A Multi-Time-Scale, Multi-Source AGC Coordinated Control Method for Thermal Power Units
By using a multi-timescale, multi-source AGC coordinated control method for thermal power units, the usage weights of condensate throttling, battery energy storage, and flywheel energy storage are dynamically adjusted. This solves the problem of balancing the adjustment capability and thermal parameter safety of thermal power units when rapidly responding to changes in AGC target power. It achieves a balance between response speed and tracking accuracy, as well as safe control of resource status.
Patent Information
- Application Number
- CN202611636197.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-10-22
- Publication Date
- 2026-11-17
AI Technical Summary
When existing thermal power units respond quickly to changes in AGC target power, it is difficult to balance regulation capability and thermodynamic parameter safety. The allocation of rapid regulation resources is biased, making it difficult to balance response speed and tracking accuracy. When resources are close to the state boundary, they are prone to exceeding the limit, and the actual power lag causes unidirectional overshoot.
A multi-timescale, multi-source AGC coordinated control method for thermal power units is adopted. By establishing a predictive model and a rolling optimization objective function, the usage weights of condensate throttling, battery energy storage, and flywheel energy storage are dynamically adjusted. Combined with resource state margin and time scale characteristics, a reference power command is generated to achieve coordinated resource allocation and safe control.
It achieves a balance between response speed and tracking accuracy during step, directional reversal and continuous load increases and decreases, avoids resource status exceeding limits, suppresses unidirectional overshoot, and improves the regulation capability and safety of thermal power units.
Smart Images

Figure CN123121735A_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of multi-timescale automatic power generation control and multi-source energy storage coordinated control technology for thermal power units, specifically involving multi-timescale multi-source AGC coordinated control technology for thermal power units. Background Technology
[0002] With the increasing proportion of fluctuating power sources such as wind and solar power being integrated into the grid, the number of step changes, directional reversals, and continuous load increases / decreases in the grid's AGC target power is rising, requiring thermal power units to undertake frequency regulation tasks more frequently. Current thermal power coordinated control mainly alters unit output through fuel quantity, feedwater flow, and turbine regulating valves. However, the boiler combustion, pulverizing, steam-water circulation, and turbine operation processes exhibit significant inertia and delay. When the AGC target power changes rapidly, relying solely on the thermal power unit itself is insufficient to compensate for power deviations in a timely manner. Furthermore, accelerating changes in control variables can easily cause fluctuations in main steam pressure and intermediate-point enthalpy, making it difficult to balance response speed with the safety of thermal parameters.
[0003] To improve the regulation capability of thermal power units, related technologies introduce rapid regulation resources such as condensate throttling, battery energy storage, and flywheel energy storage into the frequency regulation process. Condensate throttling reduces the condensate flow rate of the low-pressure regenerative system and decreases the amount of regenerative steam extracted, allowing more steam to enter the subsequent stages of the turbine to perform work, thereby increasing unit output without increasing fuel consumption. Battery energy storage achieves rapid power throughput through a bidirectional converter. Flywheel energy storage provides short-term power support through changes in rotor kinetic energy. The response speed of these three types of resources is faster than that of the thermal power unit itself, but their response speed, sustainability, and available capacity differ significantly and change continuously with operating conditions, making it difficult to coordinate the use of these three types of resources with the unit itself.
[0004] However, the adjustability of the above three types of resources is limited. Condensate throttling is constrained by the deaerator water level and its duration. Frequent reversals of battery energy storage increase cycle losses. Flywheel energy storage has limited usable energy capacity. The capacity boundaries of each resource also change continuously with the operating state. Existing multi-energy storage frequency regulation methods mostly use fixed ratios, fixed frequency bands, or preset call sequences to allocate power. There is a deviation between the allocation result and the actual executable capacity of each resource. During processes such as step changes, direction reversals, and continuous load increases and decreases, the adjustment power required by the unit varies significantly. Fixed allocation methods cause a mismatch between the power allocation and adjustment needs among the resources. Output lag is likely to occur in the later stages of load adjustment, and it is difficult to balance response speed and tracking accuracy.
[0005] Furthermore, when a resource approaches its state boundary, the fixed allocation method still issues power commands according to the original proportion. These commands exceed the power range that the resource can continue to execute in the current state, which can easily cause the deaerator water level or charge state to exceed the limit. The actual power of each fast resource lags behind its power command. When the total power of the unit approaches the AGC target power, the lagging actual power still increases in the original direction, which can easily lead to overshoot in the same direction. Summary of the Invention
[0006] To address the problems in existing technologies, such as discrepancies between resource adjustability and allocation results, difficulty in balancing response speed and tracking accuracy, resource overshooting when approaching state boundaries, and inverse overshoot caused by actual power lag, this invention proposes the following solution: A multi-timescale, multi-source AGC coordinated control method for thermal power units includes: S1. Collect AGC target power, actual power of thermal power unit, main steam pressure, intermediate point enthalpy, condensate flow rate, deaerator water level, battery state of charge and flywheel state of charge. S2. Based on the actual power of the thermal power unit, the main steam pressure, the intermediate point enthalpy value and the condensate flow rate, establish a thermal power unit prediction model and a condensate throttling power model; based on the battery state of charge and the flywheel state of charge, establish a battery power model and a flywheel power model; and construct an equivalent condensate state of charge based on the deaerator water level. S3. Determine the auxiliary power demand based on the difference between the AGC target power and the thermal power unit prediction power obtained from the thermal power unit prediction model, and determine the time scale characteristics including the duration of the auxiliary power demand in the same direction, the typical duration scale of similar historical events, and the change in AGC target power in adjacent control cycles. S4. Based on the time scale characteristics and the state margins of each resource determined by the equivalent state of charge of the condensate, the state of charge of the battery, and the state of charge of the flywheel, update the dynamic usage weights of condensate throttling, battery, and flywheel, use the dynamic usage weights as weight coefficients of resource usage tendency, and establish a rolling optimization objective function based on the resource usage tendency and the predicted power of each resource in the prediction time domain. S5. According to the first control cycle, under the constraints of the thermal power unit's ramp rate and thermal state, condensate throttling, battery and flywheel power, as well as the state of charge and deaerator water level, solve the rolling optimization objective function to obtain the thermal power unit reference power, condensate throttling reference power, battery reference power and flywheel reference power. S6. In a second control cycle shorter than the first control cycle, generate control quantities for fuel, feedwater, and turbine regulating valves based on the reference power of the thermal power unit. In a third control cycle shorter than the second control cycle, generate condensate flow command, battery power command, and flywheel power command based on the condensate throttling reference power, battery reference power, flywheel reference power, and the state boundaries of each resource. Then, re-execute S1 to S5 in the next control cycle.
[0007] Furthermore, the equivalent state of charge of the condensate is the ratio of the current available condensate mass to the total available condensate mass within the allowable water level range. The equivalent state of charge of the condensate is linearly mapped to the deaerator water level, and is 0 when the deaerator water level is at the lower limit of the allowable water level and 1 when it is at the upper limit of the allowable water level. When the condensate throttling increases the power, the equivalent state of charge of the condensate decreases, and when the condensate throttling replenishes the condensate, the equivalent state of charge of the condensate increases.
[0008] Furthermore, the duration of the same direction is the cumulative time during which the auxiliary power demand exceeds the preset dead zone and the direction remains unchanged, and it is reset to zero when the direction of the auxiliary power demand changes or the absolute value of the auxiliary power demand is lower than the preset dead zone. The typical duration scale of the historical similar events is obtained by classifying and statistically analyzing the historical deviation events according to the deviation direction, amplitude range and unit load range. The change in AGC target power between adjacent control cycles is the difference between the AGC target power of the current control cycle and the AGC target power of the previous control cycle.
[0009] Furthermore, the update rule for the dynamic usage weight is as follows: when the mutation factor increases, the dynamic usage weight of the flywheel is increased; when the continuity factor increases, the dynamic usage weight of the flywheel is decreased and the dynamic usage weight of the battery is increased; when the expected energy factor increases, the dynamic usage weight of condensate throttling is increased.
[0010] Furthermore, the rolling optimization objective function is a weighted sum of the total power tracking error term, condensate throttling, dynamic usage cost of battery and flywheel, state recovery cost, reference power change cost, and one-sided overshoot cost. Each cost term is normalized according to the rated power or allowable state range of the corresponding resource before weighting. The one-sided overshoot cost term takes a value along the current adjustment direction. In S5, only the first control quantity in the optimal control sequence obtained by solving is executed each time.
[0011] Furthermore, the state boundary mentioned in S6 includes a hard boundary for the state of charge and the deaerator water level, as well as an execution protection zone set inside the hard boundary. When the resource state enters the execution protection zone, a limiting coefficient between 0 and 1 is determined based on the margin between the resource state and the hard boundary, and the corresponding reference power is corrected according to the limiting coefficient. Specifically, when the condensate equivalent state of charge is close to the lower boundary, the positive throttling power is reduced, and when it is close to the upper boundary, the negative replenishment power is restricted. When the battery state of charge and the flywheel state of charge are close to the lower boundary, the discharge is restricted, and when they are close to the upper boundary, the charging is restricted.
[0012] Furthermore, in S6, the actual power that has not yet been released at the next moment is predicted based on the first-order response model of each resource, and the predicted total power at the next moment is obtained. When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the reference power of the corresponding resource is reduced in advance.
[0013] Based on the same inventive concept, the present invention also proposes an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the above-described method.
[0014] Based on the same inventive concept, the present invention also proposes a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described method.
[0015] Based on the same inventive concept, the present invention also proposes a computer program product, including a computer program that, when executed by a processor, implements the above-described method.
[0016] Compared with the prior art, the present invention has the following beneficial effects: By incorporating the thermal power unit itself, the internal thermal energy storage formed by condensate throttling, battery energy storage, and flywheel energy storage into the same rolling optimization objective function, and normalizing each cost term according to the rated power or allowable state range of each resource, the power carrying capacity of each resource can change with its response speed, sustainability, and current state, thereby unifying the capacity boundary of the resources into power allocation.
[0017] The time scale characteristics are determined by the duration of auxiliary power demand in the same direction, the typical duration of similar historical events, and the change in AGC target power in adjacent control cycles. This allows the flywheel to take priority in step and direction reversal power, and the battery to take priority in continuous power, thus balancing response speed and tracking accuracy during step, direction reversal, and continuous load increases and decreases.
[0018] By setting up an execution protection zone inside the hard boundary of the charged state and deaerator water level, and correcting the reference power by determining a limiting coefficient between 0 and 1 according to the margin between the resource state and the hard boundary when the resource state enters the execution protection zone, each resource can continuously exit when approaching the boundary without generating an over-limit command, thereby avoiding over-limits of the deaerator water level and charged state.
[0019] The first-order response model of each resource is used to predict the actual power that has not yet been released at the next moment. When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the reference power of the corresponding resource is reduced in advance so that the total power will not continue to increase in the same direction when it is close to the target, thereby suppressing the same-direction overshoot and improving the tracking accuracy.
[0020] This invention features hierarchical coordination and rolling updates. It can allocate adjustment tasks based on the time scale characteristics of AGC target power and the real-time status of each resource, and output power commands for thermal power units, condensate throttling, battery energy storage, and flywheel energy storage. It is applicable to fields such as automatic power generation control of thermal power units, deep peak shaving of thermal power units, joint frequency regulation of thermal power and energy storage, secondary frequency regulation of power grids with a high proportion of new energy, and safety control of thermal parameters of thermal power units. Attached Figure Description
[0021] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings: Figure 1 This is a flowchart illustrating the method described in the implementation method; Figure 2 This is a structural diagram of the coordination and control system module described in the implementation method; Figure 3 This is a diagram showing the condensate throttling device and equivalent state of charge calculation relationship described in the implementation method; Figure 4 This is a flowchart of the resource coordination layer rolling optimization process described in the implementation method; Figure 5 This is a flowchart of the fast execution layer boundary protection and terminal braking described in the implementation method; Figure 6 This is the diagram showing the three-layer control cycle and prediction time domain relationship described in the implementation method; Figure 7 These are comparison curves of large step rapid tracking under two sets of working conditions as described in the implementation method; (a) is the process of AGC command step increase, and (b) is the process of AGC command step decrease. Detailed Implementation
[0022] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. It should be noted that, unless otherwise specified, the embodiments and features in the embodiments of the present invention can be combined with each other, and the described embodiments are only some embodiments of the present invention, not all embodiments.
[0023] Implementation Method 1: This implementation method provides a multi-timescale, multi-source AGC coordinated control method for thermal power units, including: S1. Collect AGC target power, actual power of thermal power unit, main steam pressure, intermediate point enthalpy, condensate flow rate, deaerator water level, battery state of charge and flywheel state of charge. S2. Based on the actual power of the thermal power unit, the main steam pressure, the intermediate point enthalpy value and the condensate flow rate, establish a thermal power unit prediction model and a condensate throttling power model; based on the battery state of charge and the flywheel state of charge, establish a battery power model and a flywheel power model; and construct an equivalent condensate state of charge based on the deaerator water level. S3. Determine the auxiliary power demand based on the difference between the AGC target power and the thermal power unit prediction power obtained from the thermal power unit prediction model, and determine the time scale characteristics including the duration of the auxiliary power demand in the same direction, the typical duration scale of similar historical events, and the change in AGC target power in adjacent control cycles. S4. Based on the time scale characteristics and the state margins of each resource determined by the equivalent state of charge of the condensate, the state of charge of the battery, and the state of charge of the flywheel, update the dynamic usage weights of condensate throttling, battery, and flywheel, use the dynamic usage weights as weight coefficients of resource usage tendency, and establish a rolling optimization objective function based on the resource usage tendency and the predicted power of each resource in the prediction time domain. S5. According to the first control cycle, under the constraints of the thermal power unit's ramp rate and thermal state, condensate throttling, battery and flywheel power, as well as the state of charge and deaerator water level, solve the rolling optimization objective function to obtain the thermal power unit reference power, condensate throttling reference power, battery reference power and flywheel reference power. S6. In a second control cycle shorter than the first control cycle, generate control quantities for fuel, feedwater, and turbine regulating valves based on the reference power of the thermal power unit. In a third control cycle shorter than the second control cycle, generate condensate flow command, battery power command, and flywheel power command based on the condensate throttling reference power, battery reference power, flywheel reference power, and the state boundaries of each resource. Then, re-execute S1 to S5 in the next control cycle.
[0024] Preferably, the first control cycle, the second control cycle, and the third control cycle can be adjusted proportionally according to the device communication rate and dynamic response time, provided that the third control cycle is shorter than the second control cycle and the second control cycle is shorter than the first control cycle.
[0025] By using the above method, the internal thermal energy storage formed by the thermal power unit body, condensate throttling, battery energy storage, and flywheel energy storage are incorporated into the same rolling optimization problem. The optimization is performed in three control cycles and then re-optimized in the next control cycle, thereby achieving the coordinated allocation of multi-source frequency regulation resources.
[0026] Furthermore, the equivalent state of charge of the condensate is the ratio of the current available condensate mass to the total available condensate mass within the allowable water level range. The equivalent state of charge of the condensate is linearly mapped to the deaerator water level, and is 0 when the deaerator water level is at the lower limit of the allowable water level and 1 when it is at the upper limit of the allowable water level. When the condensate throttling increases the power, the equivalent state of charge of the condensate decreases, and when the condensate throttling replenishes the condensate, the equivalent state of charge of the condensate increases.
[0027] By defining the equivalent state of charge of condensate as the ratio of the mass of condensate available within the allowable water level range to the total mass of condensate available, the deaerator water level margin is converted into a state variable with the same dimensions as the state of charge, thereby allowing condensate throttling and energy storage to be included in the rolling optimization objective function.
[0028] Furthermore, the duration of the same direction is the cumulative time during which the auxiliary power demand exceeds the preset dead zone and the direction remains unchanged, and it is reset to zero when the direction of the auxiliary power demand changes or the absolute value of the auxiliary power demand is lower than the preset dead zone. The typical duration scale of the historical similar events is obtained by classifying and statistically analyzing the historical deviation events according to the deviation direction, amplitude range and unit load range. The change in AGC target power between adjacent control cycles is the difference between the AGC target power of the current control cycle and the AGC target power of the previous control cycle.
[0029] By determining the duration in the same direction by the cumulative time exceeding the preset dead zone while maintaining the same direction, and by statistically analyzing the typical duration scale of similar historical events according to the deviation direction, amplitude, and load range, the duration of auxiliary power demand can be quantified, thereby providing a basis for updating dynamic usage weights.
[0030] Furthermore, the update rule for the dynamic usage weight is as follows: when the mutation factor increases, the dynamic usage weight of the flywheel is increased; when the continuity factor increases, the dynamic usage weight of the flywheel is decreased and the dynamic usage weight of the battery is increased; when the expected energy factor increases, the dynamic usage weight of condensate throttling is increased.
[0031] By adjusting the dynamic usage weights of condensate throttling, batteries, and flywheels according to the mutation factor, duration factor, and expected energy factor, the flywheel is given priority to undertake step and reverse power, and the battery is given priority to undertake continuous power, thereby matching resource capacity with time scale characteristics.
[0032] Furthermore, the rolling optimization objective function is a weighted sum of the total power tracking error term, condensate throttling, dynamic usage cost of battery and flywheel, state recovery cost, reference power change cost, and one-sided overshoot cost. Each cost term is normalized according to the rated power or allowable state range of the corresponding resource before weighting. The one-sided overshoot cost term takes a value along the current adjustment direction. In S5, only the first control quantity in the optimal control sequence obtained by solving is executed each time.
[0033] By weighting and summing the tracking error term, dynamic usage cost term, state recovery cost term, reference power change cost term, and one-sided overshoot cost term, and normalizing them according to the rated power or allowable state range, the cost terms are made comparable, thereby simultaneously taking into account tracking accuracy, resource consumption, and overshoot suppression.
[0034] Furthermore, the state boundary mentioned in S6 includes a hard boundary for the state of charge and the deaerator water level, as well as an execution protection zone set inside the hard boundary. When the resource state enters the execution protection zone, a limiting coefficient between 0 and 1 is determined based on the margin between the resource state and the hard boundary, and the corresponding reference power is corrected according to the limiting coefficient. Specifically, when the condensate equivalent state of charge is close to the lower boundary, the positive throttling power is reduced, and when it is close to the upper boundary, the negative replenishment power is restricted. When the battery state of charge and the flywheel state of charge are close to the lower boundary, the discharge is restricted, and when they are close to the upper boundary, the charging is restricted.
[0035] By setting an execution protection zone inside the hard boundary and determining a limiting coefficient between 0 and 1 based on the margin between the resource state and the hard boundary, the reference power is continuously corrected when the resource state approaches the boundary, thereby preventing the water level and state of charge from exceeding the limit and absorbing the residual power of the first-order response.
[0036] Furthermore, in S6, the actual power that has not yet been released at the next moment is predicted based on the first-order response model of each resource, and the predicted total power at the next moment is obtained. When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the reference power of the corresponding resource is reduced in advance.
[0037] The first-order response model of each resource is used to predict the actual power that has not yet been released at the next moment. When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the reference power is reduced in advance so that the total power will no longer increase in the same direction when it is close to the target, thereby suppressing the same-direction overshoot.
[0038] The method described in this embodiment can be executed by a processor calling a computer program, which can be stored in a computer storage medium. When the computer program is executed by the processor, the above-mentioned multi-timescale, multi-source AGC coordinated control method for thermal power units can be implemented.
[0039] Implementation Method Two: This implementation method uses a multi-timescale, multi-source AGC coordinated control method for thermal power units as a specific example, combined with... Figures 1 to 6 The first implementation method will be further explained below.
[0040] like Figure 1 As shown, the method described in this embodiment is executed in the order of S1 to S6. The resource coordination layer solves and outputs the reference power of the thermal power unit, the reference power of condensate throttling, the reference power of the battery, and the reference power of the flywheel according to the first control cycle. The thermal power control layer generates the control quantities of fuel, feedwater, and turbine regulating valves according to the second control cycle. The fast execution layer generates the condensate flow command, the battery power command, and the flywheel power command according to the third control cycle, and re-executes S1 to S5 using the latest measurement status in the next control cycle.
[0041] In S1, the following data are collected: AGC target power, actual power of the thermal power unit, main steam pressure, intermediate point enthalpy, condensate flow rate, deaerator water level, battery state of charge, and flywheel state of charge. Simultaneously, fuel commands, feedwater flow rate, turbine regulating valve opening, and feedback status of each actuator are also collected. The collected signals are updated synchronously according to the third control cycle, serving as inputs to the thermal power unit prediction model, condensate throttling power model, battery power model, and flywheel power model. These signals also serve as the basis for constructing the condensate equivalent state of charge, calculating the state margin of each resource, and determining whether a resource has entered the execution protection zone.
[0042] In S2, a prediction model for the thermal power unit and a condensate throttling power model are established based on the actual power of the thermal power unit, the main steam pressure, the intermediate point enthalpy, and the condensate flow rate. A battery power model and a flywheel power model are established based on the battery state of charge and the flywheel state of charge. An equivalent state of charge for the condensate is constructed based on the deaerator water level. The thermal power unit prediction model adopts a low-order nonlinear boiler turbine model suitable for online prediction, and its state vector, input vector, and output vector are:
[0043] in, The state vector consists of the combustion rate state, the steam-water separator pressure, and the midpoint enthalpy. In terms of combustion rate, For the pressure of the steam-water separator, The enthalpy value is the value at the midpoint of the heated section. The input vector consists of fuel command, feedwater flow rate, and turbine control valve opening. For fuel directive, For water supply flow rate, For the turbine regulating valve opening, The output vector is composed of main steam pressure, midpoint enthalpy, and actual power of the thermal power unit. This refers to the main steam pressure at the turbine inlet. This represents the actual power output of the thermal power unit.
[0044] The dynamics of pulverization, pressure, and midpoint enthalpy are as follows:
[0045] in, The coefficient of inertia in milling. For the combustion rate state, the dot above the variable represents the derivative of that variable with respect to time. It is a natural constant. The pure time delay of the pulverizing system. For the Laplace operator, For fuel directives.
[0046]
[0047] in, The energy storage coefficient of the heated section of the pressure channel. For the pressure of the steam-water separator, To supply water enthalpy, The first equivalent enthalpy parameter, For water supply flow rate, This is the enthalpy correction factor. The enthalpy value is the value at the midpoint of the heated section. Main vapor enthalpy The main steam flow rate entering the turbine. For heat absorption gain, This is the combustion rate state.
[0048]
[0049] in, The energy storage coefficient of the heated section of the enthalpy channel. The enthalpy value is the value at the midpoint of the heated section. To supply water enthalpy, This is the second equivalent enthalpy parameter. For water supply flow rate, This is the enthalpy correction factor. Main vapor enthalpy The main steam flow rate entering the steam turbine. For heat absorption gain, This is the combustion rate state.
[0050] The superheater outlet condition and turbine power meet the following requirements:
[0051] in, For the pressure of the steam-water separator, This refers to the main steam pressure at the turbine inlet. The nonlinear function of pressure drop is identified from unit test or historical operating data. The main steam flow rate entering the steam turbine. For the turbine regulating valve opening, The main steam flow nonlinear function is identified from unit test or historical operating data. Main vapor enthalpy It is a nonlinear function of the main steam enthalpy value identified from unit test or historical operating data.
[0052]
[0053] in, This represents the actual power output of the thermal power unit. To increase the power gain of the steam turbine. The main steam flow rate entering the steam turbine. Main vapor enthalpy For water enthalpy.
[0054] The resource coordination layer performs numerical integration on the above model within each first control cycle to obtain the power and thermal state of the thermal power units in the predicted time domain. The actual additional power of condensate throttling is calculated using a first-order inertial model.
[0055] in, The time constant of the condensate throttling power channel. The actual additional power of condensate, This is the reference power for condensate throttling.
[0056] Combination Figure 3 The changes in condensate flow rate and the actual additional power of condensate flow satisfy the following within the operating range:
[0057] in, The actual additional power of condensate, The conversion gain between condensate power and condensate flow rate. This represents the change in condensate flow rate relative to the baseline operating condition.
[0058] definition To allow for the total available condensate mass within the water level range, Given the current condensate quality that can still be used for throttling and frequency regulation, the equivalent state of charge of the condensate is:
[0059] in, This represents the equivalent charged state of condensate. For the condensate quality that can still be used for throttling and frequency regulation, The total amount of condensate that can be called up within the allowed water level range.
[0060] From the conservation of condensate mass, we can obtain the continuous state equation and the discrete update equation, where 3600 is the conversion factor between hours and seconds:
[0061] in, This represents the equivalent charged state of condensate. The actual additional power of condensate, To allow for the total available condensate mass within the water level range, The formula represents the conversion gain between condensate power and condensate flow rate, where 3600 is the conversion factor between hours and seconds.
[0062]
[0063] in, This represents the equivalent charged state of condensate. To control the cycle number, This represents the equivalent charge state of the condensate in the (k+1)th control cycle. The equivalent charge state of condensate in the kth control cycle is... This represents the actual additional power of condensate during the k-th control cycle. This is the third control cycle. The actual additional power of condensate, To allow for the total available condensate mass within the water level range, The formula represents the conversion gain between condensate power and condensate flow rate, where 3600 is the conversion factor between hours and seconds.
[0064] If the effective cross-sectional area of the deaerator remains approximately constant within the allowable range, then there is a linear mapping relationship between the equivalent state of charge of the condensate and the water level in the deaerator:
[0065] in, This represents the equivalent charged state of condensate. To control the cycle number, This refers to the deaerator water level. This is the lower limit of the allowable water level for the deaerator. This is the upper limit of the allowable water level for the deaerator. The actual water level of the deaerator during the kth control cycle.
[0066] Both battery energy storage and flywheel energy storage employ a first-order power stage and update the state of charge according to the charging and discharging direction:
[0067] in, The time constant of the resource power channel. For the actual power of the resources, Here, j represents the resource reference power, and j represents the resource number.
[0068]
[0069] in, The resource is in a state of charge. For the actual power of the resources, For the rated energy capacity of resources, For resource discharge efficiency, The formula represents the resource charging efficiency, where 3600 is the conversion factor between hours and seconds.
[0070] The aforementioned resources include condensate throttling, battery energy storage, and flywheel energy storage. Power is positive when it increases power to the grid and negative when it absorbs power from the grid. In this embodiment, the condensate throttling power range is -3MW to 15MW, the total available condensate mass within the allowable water level range is 82.99t, the power channel time constant is 8s, and the condensate equivalent state of charge range is 2% to 98%, mainly undertaking short-term thermal energy storage; the battery energy storage power range is -30MW to 30MW, the rated energy capacity is 15MWh, the power channel time constant is 2s, and the state of charge range is 15% to 85%, mainly undertaking continuous power and energy support; the flywheel energy storage power range is -15MW to 15MW, the rated energy capacity is 1.5MWh, the power channel time constant is 0.15s, and the state of charge range is 10% to 90%, mainly undertaking step and direction reversal.
[0071] In S3, the auxiliary power demand is determined based on the difference between the AGC target power and the predicted power of the thermal power unit obtained from the thermal power unit prediction model. Furthermore, based on the duration of the auxiliary power demand in the same direction, the typical duration of similar historical events, and the change in AGC target power between adjacent control cycles, time-scale characteristics including a duration factor, a projected energy factor, and a mutation factor are determined. The resource coordination layer first calculates the current auxiliary power demand:
[0072] in, To meet auxiliary power requirements, To control the cycle number, For AGC target power, The predicted power of the thermal power unit is obtained from the thermal power unit prediction model. The auxiliary power demand for the k-th control cycle, The target power for AGC in the k-th control cycle is... The predicted power of the thermal power unit is obtained from the thermal power unit prediction model in the kth control cycle.
[0073] The controller records the duration of the current power deviation in the same direction. And obtain typical persistence scales based on similar historical deviation events. This yields the duration estimate and projected energy demand:
[0074] in, For duration estimation, To control the cycle number, As a typical duration scale for similar historical events, Duration in the same direction, This is the third control cycle. To obtain the maximum value, This is the estimated duration during the k-th control period.
[0075]
[0076] in, To predict energy demand, To control the cycle number, This represents the projected energy demand during the k-th control cycle. To meet auxiliary power requirements, For duration estimation, 3600 in the formula is the conversion factor between hours and seconds.
[0077] When the direction of auxiliary power demand changes or its absolute value falls below the preset dead zone, the duration in the same direction is reset to zero and re-statistically recorded. Historical similar events are categorized by deviation direction, amplitude range, and unit load range. In the... Within the first control cycle, the thermal power unit prediction model provides the future... The predicted power of the three types of fast resources jointly compensates for the auxiliary power demand, and the prediction error is:
[0078] in, For power prediction error, To control the cycle number, For the prediction step number, To meet auxiliary power requirements, The actual additional power of condensate, This represents the actual power of the battery. This represents the actual power of the flywheel. Let the power prediction error for the (k+i)th control cycle be the prediction value for the kth control cycle. To predict the auxiliary power demand for the (k+1)th control cycle from the kth control cycle, To determine the actual additional power of condensate in the (k+1)th control cycle as predicted in the kth control cycle, The actual battery power in the (k+1)th control cycle is predicted in the kth control cycle. The actual flywheel power for the (k+1)th control cycle is predicted in the kth control cycle.
[0079] To identify short-term step jumps, persistence biases, and direction reversals, persistence factors, predicted energy factors, and mutation factors are constructed respectively:
[0080] in, Let be the persistence factor for the k-th control period. Let be the duration of the same direction during the k-th control cycle. As a persistent factor, To control the cycle number, Duration in the same direction, As a short-time process discrimination metric, The smoothing time constant, It is an exponential function.
[0081]
[0082] in, Let be the predicted energy factor for the k-th control cycle. To predict the energy factor, To control the cycle number, For duration estimation, As a long-term process discrimination metric, To obtain the minimum value.
[0083]
[0084] in, For the mutation factor in the k-th control cycle, As a mutation factor, To control the cycle number, This represents the change in target power for AGC in adjacent control cycles. As a step normalization benchmark, To obtain the minimum value.
[0085] All three factors are limited to between 0 and 1.
[0086] In S4, based on the time scale characteristics and the resource state margins determined by the condensate equivalent state of charge, the battery state of charge, and the flywheel state of charge, the dynamic usage weights of condensate throttling, battery, and flywheel are updated. These dynamic usage weights are used as weight coefficients for resource usage tendency, and a rolling optimization objective function is established based on the resource usage tendency and the predicted power of each resource in the prediction time domain. Combined with... Figure 4 The dynamic usage weights for condensate throttling, batteries, and flywheel are:
[0087] in, The dynamic weighting of condensate throttling in the k-th control cycle. The dynamic usage weight of the battery in the k-th control cycle. The dynamic weights used for the flywheel in the k-th control cycle are determined. The dynamic use weight for condensate throttling, Assign dynamic usage weights to the battery. Weights are used dynamically for the flywheel. To control the cycle number, The basic weighting coefficient for condensate throttling. The basic weighting coefficient for the battery. This is the basic weighting coefficient of the flywheel. The reference coefficient for condensate throttling is... The coefficient for the duration of condensate throttling. The energy demand coefficient for condensate throttling. This is the baseline coefficient for the battery. This is the coefficient for the battery duration term. This is the reference coefficient for the flywheel. The coefficient for the flywheel's instruction mutation term. For the duration term coefficient of the flywheel, As a persistent factor, To predict the energy factor, It is a mutation factor.
[0088] In this embodiment, the basic weighting coefficient and each coefficient are respectively taken as: It is 0.20. It is 0.55. It is 3.5. It is 1.2; It is 0.40. It is 1.8. It is 1.15; It is 0.10. It is 1.5. It is 1.25. It is 1.5.
[0089] A higher dynamic utilization weight indicates that the corresponding resource has a higher priority in consuming power. When the mutation factor increases, the dynamic utilization weight of the flywheel increases; when the persistence factor increases, the dynamic utilization weight of the flywheel decreases, while the dynamic utilization weight of the battery increases; when the expected energy factor increases, the dynamic utilization weight of condensate throttling increases. The resource coordination layer establishes the following rolling optimization objective function:
[0090] in, The objective function for the rolling optimization of the resource coordination layer, To obtain the minimum value, In order to reach a settlement, For the prediction step number, For the number of prediction steps in the upper layer, To track error weights, For the first The power prediction error of the step, For resource serial number, This represents the dynamic usage weight of resources; a larger value indicates that the corresponding resource has higher priority in receiving power. To control the cycle number, For the first Predicted power of step resources For the first The step is a state vector composed of the charged states of each resource. The target vector for state recovery. For norm, The state recovery weight matrix is... For one-sided overshoot weights, This is the unilateral overshoot along the current adjustment direction. This is a vector of changes in the reference power of each resource. This is the reference power change weight matrix.
[0091] Before weighting, each cost term in the objective function is normalized according to the rated power or allowable state range of the corresponding resource. The resource coordination layer only executes the first reference power in the optimal control sequence each time, and re-solves using the latest state in the next first control cycle.
[0092] In S5, according to the first control cycle, the rolling optimization objective function is solved under constraints of the thermal power unit's ramp rate and thermal state, condensate throttling, battery and flywheel power constraints, and state of charge and deaerator water level constraints, to obtain the reference power of the thermal power unit, the reference power of condensate throttling, the reference power of the battery, and the reference power of the flywheel. Combined with... Figure 6 The first control period is 30 s, and the first prediction time domain is 300 s; the rolling optimization also satisfies the following constraints:
[0093] in, This represents the actual power output of the thermal power unit during the (k+1)th control cycle. This represents the actual power output of the thermal power unit during the k-th control cycle. This is the lower limit of resource power. For the actual power of the resources, This is the upper limit of resource power. This is the limit of the resource's state of charge. The resource is in a state of charge. This represents the upper limit of the resource's state of charge. This represents the actual power output of the thermal power unit. To control the cycle number, The allowable gradient for thermal power units. The first control cycle is measured in minutes.
[0094] In this embodiment, the allowable ramp rate for the thermal power unit is 12 MW / min; the condensate equivalent state of charge adopts a hard boundary of 0.02 to 0.98, with an execution protection zone set at 0.06 to 0.94; the battery and flywheel adopt the above-mentioned hard state of charge range. The actual total power of the system is:
[0095] in, The actual total power of the system in the k-th control cycle is... The actual additional power of condensate in the kth control cycle. The actual battery power in the k-th control cycle. The actual power of the flywheel in the k-th control cycle. This represents the actual total power of the system. To control the cycle number, This represents the actual power output of the thermal power unit. The actual additional power of condensate, This represents the actual power of the battery. This represents the actual power of the flywheel.
[0096] In S6, according to a second control cycle shorter than the first control cycle, control quantities for fuel, feedwater, and turbine regulating valves are generated based on the reference power of the thermal power unit. In a third control cycle shorter than the second control cycle, according to the condensate throttling reference power, the battery reference power, the flywheel reference power, and the state boundaries of each resource, condensate flow command, battery power command, and flywheel power command are generated. S1 to S5 are then re-executed in the next control cycle. (Combined with...) Figure 5 The second control cycle is 10 s, the second prediction time domain is 100 s, and the third control cycle is 1 s with single-step safety prediction. The thermal power control layer converts the reference power of the thermal power unit into fuel quantity, feedwater flow rate, and turbine regulating valve opening. Its optimization objective is:
[0097] in, The objective function for optimizing the thermal power control layer is... To obtain the minimum value, In order to reach a settlement, For the prediction step number, For thermal power prediction steps, For the first The output vector consists of thermal power output, main steam pressure, and intermediate point enthalpy. For the corresponding reference vector, For norm, To output the bias weight matrix, For the first The vector of changes in the opening of the fuel, feedwater, and turbine regulating valves. This is the incremental weight matrix for fuel, water supply, and control valves.
[0098] The control quantity is also limited by its amplitude and rate of change. The fast execution layer calculates the terminal auxiliary power target based on the predicted power of the thermal power unit at the next moment:
[0099] in, The terminal auxiliary power target for the k-th control cycle is... The target power for AGC in the k-th control cycle is... The predicted power of the thermal power unit is obtained from the thermal power unit prediction model in the (k+1)th control cycle. For terminal auxiliary power target, To control the cycle number, For AGC target power, The predicted power of the thermal power unit is obtained from the thermal power unit prediction model.
[0100] And based on the first-order response of each fast resource, estimate the actual power that has not yet been released at the next time step:
[0101] in, This is the predicted auxiliary power value for the next time step. The resource reference power for the k-th control cycle. To control the cycle number, In order to reach a settlement, For resource serial number, These are the first-order preservation coefficients. For the actual power of the resources, For resource reference power, This is the third control cycle. The time constant of the resource power channel. It is an exponential function.
[0102] When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the fast execution layer reduces the reference power of the corresponding resources in advance to reduce in-direction overshoot. When the resource state enters the execution protection zone inside the hard boundary, a limiting coefficient between 0 and 1 is calculated based on the resource state margin, and the resource command is corrected.
[0103] in, This refers to the corrected resource instruction power. This is the amplitude limiting factor. The resource is in a state of charge. This is the reference power for the resource.
[0104] For condensate throttling, the positive throttling power is gradually reduced when approaching the lower limit of the deaerator water level, and the negative replenishment power is limited when approaching the upper limit of the deaerator water level. For the battery and flywheel, discharge is limited when approaching the lower limit of the state of charge, and charging is limited when approaching the upper limit of the state of charge. The rolling optimization objective function uses hard boundary constraints, and the fast execution layer uses an execution protection zone. The two levels of protection jointly absorb the residual power of the first-order response and the prediction model error.
[0105] like Figure 2As shown, the method described in this embodiment can be implemented by a coordinated control system including an AGC command acquisition module, a resource coordination layer, a thermal power control layer, a fast execution layer, and a status acquisition and protection module. The AGC command acquisition module is used to receive the AGC target power issued by the power grid; the status acquisition and protection module is used to acquire the thermal state of the thermal power unit, the deaerator water level, the battery charge state, and the flywheel charge state, and feed the above states back to each control layer; the resource coordination layer includes a deviation characteristic calculation unit and a rolling optimization unit, used to calculate the auxiliary power demand and the above time scale characteristics, update the dynamic usage weight, and solve the rolling optimization objective function; the thermal power control layer includes a nonlinear prediction model and a coordinated control unit, used to generate fuel commands, feedwater commands, and turbine regulating valve commands based on the thermal power unit reference power and apply them to the thermal power unit body; the fast execution layer includes a condensate throttling device, battery energy storage, and flywheel energy storage, used to generate condensate flow commands, battery power commands, and flywheel power commands and implement state boundary protection and terminal braking, wherein the condensate throttling device includes a condenser, a low-pressure heater, and a deaerator. The above modules exchange target power, predicted status, reference power, and execution feedback through the power plant control network.
[0106] Implementation Method 3: This implementation method further illustrates the control effect of the multi-time-scale multi-source AGC coordinated control method for thermal power units described in Implementation Method 2 through experiments.
[0107] In this implementation, the actual AGC target power of two 240-minute intervals of a 1050MW thermal power unit is selected as input. The original AGC step value is converted to 30-second sampling in the previous value holding method without smoothing, so as to retain the step and direction reversal. In addition, a 540-minute instruction is used to compare the battery frequency regulation load. All comparison schemes use the same AGC target power, unit model, control cycle and state boundary.
[0108] The control process is executed according to the steps described in Implementation Method 2. The hard boundaries, execution protection zones, and first control cycles of each resource remain unchanged, and the dynamic usage weights are updated online according to the time scale characteristics.
[0109] Under two 240-minute operating conditions, the comprehensive frequency regulation index of the complete coordinated scheme was 3.0694 and 2.9621, respectively. Compared with conventional thermal power control, the comprehensive frequency regulation index improved by 48.40% and 43.85%, the speed index improved by 22.80% and 26.11%, and the accuracy index improved by 19.76% and 11.56%, respectively. All the above improvements are relative to the corresponding indicators of conventional thermal power control. The comparison used the same AGC target power, unit model, control cycle, and state boundaries. In the resource withdrawal test, after removing the flywheel, the comprehensive frequency regulation index of the two operating conditions decreased by 5.95% and 12.84%, respectively; after removing the battery, it decreased by 6.81% and 4.13%, respectively; and after removing the condensate, it decreased by 1.39% and 0.75%, respectively. This shows that the flywheel has a significant impact on the step and commutation processes, the battery mainly affects continuous power tracking, and the condensate mainly provides short-term support. During the 540-minute statistical period, when the power previously borne by the condensate and flywheel was shifted to the battery, the battery energy throughput, equivalent complete cycle count of rainflow, and maximum power demand increased by 39.82%, 39.74%, and 30.07%, respectively. During the full operation of the system, the states of charge of the condensate, battery, and flywheel remained within the set ranges, with the main steam pressure ranging from 22.770 MPa to 22.822 MPa and from 22.761 MPa to 22.828 MPa, respectively.
[0110] like Figure 7 (a) and Figure 7 As shown in (b), this embodiment presents local tracking curves of AGC commands and total power under two sets of operating conditions. The horizontal axis represents time, and the vertical axis represents power. AGC commands are represented by solid lines, the total power with fast tracking enabled is represented by solid lines with circular markers, and the total power with fast tracking disabled is represented by dashed lines. Both sets of operating conditions were obtained under the condition of enabling terminal prediction. Figure 7 (a) The process corresponding to the step inversion of the AGC instruction. Figure 7 (b) Corresponding to the step descent process of the AGC instruction, the AGC instruction exhibits a segmented step change in both processes. Figure 7 (a) It is evident that after a step jump in the AGC command, compared to the case where fast tracking is disabled, the total power with fast tracking enabled rises faster in the initial stage of the step jump, and the power deviation between the AGC command and the total power decreases more quickly. Conversely, the total power with fast tracking disabled rises more slowly, and the power deviation persists for a longer period. Figure 7 (b) It can be seen that after the AGC command step descent, compared with the case where fast tracking is disabled, the total power with fast tracking enabled decreases faster and the power deviation shrinks faster. However, the total power with fast tracking disabled is higher than that with fast tracking enabled in the initial stage of the step descent. Under the above two sets of operating conditions, the total power with fast tracking enabled fluctuates slightly around the AGC command when approaching it, without significant overshoot in the same direction.
[0111] The above results show that, under the condition of unchanged state constraints, this embodiment can improve the AGC target power tracking performance of thermal power units by coordinating internal thermal energy storage and external energy storage across multiple time scales, and reduce the fast commutation power, energy exchange and peak power requirements borne by the battery.
[0112] The specific embodiments of the present invention disclosed above are merely illustrative of the invention. These embodiments do not exhaustively describe all details, nor do they limit the invention to the specific embodiments described. Many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention.
Claims
1. A multi-time scale multi-source AGC coordinated control method for a thermal power unit, characterized in that, include: S1. Collect AGC target power, actual power of thermal power unit, main steam pressure, intermediate point enthalpy, condensate flow rate, deaerator water level, battery state of charge and flywheel state of charge. S2. Based on the actual power of the thermal power unit, the main steam pressure, the intermediate point enthalpy value and the condensate flow rate, establish a thermal power unit prediction model and a condensate throttling power model; based on the battery state of charge and the flywheel state of charge, establish a battery power model and a flywheel power model; and construct an equivalent condensate state of charge based on the deaerator water level. S3. Determine the auxiliary power demand based on the difference between the AGC target power and the thermal power unit prediction power obtained from the thermal power unit prediction model, and determine the time scale characteristics including the duration of the auxiliary power demand in the same direction, the typical duration scale of similar historical events, and the change in AGC target power in adjacent control cycles. S4. Based on the time scale characteristics and the state margins of each resource determined by the equivalent state of charge of the condensate, the state of charge of the battery, and the state of charge of the flywheel, update the dynamic usage weights of condensate throttling, battery, and flywheel, use the dynamic usage weights as weight coefficients of resource usage tendency, and establish a rolling optimization objective function based on the resource usage tendency and the predicted power of each resource in the prediction time domain. S5. According to the first control cycle, under the constraints of the thermal power unit's ramp rate and thermal state, condensate throttling, battery and flywheel power, as well as the state of charge and deaerator water level, solve the rolling optimization objective function to obtain the thermal power unit reference power, condensate throttling reference power, battery reference power and flywheel reference power. S6. In a second control cycle shorter than the first control cycle, generate control quantities for fuel, feedwater, and turbine regulating valves based on the reference power of the thermal power unit. In a third control cycle shorter than the second control cycle, generate condensate flow command, battery power command, and flywheel power command based on the condensate throttling reference power, battery reference power, flywheel reference power, and the state boundaries of each resource. Then, re-execute S1 to S5 in the next control cycle.
2. The method of claim 1, wherein, The equivalent state of charge (SBC) of the condensate is the ratio of the current available condensate mass to the total available condensate mass within the allowable water level range. The SBC is linearly mapped to the deaerator water level, and is 0 when the deaerator water level is at the lower limit of the allowable water level and 1 when it is at the upper limit of the allowable water level. When the condensate throttling increases the power, the SBC decreases, and when the condensate throttling replenishes the condensate, the SBC increases.
3. The method of claim 1, wherein, The duration of the same direction is the cumulative time during which the auxiliary power demand exceeds the preset dead zone and the direction remains unchanged. It is reset to zero when the direction of the auxiliary power demand changes or the absolute value of the auxiliary power demand is lower than the preset dead zone. The typical duration scale of the historical similar events is obtained by classifying and statistically analyzing the historical deviation events according to the deviation direction, amplitude range and unit load range. The change in AGC target power between adjacent control cycles is the difference between the AGC target power of the current control cycle and the AGC target power of the previous control cycle.
4. The method of claim 3, wherein, The update rule for the dynamic usage weight is as follows: when the mutation factor increases, the dynamic usage weight of the flywheel is increased; when the continuity factor increases, the dynamic usage weight of the flywheel is decreased and the dynamic usage weight of the battery is increased; when the expected energy factor increases, the dynamic usage weight of condensate throttling is increased.
5. The method according to claim 1, characterized in that, The rolling optimization objective function is a weighted sum of the total power tracking error term, condensate throttling, dynamic usage cost of battery and flywheel, state recovery cost, reference power change cost, and one-sided overshoot cost. Each cost term is normalized according to the rated power or allowable state range of the corresponding resource before weighting. The one-sided overshoot cost term takes a value along the current adjustment direction. In S5, only the first control quantity in the optimal control sequence obtained by solving is executed each time.
6. The method according to claim 1, characterized in that, The state boundary mentioned in S6 includes a hard boundary for the state of charge and the deaerator water level, as well as an execution protection zone set inside the hard boundary. When the resource state enters the execution protection zone, a limiting coefficient between 0 and 1 is determined based on the margin between the resource state and the hard boundary, and the corresponding reference power is corrected according to the limiting coefficient. Specifically, when the condensate equivalent state of charge is close to the lower boundary, the positive throttling power is reduced, and when it is close to the upper boundary, the negative replenishment power is restricted. When the battery state of charge and the flywheel state of charge are close to the lower boundary, the discharge is restricted, and when they are close to the upper boundary, the charging is restricted.
7. The method according to claim 1, characterized in that, In S6, the actual power that has not yet been released at the next moment is predicted based on the first-order response model of each resource, and the predicted total power at the next moment is obtained. When the predicted total power is close to the AGC target power and there is still residual auxiliary power along the current adjustment direction, the reference power of the corresponding resource is reduced in advance.
