time optimization decision-making, setting a single decision cycle. The duration is 15 minutes, with a rolling time domain. The duration is 4 hours, corresponding to the total number of time periods within the rolling time domain. Time period number The specific implementation process of the multi-objective real-time optimization decision-making method is as follows: Step S1: Real-time acquisition and status awareness of multi-source data The perception layer collects four types of multi-source heterogeneous data in real time through multiple communication interfaces to achieve a comprehensive perception of the system's operating status and the external market status, providing multi-dimensional input parameters for optimization decisions. The data collection content is as follows: 1. Market and Policy Data The real-time electricity price forecast sequence for 16 time periods within the future rolling time domain is obtained through the dedicated data interface of the power trading center. Deep peak shaving compensation price Ancillary service winning bid price and corresponding winning bid capacity Obtain carbon trading prices through carbon trading market data interfaces. Obtain the unit price of biomass power generation subsidies from the data interface of the energy authority. Simultaneously, sulfur dioxide emission concentration limits are obtained from the environmental monitoring data platform. and nitrogen oxide emission concentration limits .

[0050] 2. Unit operating data Real-time acquisition of current unit power generation through the standard interface of the unit's distributed control system (DCS). Main steam parameters used to correct the thermal efficiency of steam turbines (Including main steam pressure and main steam temperature), boiler furnace temperature Oxygen content in flue gas Inlet and outlet concentrations of the desulfurization system and Inlet and outlet concentrations of the denitrification system and Simultaneously, obtain the unit's rated maximum output from the DCS or equipment specifications. Minimum technical output Maximum load increase rate within a single decision cycle and maximum load reduction rate .

[0051] 3. Fuel inventory and characteristic data Obtain industrial analysis data on the current coal type from the plant-level Management Information System (MIS), including calorific value. Ash content , sulfur content Unit calorific value coal cost (Cost per ton of coal delivered to the plant divided by) (Calculated); simultaneously obtain biomass fuel reserves and biomass calorific value. Alkali metal content in biomass that affects ash melting characteristics and Unit calorific value biomass cost (Cost per ton of biomass delivered to the plant divided by) (Calculated).

[0052] 4. By-product market data Real-time data collection of current fly ash market prices through the enterprise management system. With solid waste disposal costs .

[0053] Step S2: Construct a refined full-cost-benefit model Based on real-time collected data, the decision-making team constructs a full-condition cost-benefit function covering fuel costs, environmental costs, and fly ash revenue, with a 15-minute decision cycle. Using time steps, the unit output is calculated for each time period. Proportion of biomass co-firing As the core input variable, the specific implementations of each sub-model are as follows: 1. Multi-fuel coupling cost model The formula for calculating the total fuel cost under different unit outputs and different biomass blending ratios is as follows:

[0054] Among them, boiler efficiency The load factor is a nonlinear function of the unit load rate, obtained by fitting the full-condition coal consumption characteristic curve based on boiler thermal performance tests. It exhibits a nonlinear decreasing trend during the deep peak-shaving low-load segment; the thermal efficiency of the steam turbine generator unit... It is inversely proportional to the turbine heat rate, and is obtained through real-time acquisition of main steam parameters. Online correction of design values ​​ensures the accuracy of thermal efficiency calculations under different operating conditions. Biomass co-firing ratio. The proportion of biomass heat to the total heat of the fuel fed into the furnace is determined by the unit calorific value cost. , Heat-weighted calculations are performed to achieve coupled calculations of coal and biomass costs.

[0055] 2. Environmental Cost Model under All Operating Conditions Using a mechanism-data hybrid modeling approach, desulfurization and denitrification cost functions are constructed to quantify the operating costs of environmental protection facilities under all operating conditions. The formula for calculating the total environmental protection cost is as follows:

[0056] Among them, the desulfurization cost function Based on unit output, biomass co-firing ratio, and desulfurization inlet Concentration and emission limits are inputs; the output is the total cost of absorbent consumption and plant power consumption in the desulfurization system; denitrification cost function. Based on unit output, biomass co-firing ratio, and denitrification inlet The concentration and emission limits are inputs, and the output is the total cost of reducing agent consumption and plant power consumption in the denitrification system. In this model, the pollutant concentrations at the desulfurization and denitrification inlets are first determined by the sulfur content of the coal combustion. The mechanism model of biomass co-firing ratio and unit load is calculated, and the model parameters are calibrated online using real-time collected inlet and outlet concentration data. This ensures the generalization ability under all operating conditions and achieves accurate adaptation to the current operating conditions. It can accurately reflect the piecewise nonlinear characteristics of environmental protection costs under different load segments and solves the problem of distortion in the calculation of environmental protection costs in existing models.

[0057] 3. Net Profit Model for Fly Ash The calculation of fly ash production is coupled with the prediction of fly ash quality qualification rate, quantifying the sales revenue of qualified fly ash and the disposal cost of unqualified fly ash. The calculation formula is as follows:

[0058] Among them, the ash production model The total ash production mass per time period is calculated based on the fuel consumption corresponding to the unit load, the ash content of the coal, and the ash blending characteristics of biomass, including time integration calculations. Fly ash qualification rate model. One of the core innovations of this invention is to quantitatively describe the dynamic changes in quality indicators such as fly ash loss on ignition and water demand ratio under combustion conditions and fuel characteristics, and its expression is as follows:

[0059] In the formula, the furnace temperature is... and oxygen content in flue gas Power output from the generator unit Proportion of biomass co-firing Calculations based on the boiler combustion mechanism model show that: when the unit load increases, the furnace heat load increases and the furnace temperature rises; when the excess air coefficient increases under low load, the oxygen content in the flue gas increases; and the alkali metal content in biomass... , By altering the ash fusion characteristics and combustion reaction process to influence the loss on ignition of fly ash, the results are incorporated into a qualification rate model for quantitative calculation. When the predicted fly ash quality falls below the qualification standard, the corresponding ash production revenue becomes negative, representing solid waste disposal costs. This achieves full quantification of by-product revenue and risk, avoiding the economic risk of "profit from power generation but loss from ash disposal."

[0060] Step S3: Construct a multi-objective rolling optimization mathematical model In each decision cycle, the optimization objective is to maximize the total net system revenue over the next 4 hours in the rolling time domain. A multi-objective rolling optimization mathematical model is constructed, incorporating multi-dimensional constraints. The decision variables are the unit output over 16 time periods. Proportion of biomass co-firing .

[0061] 1. Objective function The optimization objective is to maximize the total net system revenue over the rolling time domain. The objective function is as follows:

[0062] in: Revenue from power generation and grid connection It is calculated from the real-time electricity price and the power generation output; Ancillary services and subsidy revenue It comprises four parts: revenue from winning bids for ancillary services, revenue from deep peak shaving compensation, revenue from biomass power generation subsidies, and revenue from carbon emission reduction. The calculation formula is as follows:

[0063] In the formula, For the nonlinear compensation function of deep peak shaving, based on the deep peak shaving compensation price The compensation amount is calculated based on the peak shaving depth of the generating unit, and increases as the peak shaving depth of the generating unit increases. The carbon emission reduction per unit time generated by co-firing biomass to replace coal is related to the biomass co-firing ratio. Positive correlation, calculated based on the amount of coal replaced by biomass.

[0064] 2. Constraints The optimization model sets two types of boundaries: hard constraints and soft constraints. Hard constraints are the physical boundaries of the unit and equipment capabilities, which must be strictly satisfied. Soft constraints are operational compliance constraints, which are guided to convergence through an algorithmic penalty function. Unit output constraints (C1, hard constraints): The unit output is limited to within the technically feasible range; Unit ramp-up constraint (C2, hard constraint): Ensure that the rate of change of unit load is within the allowable range of the equipment; Biomass co-firing ratio constraints (C3, hard constraint): ,in The real-time upper limit of the biomass co-firing ratio is dynamically determined by the fuel conveying equipment capacity, the safety assessment results of boiler coking and heating surface corrosion, and the fuel inventory level. This quantifies safety risks into a constraint boundary and solves the problem of balancing the safety and economy of biomass co-firing. Sulfur dioxide emission concentration constraints (C4, soft constraint): ; Nitrogen oxide emission concentration constraints (C5, soft constraint): ; Fly ash quality constraints (C6, soft constraints): ,in This is the minimum acceptable threshold for fly ash qualification rate, used to avoid additional disposal costs due to a large amount of unqualified fly ash caused by improper operating conditions.

[0065] Step S4: Solving the model based on a hybrid algorithm of dynamic feasible region shrinkage and particle swarm optimization. To address the aforementioned complex optimization problem involving multiple time periods, nonlinearity, and strong coupling of continuous variables, a hybrid solution strategy of "dynamic feasible region shrinkage-particle swarm optimization" is adopted. This strategy achieves sub-second-level solutions by pre-emptively shrinking the search space through hard constraints and guiding convergence with an adaptive penalty function. The specific implementation steps are as follows: 1. Initialization settings Set particle swarm size Maximum number of iterations The position of each particle This represents a complete decision-making scheme, consisting of decision variables for 16 time periods within the rolling time domain, namely:

[0066] The particle position vector has a dimension of 32. The positions and velocities of all particles are randomly initialized in the initial search space. .

[0067] 2. Feasible region preprocessing (preprocessing of hard constraints) For each time period Based on the output of the previous period With the slope constraint C2, dynamically calculate the absolute feasible region of the output in the current time period. Based on real-time fuel data, inventory levels, and boiler safety models, the real-time upper limit of biomass blending ratio is dynamically updated. .

[0068] By strictly limiting the solution space of all particles to the aforementioned dynamically shrunk feasible region, and performing boundary absorption processing on the positions of out-of-bounds particles, the out-of-bounds dimension is directly reset to the corresponding boundary value, which greatly reduces the invalid search space and improves the optimization efficiency from the root.

[0069] 3. Construct the fitness function The objective function that maximizes total net profit is combined with soft constraints C4-C6 using an adaptive penalty function method to construct an unconstrained fitness function. :

[0070] in, This is a function for calculating the out-of-bounds amount. When the constraint is not exceeded, the out-of-bounds amount is 0. When the constraint exceeds the bounds, the out-of-bounds magnitude is output. , , The penalty factor is an adaptive factor that increases exponentially with the number of iterations: the penalty factor is small in the early stage of iteration, allowing particles to search in the infeasible region to expand the global exploration range; the penalty factor gradually increases in the later stage of iteration, guiding the algorithm to converge to the optimal solution in the feasible region.

[0071] 4. Particle position and velocity update The particle velocity and position are updated according to the standard update formula of the particle swarm optimization algorithm. Simultaneously, a strategy of linearly decreasing inertia weight and asynchronous adjustment of the learning factor is introduced to balance the algorithm's global search and local exploitation capabilities. The update formula is as follows:

[0072] In the formula, This represents the current iteration number; For the first The inertial weight of the generation; , For learning factors; , A random number whose value is between 0 and 1; For particles The individual's historical optimal position; This represents the globally optimal position for the population.

[0073] 5. Iteration and Termination Calculate the fitness value of each particle and update the individual's historical best. and global optimal Repeat the particle position and velocity update steps. When the maximum number of iterations is reached, or the global optimal fitness value has not significantly improved for several consecutive generations, terminate the iteration and output the optimal decision sequence. .

[0074] In this embodiment, the single-round optimization solution time does not exceed 3 seconds, which fully meets the real-time requirements of 15-minute electricity spot market transactions.

[0075] Step S5: Command Issuance and Rolling Closed-Loop Control A typical mechanism for rolling time-domain optimization is employed: each optimization covers 16 future time periods, taking only the first time period in the optimization sequence. ) decision value , As the current execution command: load command is transmitted via the plant-level automatic generation control (AGC) interface. The command is sent to the unit's DCS system to automatically adjust the unit's power output and to specify the biomass co-firing ratio. The data is sent to the fuel delivery control system to automatically adjust the feeding ratio of biomass and coal.

[0076] When the next 15-minute decision cycle arrives, the system rolls forward one time period, updates all real-time collected data, and repeats steps S1 to S5 to achieve rolling closed-loop optimization and continuously respond to fluctuations in spot market electricity prices and changes in unit operating conditions.

[0077] Corresponding to the above method, the multi-objective real-time optimization decision system for coal-fired power units oriented towards the electricity spot market described in this embodiment adopts a three-layer hardware and software integrated architecture of "perception layer - decision layer - execution layer", specifically including: perception layer unit: configured with electricity trading interface, DCS data acquisition interface, plant-level MIS interface, carbon market and by-product market data interface, executes the multi-source data real-time acquisition and status perception function of step S1, and transmits the data to the decision layer after completing data cleaning and time-series alignment; Decision-making unit: It has a built-in refined full-cost-benefit model library and a multi-objective rolling optimization solver, which respectively execute the sub-model construction in step S2, the optimization model building in step S3, and the intelligent algorithm solution in step S4, and output the optimal decision sequence. Execution layer unit: Configures AGC command issuance module, fuel control command issuance module and rolling cycle triggering module, executes the command issuance and rolling closed-loop control function of step S5, issues the optimal decision command to the unit control system, and triggers the optimization process of the next decision cycle to form a complete control closed loop.

[0078] This embodiment, through the aforementioned method and system, enables multi-objective real-time optimization decision-making for coal-fired power units in a spot electricity market environment: automatically increasing unit output to generate revenue during periods of high electricity prices, optimizing biomass co-firing ratios to obtain subsidies and carbon revenue during periods of low electricity prices, and dynamically ensuring fly ash quality to avoid disposal losses. Calculations show that this can improve the overall economic efficiency of the unit by 2% to 5%, while ensuring compliance with environmental emission standards and boiler operation safety. It solves the technical problems of lagging decision-making response, inability to accurately balance multiple variables, and insufficient refinement of cost-benefit models in existing technologies, demonstrating significant creativity and practicality.

Claims

1. A multi-objective real-time optimization decision-making method for coal-fired power units oriented towards the electricity spot market, characterized in that, Includes the following steps: S1: Real-time collection of multi-source heterogeneous data, including market and policy data, unit operation data, fuel inventory and characteristic data, and by-product market data; S2: Construct a refined full cost-benefit model based on the data. The refined full cost-benefit model includes a multi-fuel coupled cost model, a full-condition environmental protection cost model, and a fly ash net benefit model coupled with the prediction of fly ash quality qualification rate. S3: With the goal of maximizing the total net profit of the system in the rolling time domain, and in combination with constraints, a multi-objective rolling optimization mathematical model is constructed. S4: The multi-objective rolling optimization mathematical model is solved using a hybrid algorithm of dynamic feasible region shrinkage and particle swarm optimization to obtain the optimal decision sequence; S5: Issue the decision value corresponding to the current time period in the optimal decision sequence as an execution instruction, and repeat steps S1 to S5 in the next decision cycle to achieve rolling closed-loop control.

2. The method according to claim 1, characterized in that, In step S2: The calculation formula for the multi-fuel coupling cost model is as follows: ， in, To contribute to the unit, The proportion of biomass co-firing, For boiler efficiency, For online correction of the thermal efficiency of steam turbine generator sets, and These are the unit calorific value costs of coal and biomass, respectively. The calculation formula for the full-condition environmental protection cost model is as follows: ， in, and These are the cost functions for desulfurization and denitrification, respectively, modeled using a hybrid mechanism-data approach. The calculation formula for the fly ash net profit model is as follows: ， in, The expression for the fly ash quality qualification rate model is as follows: 。 3. The method according to claim 2, characterized in that, The objective function mentioned in step S3 is: in, This represents the total number of time periods within the rolling time domain. For the revenue from generating electricity and connecting to the grid, For ancillary services and subsidy revenue.

4. The method according to claim 3, characterized in that, The constraints described in step S3 include hard constraints and soft constraints, wherein: Hard constraints include: unit output constraints Unit ramping constraints Biomass co-firing ratio constraints ; Soft constraints include: sulfur dioxide emission concentration constraints. nitrogen oxide emission concentration constraints fly ash quality constraints .

5. The method according to claim 1, characterized in that, Step S4 specifically includes: S4.1: Initialize the particle swarm, the position of each particle. This represents a decision scheme composed of decision variables for each time period within the rolling time domain; S4.2: Perform feasible region preprocessing on hard constraints and dynamically calculate the absolute feasible region of output at each time period. Real-time upper limit of biomass co-firing ratio And the particle positions are confined to the contracted feasible region; S4.3: Constructing an unconstrained fitness function ,in , , This is an adaptive penalty factor that increases exponentially with the number of iterations. S4.4: Update the position and velocity of the particles according to the particle swarm optimization algorithm; S4.5: Iterate until the termination condition is met, and output the optimal solution sequence.

6. The method according to claim 5, characterized in that, The update formulas for particle position and velocity in step S4.4 are as follows: ， ， in, Assuming inertia weights and employing a linear decreasing strategy, , The learning factor is used and an asynchronous adjustment strategy is adopted.

7. The method according to claim 1, characterized in that, The current time period mentioned in step S5 is the first time period of the rolling time domain ( ), to optimize unit output The optimal biomass co-firing ratio is then sent to the unit's distributed control system. Issued to the fuel delivery control system.

8. A multi-objective real-time optimization decision-making system for coal-fired power units oriented towards the electricity spot market, characterized in that, The system for implementing the method as described in any one of claims 1 to 7 includes: The perception layer unit is used to perform the real-time acquisition and state perception of the multi-source data; The decision-making unit is equipped with the refined full-cost-benefit model and the multi-objective rolling optimization solver, which are used to perform model building and algorithm solving. The execution layer unit is used to issue the optimal decision command to the unit control system and trigger the rolling closed loop.

9. The system according to claim 8, characterized in that, The perception layer unit is equipped with an electricity trading interface, a distributed control system data acquisition interface, a plant-level management information system interface, a carbon market data interface, and a by-product market data interface.

10. The system according to claim 8, characterized in that, The decision-making layer unit incorporates a multi-fuel coupling cost model, a full-condition environmental protection cost model, and a fly ash net revenue model, as well as a dynamic feasible domain shrinkage-particle swarm optimization hybrid solver; the execution layer unit is equipped with an automatic power generation control command issuing module and a fuel control command issuing module.

Images