Electric power active power balance control method for coupling multi-element energy storage of coal power unit

Abstract

The invention relates to the technical field of electric power system regulation and control, and particularly discloses an electric power active balance control method for coupling multi-element energy storage of a coal power unit, which is used for solving the problem that the existing static active safety margin evaluation method cannot evaluate the safety of the coal power unit under the complex working conditions of high wind power penetration, large disturbance, system inertia decline and the like. The method solves the problems that multi-stage frequency response cannot be accurately described, online self-adaptive power distribution and continuous smooth scheduling are lacked, and consequently frequency drop is large, recovery is slow and output suddenly changes, and comprises the steps of building a collaborative characteristic model, calculating dominant characteristics on line, optimizing scheduling in a segmented mode and updating parameters in a closed loop. According to the method, by constructing a hierarchical inertia and multi-stage frequency modulation collaborative model, and adopting multi-scale online feature extraction and adaptive power distribution, segmented gradient optimization scheduling and closed-loop online parameter updating, quick response, smooth connection and accurate optimization of frequency support and economic scheduling in a wind power high-permeability scene are realized.

Description

Technical Field

[0001] The present invention relates to the technical field of power system regulation and control, and in particular to a method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage. Background Art

[0002] Chinese invention patent application publication number CN119340999A discloses a method for assessing the active power safety margin of a power system, taking into account the uncertainty of renewable energy output. This method constructs a static active power safety domain model based on active power balance, line phase angle difference, and unit output constraints. The maximum volume optimization method is used to solve the safety domain boundary. The method then combines interval mathematics and the Lagrange multiplier method to evaluate the static active power safety margin in both risk and safety states. However, this method primarily focuses on quantifying static margins and fails to reflect the system's multi-stage frequency response characteristics under large disturbances or payload fluctuations. It also lacks characterization of inertial response, primary, secondary, and tertiary frequency regulation, and ramping capability. Furthermore, the model parameters are set based on fixed operating conditions and cannot be adjusted adaptively, making the assessment results difficult to adapt to real-world scenarios with rapid changes in wind and solar power output. Furthermore, the assessment results are not closely integrated with the coordinated control and economic dispatch of the unit and energy storage, making it impossible to guide coal-fired power units coupled with multi-element energy storage systems in achieving rapid active power balance and frequency stability under dynamic conditions. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to provide a method for controlling the active power balance of coal-fired power units coupled with multi-element energy storage. By constructing a hierarchical inertia and multi-stage frequency regulation collaborative model, adopting multi-scale online feature extraction and adaptive power allocation, segmented optimization scheduling and closed-loop online parameter update, the method can achieve rapid response, smooth connection and precise optimization of frequency support and economic scheduling in scenarios with high wind power penetration.

[0004] To achieve the above object, the present invention provides the following technical solutions: A method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage comprises the following steps: Step 1: Construct a synergistic characteristic model: Based on the power dynamic characteristics of coal-fired power units and energy storage units, establish an equivalent mathematical model with hierarchical inertia and multi-stage frequency regulation capabilities; Step 2: Online calculation of dominant characteristics: Based on the power system frequency deviation and the real-time output measurements of coal-fired power units and energy storage units, the dominant dynamic characteristics are extracted and the power allocation coefficients of each coordinated unit are calculated; Step 3: Segmented Optimal Scheduling: Build a segmented optimal scheduling model that takes into account both power constraints and economic operation objectives, and solve the combined output plan of coal-fired power units and energy storage units; Step 4: Closed-loop parameter update: Based on real-time load fluctuations and frequency deviations, perform online closed-loop adjustments on the equivalent inertia coefficient, frequency modulation slope, response time parameters, and power distribution coefficient in the equivalent mathematical model of step 1. In step three, the segmented optimization scheduling model constructs a sub-model containing the cost function and energy storage charging and discharging constraints for the scheduling period. Within this sub-model, the output distribution of coal-fired power units and energy storage units is iteratively calculated through gradient descent. The joint output plan for the scheduling period is output according to the convergence criterion, and the gradient step size is dynamically adjusted. Each scheduling period is executed and connected in sequence.

[0005] It should be noted that, first, the present invention constructs an equivalent mathematical model with graded inertia and multi-stage frequency regulation capabilities to achieve inertial response to the units and energy storage devices. Second, a segmented optimization scheduling model with economic cost and energy storage charging and discharging constraints is constructed for each scheduling period, and the joint output plan is solved through dynamic step size adjustment and accelerated gradient iteration. Finally, based on recursive least squares and sliding window algorithms, the equivalent inertia coefficient, frequency regulation slope, and distribution coefficient in the model are identified and updated in real time in a closed loop to keep the parameters consistent with the operating conditions. This invention overcomes the shortcomings of the existing technology, which is limited to static active safety domain assessment and cannot characterize the multi-stage frequency response of the system.

[0006] As a further solution of the present invention, in step three, the segmented optimization scheduling model uses the combined output result of the previous period of each scheduling period as the hot start initial value, and sets a sliding overlapping time window in the sub-model; the economic cost function and the frequency deviation penalty term are introduced in parallel in the sub-model objective function, and the weights of the economic cost function and the frequency deviation are adaptively adjusted according to the real-time frequency difference ratio; an increasing / decreasing tightening strategy is adopted for the upper and lower charging and discharging limits of the energy storage unit, and the boundary change rate is set according to the current state of charge; and a smooth transition correction is performed between the combined output plans of adjacent scheduling periods.

[0007] It should be noted that the present invention introduces the following technical features in step three: the combined output result of the previous period is used as the initial value for hot start, and a sliding overlapping time window is set to ensure the continuity of the scheme; the economic cost and frequency deviation penalty terms are included in the sub-model objective function in parallel, and the weights of the two are dynamically adjusted according to the real-time frequency deviation ratio; the upper and lower limits of the charge and discharge of the energy storage unit adopt an increasing / decreasing tightening strategy based on the current state of charge to match the actual constraints; at the same time, a smooth transition correction is performed between the outputs of adjacent scheduling periods to avoid sudden changes in output, thereby realizing the dynamic connection and optimized scheduling of the combined output of multiple periods, overcoming the shortcomings of the existing technology that only quantifies the active margin based on the static safety domain boundary and fails to consider the continuity and dynamic coupling between the scheduling periods.

[0008] As a further solution of the present invention, when the installed capacity of wind power accounts for 50% to 70% of the total power generation capacity, the installed capacity of the energy storage unit accounts for 10% to 20% of the total power generation capacity, and the wind power output fluctuation rate exceeds ±15% / min, the sliding overlapping time window length of the segmented optimization scheduling model is 60 seconds, the initial value of the hot start is the average value of the joint output at the end of the previous period in the last 10 seconds, the frequency deviation penalty weight is calculated based on the real-time maximum drop rate of 0.2 Hz / s, and is doubled when the drop rate exceeds 0.2 Hz / s, the tightening rate of the upper and lower limits of the energy storage unit charge and discharge is 10% of the current state of charge, and the smooth transition correction of the joint output plan of adjacent scheduling periods adopts three-point linear interpolation.

[0009] It should be noted that when the proportion of wind power exceeds 50%, the inertia of traditional synchronous machines is largely replaced, and the initial frequency of the system drops faster; although 10% to 20% of energy storage installed capacity can provide frequency regulation support, its capacity and power release rate need to be finely constrained, otherwise it will be difficult to cope with short-term severe fluctuations exceeding ±15% / min; in a high-speed fluctuation environment, it is necessary to quickly suppress frequency drops and avoid sudden changes in dispatch instructions that affect equipment safety and market economy; in highly fluctuating conditions, the dispatch model switches frequently. If there is no overlapping window and smooth transition, it is easy to cause output jumps or "frame drops". A 60-second sliding overlapping time window eliminates transient jumps when switching between different sub-models, ensuring continuous output in highly volatile environments. The 10-second average value at the end of the previous period is used as the initial value, which can quickly respond to the current operating conditions without introducing excessive initial errors. When the frequency drop exceeds 0.2Hz / s, the penalty weight is doubled to enhance the frequency modulation response during severe fluctuations. A 10% tightening rate ensures that the energy storage constraints are dynamically matched with changes in SOC to avoid out-of-bounds operation. Three-point linear interpolation can seamlessly connect the output solutions in adjacent time periods.

[0010] As a further solution of the present invention, in step 1, the collaborative characteristic model forms an equivalent mathematical model through the following interconnected submodules: Cascaded inertia submodule: This consists of three to five inertia links connected in series, each with independently adjustable inertia and damping to simulate the multi-level inertial response of coal-fired power units and energy storage units; Parallel FM submodule: Three gain channels are arranged in parallel after each inertial link to handle primary, secondary, and tertiary FM at different time scales. Each channel is individually configured with response amplitude and delay. Derivative feedback submodule: sets a frequency change rate feedback path between the inertial cascade chain and the parallel frequency modulation channel to separate the inertial support and frequency modulation actions; Energy storage constraint mapping submodule: This module uses the upper and lower limits of the state of charge and the maximum charge and discharge power of each energy storage unit as model input constraints and maps them to the corresponding state variables. Parameter identification submodule: Based on the recursive least squares optimal estimation algorithm, frequency deviation, power response and state of charge data are regularly collected, and the parameters of each inertial link and frequency modulation channel are updated online.

[0011] It should be noted that in step 1, the present invention refines the synergistic characteristic model into five interconnected submodules: the cascade inertia submodule simulates the multi-level inertial support of the unit and energy storage through three to five series inertial links; the parallel frequency modulation submodule arranges primary, secondary, and tertiary gain channels after each inertial link to cover frequency modulation actions of different time scales; the derivative feedback submodule establishes a frequency change rate path to achieve decoupling of the inertia and frequency modulation functions; the energy storage constraint mapping submodule maps the upper and lower limits of the state of charge and the charge and discharge power as model input constraints; and the parameter identification submodule updates the parameters of each submodule online based on a recursive least squares algorithm. This modular structure can accurately characterize the multi-stage frequency support process and keep the model parameters synchronized with the actual operating conditions, compensating for the shortcomings of existing static evaluation methods in terms of dynamic response and parameter adaptability.

[0012] As a further solution of the present invention, the process of extracting the dominant dynamic characteristics and calculating the power allocation coefficient of each collaborative unit in step 2 includes: Step 21: Perform multi-scale decomposition processing on the power system frequency and output signals to obtain the dominant response components representing inertial support, primary frequency regulation, and secondary frequency regulation respectively; Step 22: identifying the corresponding modal energy weight based on the amplitude time series and energy spectrum characteristics of each dominant response component; Step 23: The modal energy weight is integrated with the real-time adjustable power capacity of the coal-fired power unit and the energy storage unit according to an adaptive mapping rule to generate a preliminary power allocation coefficient; Step 24 : Dynamically normalize the sub-allocation coefficients and perform boundary correction based on the upper and lower limit constraints of the remaining capacity of each unit to output the final power allocation coefficients.

[0013] As a further solution of the present invention, step 21 uses continuous wavelet transform to perform multi-scale decomposition of the frequency deviation and output signal, the mother wavelet used is Morlet wavelet, and three scales of 0.5 seconds, 2 seconds, and 8 seconds are set to correspond to different time constants; step 22 applies Hilbert transform to each scale component to obtain the instantaneous amplitude envelope for subsequent energy analysis; step 23 calculates the spectral entropy based on the amplitude envelope of each scale to characterize the energy distribution weight of different regulation stages; step 24 uses Mamdani fuzzy reasoning to integrate the spectral entropy weights of each scale with the real-time adjustable capacity of the unit and the energy storage unit to generate a preliminary distribution coefficient, and applies a five-second window Savitzky-Golay filter to the preliminary distribution coefficient, and outputs it as the final power distribution coefficient after smoothing.

[0014] It should be noted that in step 2, the present invention proposes: first, using Morlet wavelets to perform multi-scale decomposition of the frequency deviation and output signal at three scales, 0.5s, 2s, and 8s, to obtain the dominant response components of each stage; then, using Hilbert transform to extract the instantaneous amplitude envelope and calculate spectral entropy to characterize the energy distribution weights under different time constants; then, using Mamdani-type fuzzy reasoning, the spectral entropy weights are integrated with the real-time adjustable capacity of the units and energy storage units according to adaptive mapping rules to generate preliminary power allocation coefficients; finally, the preliminary coefficients are smoothed using a five-second window Savitzky-Golay filter to output the final power allocation coefficients, thereby achieving real-time quantification and dynamic allocation of the multi-stage frequency support process. The present invention overcomes the shortcomings of the prior art, which only evaluates the steady-state active power margin of the system through the static safety domain boundary, but fails to quantify the staged response characteristics of wind power and energy storage during frequency disturbances, ignores the time-varying impact of inertial support, primary frequency regulation, and secondary frequency regulation on power allocation, and cannot achieve real-time adaptive allocation for random payload mutations.

[0015] As a further solution of the present invention, in the segmented optimization scheduling model of step 3, the accelerated gradient algorithm is used for the gradient descent iterative process of each sub-model, specifically including: In the first iteration, only the current gradient information is used as the update direction; From the second iteration onwards, the current gradient is linearly superimposed with the previous update direction at a ratio of 0.8 as the new search direction; Before each iteration, the momentum coefficient is fine-tuned within the range of ±0.05 based on the difference between the current objective function value of the sub-model and the objective function value of the previous iteration; When the improvement of the objective function for three consecutive iterations is less than 0.001% or the number of iterations exceeds 50, the iteration of this sub-model is terminated and the results are output.

[0016] It should be noted that the present invention introduces an accelerated gradient algorithm into the segmented optimization scheduling model of step three: the current gradient is used in the first iteration, and the subsequent iterations linearly superimpose the current gradient with the previous round of update direction at 0.8, and fine-tune the momentum coefficient ±0.05 based on the difference between the current and previous objective function values ​​before each iteration to smooth the search path and accelerate convergence; at the same time, a termination criterion of three consecutive improvements of less than 0.001% or more than 50 iterations is set to ensure that each sub-model ends efficiently while meeting the accuracy requirements. The present invention overcomes the shortcomings of the prior art that segmented optimization only uses conventional gradient descent, cannot take into account both iterative convergence speed and directional stability, and is prone to slow convergence or frequent oscillations in high-volatility scenarios.

[0017] As a further solution of the present invention, the economic cost term and the frequency deviation penalty term are set in parallel in the sub-model objective function of step 3, and the frequency deviation penalty weight is dynamically adjusted in the following manner: The maximum frequency drop of the system in the last 30 seconds is sampled and averaged with a period of 5 seconds to generate the real-time maximum drop rate value; The falling speed value is multiplied by the preset proportional coefficient as the frequency penalty weight; During the same period, the weight of the economic cost item was set at 1.2 times the proportion of wind power installed capacity; The frequency penalty weight is reset every 40 seconds.

[0018] As a further solution of the present invention, in the segmented optimization scheduling model of step three, when the accelerated gradient algorithm is used for the gradient descent iterative process of each sub-model, the initial value of the gradient step is set to 0.1. After each iteration is completed, the step size is adaptively decayed based on the local second-order derivative / inverse gradient norm of the objective function at the current iteration point. When the iterative step size drops below 0.01 or the improvement of the objective function value in consecutive iterations is less than 0.0001, the convergence condition is met and the iterative process is terminated immediately.

[0019] It should be noted that the present invention sets the economic cost term and the frequency deviation penalty term in parallel in the sub-model objective function of step three, and uses a 5-second period to perform a sliding average calculation of the maximum speed drop in the last 30 seconds to calculate the real-time speed drop value, which is multiplied by a preset proportional coefficient to generate a dynamic frequency penalty weight. At the same time, the economic cost term weight is set to 1.2 times the proportion of wind power installed capacity and reset every 40 seconds. In addition, the present invention sets the initial step size to 0.1 in the segmented gradient optimization iteration, and adaptively decays after each iteration based on the local second-order derivative or the inverse of the gradient norm of the current objective function. When the step size is less than 0.01 or the improvement amplitude of the continuous iteration is less than 0.0001, it is immediately terminated to ensure that the optimization process is both efficient and stable in high-fluctuation scenarios. The present invention overcomes the shortcomings of the existing technology that only quantifies the active power margin in a static evaluation framework, fails to closely couple economic dispatch with the secondary frequency response, and lacks an adaptive acceleration mechanism for the gradient optimization process.

[0020] As a further solution of the present invention, the closed-loop parameter update process in step 4 includes: Perform rolling sampling of real-time frequency deviation and power response data using a 10-second sliding window; A recursive least squares algorithm with a forgetting factor of 0.98 is used to online identify the equivalent inertia coefficient, primary / secondary / tertiary frequency modulation slope, and power distribution coefficient in the equivalent mathematical model. An update is triggered every 5 seconds, and the parameters obtained from this identification are exponentially weighted and fused with the results of the previous update at a ratio of 70%:30%; The parameters obtained after fusion are limited to the range of ±10% of the previous updated value; A first-order exponential smoothing filter with a time constant of 0.5 seconds is applied to the above fusion results, and then loaded into the equivalent mathematical model for the next control cycle.

[0021] It should be noted that the present invention proposes a closed-loop parameter update process in step 4: using a 10-second sliding window to roll sample frequency deviation and power response data, a recursive least squares algorithm with a forgetting factor of 0.98 is used to online identify the equivalent inertia coefficient, frequency modulation slope at each level, and power allocation coefficient. An update is triggered every 5 seconds and exponentially weighted fusion is performed with the previous result at a ratio of 70%:30%. The fusion parameters are limited to a range of ±10%. Finally, a first-order exponential smoothing filter with a time constant of 0.5 seconds is applied and reloaded into the equivalent mathematical model to ensure that the model parameters are synchronized with the actual operating conditions. The present invention overcomes the shortcomings of the existing technology that uses offline fixed parameters for safety margin assessment and cannot adapt to the dynamic characteristics changes of coal-fired power units and energy storage units caused by changes in load and wind power output in high-fluctuation and high-penetration scenarios.

[0022] The technical effects of the present invention's method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage are as follows: The present invention constructs a synergistic characteristic equivalent model with hierarchical inertia and multi-stage frequency regulation capabilities, adopts multi-scale dynamic feature extraction and adaptive power allocation, and realizes rapid response in each stage of inertial support, primary / secondary / tertiary frequency regulation; through segmented optimization scheduling and dynamic adjustment of frequency / economic dual weights, a continuous and smooth joint output plan is output; through online recursive least squares identification and closed-loop parameter update, the model parameters are corrected in real time to fit the operating conditions; and with the help of sliding overlapping windows and multi-point interpolation transition, the output mutation during sub-model switching is eliminated, so that in scenarios with high wind power penetration and high fluctuation, the frequency drop of the power system is significantly reduced and the recovery time is greatly shortened, while taking into account both operating economy and energy storage safety, achieving multiple improvements in frequency stability, economic scheduling efficiency and output continuity. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Figure 1 A flow chart of the method proposed in the present invention; Figure 2 This is a process diagram for extracting the dominant dynamic characteristics and calculating the power allocation coefficient of each collaborative unit in step 2 of the present invention; Figure 3 A control interface diagram for applying the method proposed in the present invention. DETAILED DESCRIPTION

[0024] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0025] Example 1. Figure 1 As shown, the present invention proposes a method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage, comprising the following steps: Step 1: Construct a synergistic characteristic model: Based on the power dynamic characteristics of coal-fired power units and energy storage units, establish an equivalent mathematical model with hierarchical inertia and multi-stage frequency regulation capabilities; Step 2: Online calculation of dominant characteristics: Based on the power system frequency deviation and the real-time output measurements of coal-fired power units and energy storage units, the dominant dynamic characteristics are extracted and the power allocation coefficients of each coordinated unit are calculated; Step 3: Segmented Optimal Scheduling: Build a segmented optimal scheduling model that takes into account both power constraints and economic operation objectives, and solve the combined output plan of coal-fired power units and energy storage units; Step 4: Closed-loop parameter update: Based on real-time load fluctuations and frequency deviations, perform online closed-loop adjustments on the equivalent inertia coefficient, frequency modulation slope, response time parameters, and power distribution coefficient in the equivalent mathematical model of step 1. In step three, the segmented optimization scheduling model constructs a sub-model containing the cost function and energy storage charging and discharging constraints for the scheduling period. Within this sub-model, the output distribution of coal-fired power units and energy storage units is iteratively calculated through gradient descent. The joint output plan for the scheduling period is output according to the convergence criterion, and the gradient step size is dynamically adjusted. Each scheduling period is executed and connected in sequence.

[0026] It should be specifically noted that in step three, the segmented optimization scheduling model uses the combined output result of the previous period of each scheduling period as the hot start initial value, and sets a sliding overlapping time window in the sub-model; the economic cost function and the frequency deviation penalty term are introduced in parallel in the sub-model objective function, and the weights of the economic cost function and the frequency deviation are adaptively adjusted according to the real-time frequency difference ratio; an increasing / decreasing tightening strategy is adopted for the upper and lower limits of charging and discharging of the energy storage unit, and the boundary change rate is set according to the current state of charge; and a smooth transition correction is performed between the combined output plans of adjacent scheduling periods.

[0027] In addition, when the installed capacity of wind power accounts for 50% to 70% of the total power generation capacity, the installed capacity of energy storage units accounts for 10% to 20% of the total power generation capacity, and the wind power output fluctuation rate exceeds ±15% / min, the sliding overlapping time window length of the segmented optimization scheduling model is 60 seconds, the initial value of the hot start is the average value of the joint output at the end of the previous period in the last 10 seconds, the frequency deviation penalty weight is calculated based on the real-time maximum drop rate of 0.2 Hz / s, and is doubled when the drop rate exceeds 0.2 Hz / s. The tightening rate of the upper and lower limits of the energy storage unit charge and discharge is 10% of the current state of charge, and the smooth transition correction of the joint output plan in adjacent scheduling periods adopts three-point linear interpolation.

[0028] It should be noted that in the segmented optimization scheduling model of step 3, the accelerated gradient algorithm is used for the gradient descent iterative process of each sub-model, specifically including: In the first iteration, only the current gradient information is used as the update direction; From the second iteration onwards, the current gradient is linearly superimposed with the previous update direction at a ratio of 0.8 as the new search direction; Before each iteration, the momentum coefficient is fine-tuned within the range of ±0.05 based on the difference between the current objective function value of the sub-model and the objective function value of the previous iteration; When the improvement of the objective function for three consecutive iterations is less than 0.001% or the number of iterations exceeds 50, the iteration of this sub-model is terminated and the results are output.

[0029] It should be noted that the economic cost term and the frequency deviation penalty term are set in parallel in the sub-model objective function of step 3, and the frequency deviation penalty weight is dynamically adjusted as follows: The maximum frequency drop of the system in the last 30 seconds is sampled and averaged with a period of 5 seconds to generate the real-time maximum drop rate value; The falling speed value is multiplied by the preset proportional coefficient as the frequency penalty weight; During the same period, the weight of the economic cost item was set at 1.2 times the proportion of wind power installed capacity; The frequency penalty weight is reset every 40 seconds.

[0030] It should be noted that in the segmented optimization scheduling model of step three, when the accelerated gradient algorithm is used for the gradient descent iterative process of each sub-model, the initial value of the gradient step is set to 0.1. After each iteration is completed, the step size is adaptively decayed based on the local second-order derivative / inverse gradient norm of the objective function at the current iteration point. When the iterative step size drops below 0.001 or the improvement of the objective function value in consecutive iterations is less than 0.0001, the convergence condition is met and the iterative process is terminated immediately.

[0031] In order to clearly describe the implementation scheme of the above technical solution, its technical effects compared with the prior art are illustrated by the following examples.

[0032] In a power generation system with a system scale of 1000MW, there are 250MW coal-fired power units, 600MW wind power units, and 150MW energy storage units. At 10s, the power system load steps up 80MW within 5s. Two technical solutions are simulated using MATLAB. Solution 1 adopts the baseline technical solution, which is traditional segmented optimization scheduling, without hot start, no sliding window, a fixed penalty weight of 0.3, no tightening strategy, and no interpolation solution. Solution 2 adopts the technical solution of the present invention, adopts hot start, 60s sliding overlapping window, doubles the real-time speed drop threshold, and 10% tightening rate Three-point linear interpolation. In step three, the hot start takes the average value of the combined output at the end of the previous period for the initial value of each scheduling period. For the sliding overlapping window, the sub-model time window length is 60 seconds, and the window overlap is 30 seconds to maintain the continuity of the scheme. For the dynamic penalty weight, the maximum speed drop in the last 30 seconds is calculated in a 5-second cycle. When the speed drop is greater than 0.2Hz / s, the weight value doubles. For the tightening strategy, when the charge state of the energy storage unit is 50%, the upper and lower limits of charge and discharge are tightened by 10%. For smooth interpolation, three points are taken at the end and beginning of each of the three adjacent periods, and linear interpolation is used for seamless transition. The simulation results of the two technical solutions are compared, as shown in Table 1: Table 1 Comparison of simulation results of Scheme 1 and Scheme 2

[0033] As can be seen from the table, Scheme 2 significantly outperforms the baseline in frequency drop, RoCoF, and recovery speed, improving both output mutation and SoC lower limit with only a slight increase in economic cost. The above examples fully demonstrate the synergistic effect of the various technical features in step 3. The average combined output at the end of the previous period for the last 10 seconds is used as the initial value for the next period iteration, significantly narrowing the optimization search space, enabling faster convergence of gradient iterations and reducing initial power jumps. The 60-second, 30-second overlapping time window design ensures output continuity between adjacent sub-models, avoiding power command gaps or mutations caused by model switching. The penalty coefficient is dynamically adjusted based on the real-time maximum drop speed, automatically strengthening frequency compensation when the drop speed exceeds 0.2 Hz / s. This can promptly improve the frequency response strength under large disturbance conditions. A 10% state-of-charge tightening is implemented on the upper and lower limits of energy storage charging and discharging to ensure that the energy storage device does not operate beyond the limit during rapid fluctuations, while retaining a certain margin for subsequent frequency support. Three-point linear interpolation is used to achieve seamless connection between output schemes in adjacent scheduling periods, effectively suppressing command mutations at period boundaries and improving the smoothness of the overall output curve.

[0034] Example 2. The difference between Example 2 of the present invention and Example 1 is that this example introduces steps 1, 2 and 4 of a method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage.

[0035] In step 1 of the technical solution of the present invention, the collaborative characteristic model forms an equivalent mathematical model through the following interconnected sub-modules: Cascaded inertia submodule: This consists of three to five inertia links connected in series, each with independently adjustable inertia and damping to simulate the multi-level inertial response of coal-fired power units and energy storage units; Parallel FM submodule: Three gain channels are arranged in parallel after each inertial link to handle primary, secondary, and tertiary FM at different time scales. Each channel is individually configured with response amplitude and delay. Derivative feedback submodule: sets a frequency change rate feedback path between the inertial cascade chain and the parallel frequency modulation channel to separate the inertial support and frequency modulation actions; Energy storage constraint mapping submodule: This module uses the upper and lower limits of the state of charge and the maximum charge and discharge power of each energy storage unit as model input constraints and maps them to the corresponding state variables. Parameter identification submodule: Based on the recursive least squares optimal estimation algorithm, frequency deviation, power response and state of charge data are regularly collected, and the parameters of each inertial link and frequency modulation channel are updated online.

[0036] In step 1 of the present invention, the construction of the synergistic characteristic model is carried out according to the following process. Through the layer-by-layer superposition and mutual coupling of five interconnected submodules, a complete equivalent mathematical model is formed and the technical problems of insufficient system inertia, lack of frequency modulation resources and rapid changes in constraints under high wind and solar penetration are solved: the cascade inertia submodule first receives the frequency deviation signal of the power system, and passes through three to five series inertia links in turn. Each link filters and amplifies the frequency deviation with independently set inertia coefficients and damping coefficients to generate a short-term inertial support output that decays layer by layer to compensate for the lack of system inertia caused by wind power replacing synchronous machines; the parallel frequency modulation submodule connects three gain channels in parallel immediately after each inertia link, corresponding to the primary, secondary and tertiary frequency modulation respectively, and performs active adjustment of different amplitudes and delays on the primary inertial support signal on the short, medium and long time scales to ensure continuous and orderly power compensation after the initial inertial response, in the middle of the frequency drop and at the end of the recovery; the derivative feedback submodule is connected between the cascade inertia chain and the parallel A frequency rate-of-change measurement loop is arranged in parallel between the frequency modulation channels, feeding the frequency derivative signal back to the corresponding inertia link and frequency modulation gain channel. This allows for rapid differentiation between the inertia and frequency modulation support mechanisms, and dynamically adjusts the response strength and timing of each channel. The energy storage constraint mapping submodule reads the upper and lower limits of the state of charge and the maximum charge / discharge power of each energy storage unit in real time, maps these constraints into the model's input boundaries, and associates them with the available output of each inertia and frequency modulation channel. This automatically limits or relaxes the output of each channel to address fluctuations in available capacity caused by rapid changes in state of charge. The parameter identification submodule periodically collects system frequency deviation, unit and energy storage output, and state of charge data within a 10-second sliding window. A recursive least squares optimal estimation algorithm with a forgetting factor of 0.98 is used to online identify the inertia coefficient, frequency modulation slope at each level, and channel delay. Every 5 seconds, the new and old parameters are weighted and fused in a 70%:30% ratio, and then updated to the above submodules to ensure that the model parameters are synchronized with actual operating conditions.

[0037] like Figure 2 As shown, the process of extracting the dominant dynamic characteristics and calculating the power allocation coefficient of each collaborative unit in step 2 of the technical solution of the present invention includes: Step 21: Perform multi-scale decomposition processing on the power system frequency and output signals to obtain the dominant response components representing inertial support, primary frequency regulation, and secondary frequency regulation respectively; Step 22: identifying the corresponding modal energy weight based on the amplitude time series and energy spectrum characteristics of each dominant response component; Step 23: The modal energy weight is integrated with the real-time adjustable power capacity of the coal-fired power unit and the energy storage unit according to an adaptive mapping rule to generate a preliminary power allocation coefficient; Step 24 : Dynamically normalize the sub-allocation coefficients and perform boundary correction based on the upper and lower limit constraints of the remaining capacity of each unit to output the final power allocation coefficients.

[0038] Through multi-scale decomposition, the dominant response components of each stage of inertial support, primary frequency regulation and secondary frequency regulation are accurately extracted, and the modal energy weights are identified based on the amplitude time series and energy spectrum characteristics, which can achieve quantitative stratification of power demand throughout the frequency disturbance process; these weights are then adaptively integrated with the real-time adjustable capacity of the units and energy storage to generate preliminary allocation coefficients, and dynamic normalization and boundary correction are performed in combination with the upper and lower limits of the remaining capacity, which can match the available output of each collaborative unit in real time to avoid excessive or insufficient resource calls; the final output power allocation coefficient has fast response, high precision and constraint compliance, effectively improving the accuracy and stability of the system frequency support, and reducing scheduling errors and energy storage out-of-bounds risks caused by static allocation.

[0039] It should be specifically noted that step 21 uses continuous wavelet transform to perform multi-scale decomposition of the frequency deviation and output signal. The mother wavelet used is Morlet wavelet, and three scales of 0.5 seconds, 2 seconds, and 8 seconds are set to correspond to different time constants; step 22 applies Hilbert transform to each scale component to obtain the instantaneous amplitude envelope for subsequent energy analysis; step 23 calculates the spectral entropy based on the amplitude envelope of each scale to characterize the energy distribution weights in different regulation stages; step 24 uses Mamdani fuzzy reasoning to integrate the spectral entropy weights of each scale with the real-time adjustable capacity of the unit and energy storage unit to generate a preliminary distribution coefficient, and applies a five-second window Savitzky-Golay filter to the preliminary distribution coefficient, and outputs it as the final power distribution coefficient after smoothing.

[0040] By performing Morlet wavelet decomposition on the frequency deviation and output signal at three scales of 0.5s, 2s, and 8s, the dominant response components corresponding to the inertial support, primary frequency modulation, and secondary frequency modulation stages are extracted respectively. The instantaneous amplitude envelope of each component is then obtained using the Hilbert transform, and its spectral entropy is quantified to clearly characterize the dynamic energy distribution in different regulation stages. The energy weight of each stage is then adaptively integrated with the adjustable capacity of the unit and energy storage through Madani-type fuzzy reasoning to generate an accurate preliminary allocation coefficient. The coefficient is then smoothed through a five-second window Savitzky-Golay filter, enabling real-time and hierarchical allocation of active power support, significantly improving the response speed and allocation accuracy to large disturbances and random fluctuations, while suppressing command mutations and noise interference, ensuring the continuity, stability, and robustness of active power balance control.

[0041] In step 4 of the technical solution of the present invention, the closed-loop parameter update process includes: Perform rolling sampling of real-time frequency deviation and power response data using a 10-second sliding window; A recursive least squares algorithm with a forgetting factor of 0.98 is used to online identify the equivalent inertia coefficient, primary / secondary / tertiary frequency modulation slope, and power distribution coefficient in the equivalent mathematical model. An update is triggered every 5 seconds, and the parameters obtained from this identification are exponentially weighted and fused with the results of the previous update at a ratio of 70%:30%; The parameters obtained after fusion are limited to the range of ±10% of the previous updated value; A first-order exponential smoothing filter with a time constant of 0.5 seconds is applied to the above fusion results, and then loaded into the equivalent mathematical model for the next control cycle.

[0042] It should be noted that the closed-loop update process described above is used in step 4 to ensure that the model parameters can track the rapid changes in system operating conditions in real time and suppress measurement noise: first, the 10-second sliding window rolling sampling can capture the dynamic characteristics of frequency deviation and output response on a short time scale; second, the recursive least squares algorithm with a forgetting factor of 0.98 can not only improve the identification accuracy by utilizing historical data, but also quickly "forget" outdated information, realizing online adaptation of parameters such as system inertia and frequency modulation slope; third, triggering updates every 5 seconds and fusing new and old parameters in a 70%:30% ratio can balance the identification convergence speed and result stability; then, the parameters are limited to within ±10% to avoid drastic parameter jumps caused by transient disturbances; finally, a 0.5-second exponential smoothing filter is applied to further filter out high-frequency noise, ensuring smooth transition and stable convergence of the updated parameters, thereby maintaining the equivalent mathematical model's high-precision description of the actual operating state and continuous and reliable active power balance control.

[0043] like Figure 3 As shown, by applying the solution proposed in the present invention and driven by this method, the control interface centrally displays the core components and operating status of the closed-loop online parameter update module of the present invention through a visual interface: the "Optimization Parameter Configuration" area on the left can set the initial value of the gradient step, convergence threshold, parameter fusion ratio and limit range; the "Parameter Evolution Curve" in the middle plots the dynamic changes of the equivalent inertia coefficient and the primary / secondary / tertiary frequency modulation slope in real time; the "Parameter Monitoring" panel at the bottom displays a 5-second update cycle, 98% forgetting factor, 10-second window and ±10% limit amplitude, and realizes online parameter debugging and early warning monitoring through the iteration progress bar and status indicators such as "normal", "updating" and "warning".

[0044] In summary, combining Examples 1 and 2, this invention achieves rapid multi-stage frequency support and smooth active power allocation for wind power and energy storage in scenarios with high wind power penetration (50%-70%) and high power fluctuation (>±15% / min) by constructing a hierarchical inertia and multi-stage frequency regulation collaborative model, introducing multi-scale online feature extraction, segmented optimized scheduling, and closed-loop parameter updates. In Example 1, compared with the traditional baseline solution, this invention increases the system's lowest frequency from 49.48Hz to 49.62Hz, reduces the maximum RoCoF from 0.35Hz / s to 0.22Hz / s, shortens the frequency recovery time from 18.7s to 11.3s, reduces the peak output surge by over 60%, and increases the minimum energy storage system-on-chip (SoC) from 42.1% to 45.3%, with only a slight increase in economic costs of RMB 0.003 million per second. Example 2 further demonstrates that online recursive least squares identification and parameter fusion can maintain these performance characteristics across various system sizes and load disturbance conditions. It can be seen that the present invention effectively overcomes the shortcomings of existing static evaluation methods in dynamic response, continuity and parameter adaptation, and achieves a comprehensive improvement in frequency stability, output smoothness and economic dispatch efficiency.

[0045] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0046] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage, characterized in that: The steps include: Step 1: Construct a synergistic characteristic model: Based on the power dynamic characteristics of coal-fired power units and energy storage units, establish an equivalent mathematical model with hierarchical inertia and multi-stage frequency regulation capabilities; Step 2: Online calculation of dominant characteristics: Based on the power system frequency deviation and the real-time output measurements of coal-fired power units and energy storage units, the dominant dynamic characteristics are extracted and the power allocation coefficients of each coordinated unit are calculated; Step 3: Segmented Optimal Scheduling: Build a segmented optimal scheduling model that takes into account both power constraints and economic operation objectives, and solve the combined output plan of coal-fired power units and energy storage units; Step 4: Closed-loop parameter update: Based on real-time load fluctuations and frequency deviations, perform online closed-loop adjustments on the equivalent inertia coefficient, frequency modulation slope, response time parameters, and power distribution coefficient in the equivalent mathematical model of step 1. In step three, the segmented optimization scheduling model constructs a sub-model containing the cost function and energy storage charging and discharging constraints for the scheduling period. Within this sub-model, the output distribution of coal-fired power units and energy storage units is iteratively calculated through gradient descent. The joint output plan for the scheduling period is output according to the convergence criterion, and the gradient step size is dynamically adjusted. Each scheduling period is executed and connected in sequence.

2. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 1, characterized in that: In step three, the segmented optimization scheduling model uses the combined output results of the previous period of each scheduling period as the hot start initial value, and sets a sliding overlapping time window in the sub-model; the economic cost function and the frequency deviation penalty term are introduced in parallel in the sub-model objective function, and the weights of the economic cost function and the frequency deviation are adaptively adjusted according to the real-time frequency difference ratio; an increasing / decreasing tightening strategy is adopted for the upper and lower limits of charging and discharging of the energy storage unit, and the boundary change rate is set according to the current state of charge; and a smooth transition correction is performed between the combined output plans of adjacent scheduling periods.

3. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 2, characterized in that: When the installed capacity of wind power accounts for 50% to 70% of the total power generation capacity, the installed capacity of energy storage units accounts for 10% to 20% of the total power generation capacity, and the wind power output fluctuation rate exceeds ±15% / min, the sliding overlapping time window length of the segmented optimization scheduling model is 60 seconds, the initial value of the hot start is the average value of the joint output at the end of the previous period in the last 10 seconds, the frequency deviation penalty weight is calculated based on the real-time maximum drop rate of 0.2 Hz / s, and is doubled when the drop rate exceeds 0.2 Hz / s. The tightening rate of the upper and lower limits of the energy storage unit charge and discharge is 10% of the current state of charge. The smooth transition correction of the joint output plan in adjacent scheduling periods adopts three-point linear interpolation.

4. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 1, characterized in that: In step 1, the collaborative characteristic model forms an equivalent mathematical model through the following interconnected sub-modules: Cascaded inertia submodule: This consists of three to five inertia links connected in series, each with independently adjustable inertia and damping to simulate the multi-level inertial response of coal-fired power units and energy storage units; Parallel FM submodule: Three gain channels are arranged in parallel after each inertial link to handle primary, secondary, and tertiary FM at different time scales. Each channel is individually configured with response amplitude and delay. Derivative feedback submodule: sets a frequency change rate feedback path between the inertial cascade chain and the parallel frequency modulation channel to separate the inertial support and frequency modulation actions; Energy storage constraint mapping submodule: This module uses the upper and lower limits of the state of charge and the maximum charge and discharge power of each energy storage unit as model input constraints and maps them to the corresponding state variables. Parameter identification submodule: Based on the recursive least squares optimal estimation algorithm, frequency deviation, power response and state of charge data are regularly collected, and the parameters of each inertial link and frequency modulation channel are updated online.

5. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 1, characterized in that: The process of extracting the dominant dynamic characteristics and calculating the power allocation coefficient of each collaborative unit in step 2 includes: Step 21: Perform multi-scale decomposition processing on the power system frequency and output signals to obtain the dominant response components representing inertial support, primary frequency regulation, and secondary frequency regulation respectively; Step 22: identifying the corresponding modal energy weight based on the amplitude time series and energy spectrum characteristics of each dominant response component; Step 23: The modal energy weight is integrated with the real-time adjustable power capacity of the coal-fired power unit and the energy storage unit according to an adaptive mapping rule to generate a preliminary power allocation coefficient; Step 24 : Dynamically normalize the sub-allocation coefficients and perform boundary correction based on the upper and lower limit constraints of the remaining capacity of each unit to output the final power allocation coefficients.

6. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 5, characterized in that: In step 21, continuous wavelet transform is used to perform multi-scale decomposition of the frequency deviation and output signal. The mother wavelet used is Morlet wavelet, and three scales of 0.5 seconds, 2 seconds, and 8 seconds are set to correspond to different time constants. In step 22, Hilbert transform is applied to each scale component to obtain the instantaneous amplitude envelope for subsequent energy analysis. In step 23, spectral entropy is calculated based on the amplitude envelope of each scale to characterize the energy distribution weights in different regulation stages. In step 24, Mamdani fuzzy reasoning is used to integrate the spectral entropy weights of each scale with the real-time adjustable capacity of the unit and energy storage unit to generate a preliminary distribution coefficient. The preliminary distribution coefficient is filtered with a five-second window Savitzky-Golay filter and smoothed to be output as the final power distribution coefficient.

7. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 2, characterized in that: In the segmented optimization scheduling model of step 3, the accelerated gradient algorithm is used for the gradient descent iteration process of each sub-model, specifically including: In the first iteration, only the current gradient information is used as the update direction; From the second iteration onwards, the current gradient is linearly superimposed with the previous update direction at a ratio of 0.8 as the new search direction; Before each iteration, the momentum coefficient is fine-tuned within the range of ±0.05 based on the difference between the current objective function value of the sub-model and the objective function value of the previous iteration; When the improvement of the objective function for three consecutive iterations is less than 0.001% or the number of iterations exceeds 50, the iteration of this sub-model is terminated and the results are output.

8. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 7, characterized in that: In the sub-model objective function of step 3, the economic cost term and the frequency deviation penalty term are set in parallel, and the frequency deviation penalty weight is dynamically adjusted as follows: The maximum frequency drop of the system in the last 30 seconds is sampled and averaged with a period of 5 seconds to generate the real-time maximum drop rate value; The falling speed value is multiplied by the preset proportional coefficient as the frequency penalty weight; During the same period, the weight of the economic cost item was set at 1.2 times the proportion of wind power installed capacity; The frequency penalty weight is reset every 40 seconds.

9. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 7, characterized in that: In the segmented optimization scheduling model of step three, when the accelerated gradient algorithm is used for the gradient descent iterative process of each sub-model, the initial value of the gradient step is set to 0.

1. After each iteration is completed, the step size is adaptively decayed based on the local second-order derivative / inverse gradient norm of the objective function at the current iteration point. When the iterative step size drops below 0.001 or the improvement of the objective function value in consecutive iterations is less than 0.0001, the convergence condition is met and the iterative process is terminated immediately.

10. The method for controlling active power balance of a coal-fired power unit coupled with multi-element energy storage according to claim 1, characterized in that: The closed-loop parameter update process in step 4 includes: Perform rolling sampling of real-time frequency deviation and power response data using a 10-second sliding window; A recursive least squares algorithm with a forgetting factor of 0.98 is used to online identify the equivalent inertia coefficient, primary / secondary / tertiary frequency modulation slope, and power distribution coefficient in the equivalent mathematical model. An update is triggered every 5 seconds, and the parameters obtained from this identification are exponentially weighted and fused with the results of the previous update at a ratio of 70%:30%; The parameters obtained after fusion are limited to the range of ±10% of the previous updated value; A first-order exponential smoothing filter with a time constant of 0.5 seconds is applied to the fusion result, and then loaded into the equivalent mathematical model for the next control cycle.

Images