A peak regulation and frequency regulation heat supply method for multi-element energy storage coupled coal power units

Abstract

This invention relates to the field of power supply technology, specifically disclosing a peak-shaving and frequency-regulating heating method for multi-energy storage coupled coal-fired power units. This method addresses the problems of existing technologies that only optimize active power frequency, lack electrothermal coordination and state-of-charge constraints, and are insufficient in time delay compensation. The proposed method includes system modeling, controller design, coupling constraints, adaptive parameter tuning, delay compensation, and distributed consensus. This invention constructs a dual-domain power-thermal interconnection model, designs a cooperative controller based on delay-compensated distributed model predictive control, introduces electrothermal coordination penalties and state-of-charge constraints into the objective function, adaptively adjusts the prediction window and weight parameters, identifies and compensates for network delays online, and uses a consensus algorithm to achieve multi-regional power-thermal coordination optimization.

Description

Technical Field

[0001] This invention relates to the field of power supply technology, and in particular to a peak-shaving and frequency-regulating heating method for multi-energy storage coupled coal-fired power units. Background Technology

[0002] With the large-scale grid connection of renewable energy sources such as wind power and photovoltaics, power system frequency fluctuations have intensified, placing higher demands on frequency regulation resources. While traditional thermal power units possess advantages such as large rotational inertia and wide frequency regulation range, their response speed and regulation accuracy are relatively limited, making it difficult to meet the modern frequency regulation requirements of "fast adjustment, precision, and controllability." Energy storage systems, capable of millisecond-level charge-discharge switching and possessing strong bidirectional regulation capabilities, are considered an important means to improve the dynamic performance of the power grid. The publicly available CNKI document, "Research on Coordinated Control of Wide-Area Multi-Functional Energy Storage Based on Distributed Model Predictive Control," proposes for the first time, in a two-region interconnected power system, a DC-DMPC (Delay Compensation Distributed Model Predictive Control) algorithm framework for the coordinated control of distributed battery energy storage and pumped hydro storage. It constructs a two-region LFC mathematical model containing battery energy storage (BESS) and pumped hydro storage units (PSP), designs a distributed MPC (Model Predictive Control) controller, and utilizes ADMM (Alternating Direction Method of Control) to achieve coordinated control of distributed battery energy storage and pumped hydro storage. The Multipliers consensus algorithm is used to achieve power coordination between units. A simple delay compensation strategy is proposed for fixed communication delays, and simulation verification under typical operating conditions is completed in MATLAB. Although phased results have been achieved in framework design and algorithm verification, this only optimizes the active power frequency regulation and does not include the unit waste heat or heat network verification in the model, which is not conducive to the in-depth utilization of power-heat cogeneration resources. Furthermore, the SOC constraints of battery and pumped storage units are not added to the MPC objective function, which will lead to unit state of charge drift during long-term operation. The controller prediction window and weights are fixed and fail to adaptively adjust the MPC window length and weighting coefficients according to the fluctuation of new energy power and load changes. The delay compensation only adopts fixed delay compensation and lacks online estimation and dynamic adjustment mechanism for network jitter, packet loss and delay fluctuations, resulting in a large gap between the simulation environment and the actual deployment. Summary of the Invention

[0003] The technical problem to be solved by this invention is to provide a peak-shaving and frequency regulation heating method for multi-source energy storage coupled with coal-fired power units. By introducing electro-thermal coupling and SOC constraints into distributed model predictive control, supplemented by multi-condition adaptive prediction windows and weight adjustment, and online delay identification, the optimal peak-shaving and frequency regulation of multi-source energy storage and coal-fired power units under the DC-DMPC framework is achieved through the ADMM consensus algorithm.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for peak-shaving and frequency-regulating heating using multi-element energy storage coupled coal-fired power units includes the following steps:

[0006] A. Establish a two-region interconnected power-heat system model including battery energy storage units, pumped storage units, and heating network loads;

[0007] B. Design a distributed model predictive controller based on DC-DMPC, with frequency deviation, power balance, energy storage unit SOC and heating network temperature deviation as optimization objectives;

[0008] C. Introduce an electro-thermal coupling term and a storage unit SOC constraint into the MPC objective function;

[0009] D. The controller adaptively adjusts the prediction window length and weighting coefficients based on the output of new energy sources and load fluctuations;

[0010] E. A network jitter and packet loss dynamic delay identification strategy is adopted, and a delay compensation model is embedded in the prediction model;

[0011] F. Each area controller coordinates the power and heat among units through the ADMM consensus algorithm and uses the alternating direction multiplier method to decompose and solve the problem.

[0012] To significantly shorten the system's frequency response time under renewable energy disturbances, improve the discharge / charge efficiency of energy storage units, accurately reproduce the spatiotemporal dynamic processes of hydraulic-electromechanical systems and heating networks, and ensure effective control of system frequency stability and heating network temperature deviation under large-scale grid connection and sudden heat load conditions:

[0013] As a further aspect of this invention, in step A, based on the classic first-order RC circuit SOC-voltage model, combined with the field-calibrated nonlinear efficiency curve and temperature-dependent internal resistance, temperature, SOC, and power are dynamically coupled and introduced into the state equation to adaptively optimize discharge / charge losses and response speed under different operating conditions. This is then incorporated into a capacity decay model based on the cumulative cycle count and depth, with the decay rate set as the decision variable. To balance millisecond-level frequency regulation and minute-level energy dispatch requirements, a hybrid time-domain model is constructed using dual discrete step lengths of 0.1s and 60s. A dual-mode state equation is designed for both pumping and power generation modes, embedding a hydraulic-electromechanical coupling term based on changes in head. The system employs a first-order hysteresis pipeline water hammer approximation term to capture pressure oscillations and power fluctuations during sudden flow changes. Simultaneously, it dynamically calculates the effective head and outlet capacity based on the elevation difference between the two regions and the pipeline length, while also coordinating with grid demand. For the heating network, the system divides the regional heating network into multi-node pipe segments to establish heat capacity and thermal resistance equations, simulating the spatiotemporal diffusion of pipeline heat loss, flow velocity, and supply / return water temperature difference. Furthermore, in the dynamic mapping model of the heat pump unit, the MPC is allowed to directly adjust the power output of the battery energy storage unit and the pumped storage unit to meet the heating network's temperature and heat demands. Combining local meteorological data and historical heat consumption curves, the system uses recursive least squares to update the load forecasting model online and quickly correct for sudden heating demands.

[0014] The scenario involves a two-region interconnected power-heat coupling system where wind power and photovoltaic bases and regional heat load centers are distributed and connected through dual energy storage systems of battery energy storage and pumped hydro storage. In severe cold or when the load fluctuates drastically, the system frequency must be kept stable, the energy storage must be safe and the temperature of the heat network must be balanced. At the same time, the system faces challenges of network jitter and packet loss.

[0015] As a further aspect of the present invention, step B includes, but is not limited to, the following specific operating conditions:

[0016] Stable operating condition: The system frequency standard deviation is less than 0.02Hz, the power deviation standard deviation is less than 5% of the rated power, and the communication delay jitter is always less than 50ms within five consecutive minutes. In this case, the prediction window is fixed at 10 minutes. The four optimization weights of frequency, power, energy storage unit SOC and heat network temperature deviation in the controller are allocated according to the established empirical ratio. The ADMM consensus iteration adopts a "lightweight" mode with a relaxation factor of 1 and 5 iterations.

[0017] Rapid fluctuation condition: If the load or new energy output change rate exceeds 20% of the rated power within any minute, or the absolute value of the frequency slope is greater than 0.05Hz / s, it is judged as a rapid fluctuation condition. At this time, the prediction window is shortened to 5 minutes, the frequency deviation weight is increased to 2, and the remaining weights are adjusted proportionally. The ADMM consensus algorithm automatically increases the relaxation factor from 1 to 5 and increases the number of iterations to 8 based on the local residual. If the local MPC solution takes more than 0.2 seconds, the last available solution is automatically adopted to ensure real-time response.

[0018] Peak load condition of heating network: When the return water temperature deviation of the heating network exceeds 3°C for three consecutive minutes and the heat load growth rate exceeds 10% of the rated heat within five minutes, it is identified as peak load condition of heating network. At this time, the weight of heating network temperature deviation is increased to 3 and the weight of frequency deviation is reduced to 0.5. The dual-channel recursive least squares prediction based on weather forecast and historical heat consumption curve is run in parallel. In addition, a lower limit constraint of heat is added to the MPC target to ensure priority response to peak demand of heating network.

[0019] SOC Critical Condition: When the SOC of any energy storage unit falls below 20% or rises above the 90% threshold, it is immediately identified as a SOC critical condition. A SOC trend penalty term with a weight of 10 times that of the normal value is added to the objective function to force the SOC back to the safe range. If the BESS reaches the limit, the peak load is automatically switched from BESS to pumped storage unit and a local alarm and compensation strategy is triggered. The SOC limit exceeding term is incorporated into the ADMM iterative constraint in the form of a slack variable.

[0020] Communication delay fluctuation condition: When the estimated communication delay fluctuates between 100ms and 500ms within a sliding two-minute window and the variance exceeds 50ms. 2 When the condition is identified as a communication delay fluctuation, the network delay estimate is updated online every 30 seconds using the sliding window least squares method, and the current delay is introduced into the MPC prediction equation as a first-order lag mapping term. If the delay exceeds 300ms, the consensus weight is immediately reduced to 0.5 and the number of iterations is reduced to 3 to prioritize local optimization execution.

[0021] The above operations are limited to fine-grained configurations of prediction window length, optimized weight allocation, dynamic delay compensation, and safety constraints for five typical operating conditions: stable operation, rapid fluctuations, peak load of the heating network, SOC criticality, and communication delay fluctuations. Combined with dynamic mapping of heat pumps and online prediction of meteorological and historical heating data, the support for the accuracy of thermal regulation is further enhanced.

[0022] To ensure rapid tracking and stable compensation for dynamic changes in network latency:

[0023] As a further aspect of the present invention, in step B, the online delay identification module in the communication delay fluctuation condition adopts the recursive least squares method with a forgetting factor of 0.95 to perform weighted estimation of the communication delay samples within the most recent 120 seconds, and maps the identified dynamic delay parameters to the delay compensation coefficient in the MPC prediction equation through a first-order inertial model, so as to accurately compensate for network jitter and delay fluctuation.

[0024] To ensure precise matching between power output and heating network demand within the same optimization framework, thereby maximizing the overall energy utilization efficiency of the system, a high-weight penalty is applied to the state of charge of energy storage units to prevent safety hazards and lifespan degradation caused by overcharging or over-discharging. This ensures that the system maintains a safe and reliable operating state throughout the dynamic peak shaving, frequency regulation, and heating coupling process.

[0025] As a further aspect of the present invention, in step C, the additional electro-thermal coupling term added to the objective function multiplies the power output of the battery energy storage unit and the pumped storage unit by their respective electro-thermal conversion efficiencies, and uses the square of the difference between this difference and the predicted load deviation of the heating network as a penalty, thereby achieving optimal synergy between power output and heat demand. The additional SOC constraint term added to the objective function calculates a penalty term based on the square of the absolute value of the SOC deviation from the safe zone when the SOC of any energy storage unit is below 20% or above 90%, and the penalty weight is not less than 10.

[0026] To prioritize and quickly suppress frequency disturbances and improve system dynamic response speed and stability during periods of severe fluctuation in renewable energy output or load, while maintaining a longer forecast horizon under stable operating conditions to balance dispatch efficiency and economy:

[0027] As a further aspect of the present invention, in step D, in each calculation cycle, the controller first detects the fluctuation range of the new energy output and load in the most recent minute and compares it with the preset fluctuation threshold. If it exceeds the preset threshold, the controller shortens the time range used by the model prediction proportionally according to the current fluctuation size, raises the priority of frequency deviation in the optimization target to the highest level, and correspondingly lowers the priority of frequency deviation, energy storage state of charge deviation and heating network temperature deviation. If it is below the threshold, the controller automatically restores the prediction time to the original set long cycle and restores the priority of each target to the initial setting.

[0028] To track and compensate for control command delays caused by network latency jitter and packet loss in real time, and to ensure that model-predicted control commands can be synchronously issued and executed in a distributed network environment, thereby improving the timeliness, stability, and robustness of system peak shaving, frequency regulation, and heating scheduling:

[0029] As a further aspect of the present invention, in step E, the delay identification module uses a weighted sliding window least squares method to estimate the communication delay and packet loss in the most recent 120 seconds, and automatically resamples when the packet loss rate exceeds 5%; the obtained delay estimate is smoothed by an exponential weighted average method to obtain a dynamic delay parameter; in the model prediction algorithm, this dynamic delay parameter is used as the time constant of the first-order inertial delay model, and the control input is compensated for the corresponding delay before being used for optimization to offset the instruction delay caused by network jitter and packet loss.

[0030] To achieve globally consistent coordination of power generation and heat output among regional controllers in a distributed architecture and constrained communication environment, ensuring both global power and heat balance and robustness against packet loss and latency jitter, thereby guaranteeing the real-time performance, stability, and optimality of peak shaving, frequency regulation, and heating scheduling:

[0031] As a further aspect of the present invention, in step F, the local decision quantities for each region include the active power of the generating units, the charging and discharging power of the energy storage units, and the thermal power adjustment of the heating network. The active power increment and thermal power increment to be coordinated between regions are defined as global consensus variables. Each region controller introduces a Lagrange dual term and a penalty factor term into its local objective function. The Lagrange dual term is composed of the product of the difference between the local decision quantity and the global consensus variable and the corresponding dual variable. The penalty factor term is composed of the square of the difference between the local decision quantity and the global consensus variable multiplied by a penalty factor, to form an augmented Lagrange function. The initial value of the penalty factor is within a preset range and is dynamically adjusted during the iteration process based on the ratio of the local residual to the dual residual (when the local residual is more than ten times the dual residual, the penalty factor is doubled; when the dual residual is more than ten times the local residual, the penalty factor is halved. The iteration process includes: solving the corresponding augmented Lagrange problem in parallel for all regions; summarizing the solutions of each subproblem and taking the average to update the global consensus variable; updating the dual variable; the maximum number of iterations is ten, and the process is terminated early when both the local residual and the dual residual are below a preset threshold). Each region controller exchanges local decision variables, global consensus variables, and dual variables via IEC61850 GOOSE messages, with an exchange period not exceeding 100ms. The global consensus variable and the dual variable are initially set as zero vectors, and if communication packet loss occurs or a single message delay exceeds 200ms, the previous valid value is used to continue the iteration.

[0032] In the context of large-scale renewable energy grid integration and centralized heating, power system frequency control and heating network temperature regulation are mutually restrictive, and a single "calculate first, then execute" mode cannot simultaneously ensure real-time coordination between the two. The two area controllers communicate in real-time via IEC61850 GOOSE, and network jitter, packet loss, and latency fluctuations severely threaten the collaborative effect; one-time optimization alone cannot guarantee command synchronization and convergence stability. Furthermore, the output of renewable energy sources changes rapidly, and the dynamic differences between energy storage units and the heating network are significant, with continuous changes in weather, load, and equipment status. Therefore, the following specific improvements are made:

[0033] As a further aspect of the present invention, steps A to F are mutually mapped and fed back according to the following cyclical and circuitous logic:

[0034] Step 1, Feature Encoding: Map the power system frequency, power distribution, energy storage unit SOC, and heating network temperature status data collected in Step A into low-dimensional feature vectors;

[0035] Step 2, Generate initial control commands: Based on the low-dimensional feature vector from Step 1, generate initial control commands in Step B. The initial control commands include battery energy storage discharge / charge power, pumped storage unit output, and heating network heat supply distribution.

[0036] Step 3, Constraint Correction: In step C, the initial control quality is corrected based on electro-thermal coupling and SOC constraints to obtain the corrected control vector;

[0037] Step 4, Adaptive parameter tuning: Step D adaptively adjusts the prediction window length and optimization weights based on the output of new energy sources and load fluctuations, refines the parameters of the corrected control vector, and forms an adaptive control vector;

[0038] Step 5, Delay Compensation: Step E performs dynamic delay identification and compensation on the adaptive control vector, and outputs the compensated control vector;

[0039] Step 6, Distributed Consensus: Step F uses the ADMM consensus algorithm to solve the compensated control vector in parallel iteratively to obtain the final control results for each region;

[0040] Step 7, Response Prediction and Deviation Feedback: Input the final control result into the system model to generate a predicted response, and compare it with the expected frequency, power, SOC and temperature targets. Calculate the deviation vector between the predicted response and the target, and feed it back to the mapping unit as a regularization constraint for the next round of low-dimensional feature vector update.

[0041] Step 8, Cyclic Convergence: Repeat steps 1 to 7 until all deviations converge to the preset threshold.

[0042] In this system, due to the rapid fluctuations in wind and solar power output, measurement noise in the heating network temperature sensor, and model uncertainties caused by the head of the pumped storage unit and the water hammer effect in the pipeline, if only deterministic mapping is used, the generated control commands will fluctuate violently with the input noise, leading to frequent switching of energy storage units and increased equipment wear. In multi-region distributed iteration, traditional control methods suffer from communication delays and packet loss, resulting in incomplete information exchanged between controllers. This makes them prone to getting trapped in local optima, slow iteration convergence, and overfitting historical load / heat load curves, resulting in insufficient generalization performance during seasonal or sudden weather changes and difficulty in maintaining both frequency and temperature steady states.

[0043] As a further embodiment of the present invention, the mapping unit in step 7 specifically includes an encoder, a sampling unit, a decoder, and a distribution regularization module. The encoder extracts features from the state variables collected in step A using a multi-layer neural network and outputs the mean and variance parameters of the posterior distribution of the latent variable space. The sampling unit generates latent data based on the mean and variance parameters output by the encoder and independent and identically distributed random noise through random sampling operations. The decoder maps the latent vector back to the control command space to generate the initial control vector required in step B. The distribution regularization module constrains the posterior distribution parameters by minimizing the divergence between the posterior distribution and the predefined prior distribution.

[0044] The technical advantages of the present invention for a peak-shaving and frequency-regulating heating method for multi-element energy storage coupled coal-fired power units are as follows:

[0045] This invention constructs a dual-time-domain, multi-mode power-thermal hybrid time-domain model that balances millisecond-level frequency response and minute-level energy dispatch within the DC-DMPC distributed model predictive control framework. By combining the SOC-voltage dynamic coupling state equation based on a first-order RC circuit with nonlinear efficiency curves, temperature-dependent internal resistance, and capacity decay models, it achieves adaptive optimization of discharge / charge losses and response speeds for battery energy storage and pumped hydro storage. Simultaneously, it introduces an electro-thermal coupling penalty term and a high-weighted SOC constraint into the MPC objective function to ensure efficient matching of power output with the heating network load and prevent energy storage over-limitation. This invention also utilizes real-time online delay identification and first-order inertia compensation, possessing… The system employs an adaptive prediction window and weight adjustment encoding-sampling-decoding-regularization loop mapping, along with an ADMM-based global consensus iteration mechanism. This approach accurately compensates for network latency jitter and dynamically coordinates battery, energy storage unit, and heat pump heating strategies, addressing typical fluctuations in wind and solar power, rapid changes in heat load, and communication jitter / packet loss environments. Through multi-round deviation feedback and loop optimization, it achieves coordinated convergence of frequency deviation, power balance, SOC, and heating network temperature deviation, significantly improving the system's dynamic robustness, peak shaving, frequency regulation, and heating coordination efficiency. It supports real-time online adaptive model updates, while significantly reducing system operational risks and extending the lifespan of energy storage and heat pump equipment, enabling safe, efficient, and long-term continuous operation. Attached Figure Description

[0046] Figure 1 A diagram of the human-machine interface for applying the method proposed in this invention. Detailed Implementation

[0047] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0048] Example 1. As... Figure 1 As shown, the present invention proposes a method for peak-shaving and frequency-regulating heating for multi-element energy storage coupled coal-fired power units, comprising the following steps:

[0049] A. Establish a two-region interconnected power-heat system model including battery energy storage units, pumped storage units, and heating network loads;

[0050] B. Design a distributed model predictive controller based on DC-DMPC, with frequency deviation, power balance, energy storage unit SOC and heating network temperature deviation as optimization objectives;

[0051] C. Introduce an electro-thermal coupling term and a storage unit SOC constraint into the MPC objective function;

[0052] D. The controller adaptively adjusts the prediction window length and weighting coefficients based on the output of new energy sources and load fluctuations;

[0053] E. A network jitter and packet loss dynamic delay identification strategy is adopted, and a delay compensation model is embedded in the prediction model;

[0054] F. Each area controller coordinates the power and heat among units through the ADMM consensus algorithm and uses the alternating direction multiplier method to decompose and solve the problem.

[0055] exist Figure 1In this platform, for the peak-shaving and frequency-regulating heating method for multi-energy storage coupled coal-fired power units proposed in this invention, the real-time data such as active power output, heating network power, system frequency, and energy storage unit SOC are dynamically displayed in the "System Status Overview" area. The actual power and predicted power lines are overlaid in the "Real-time Power Curve" to verify the model accuracy online. The "System Operating Conditions" panel displays real-time communication delay identification parameters (forgetting factor, delay estimation, packet loss rate), sampling period, frequency, and power stability indicators, while also indicating the current operating condition. In the "Control Strategy and Parameters" area, the prediction window length and optimization target weights (frequency deviation, power deviation, SOC deviation, temperature deviation) and ADMM consensus algorithm relaxation factors are visually adjusted via sliders. The system displays the number of iterations and the status of the first-order time delay compensation model. The "Heating Network Temperature Response" and "Multi-Region Power Allocation" charts reflect the real-time response rate of the heating network area temperature deviation, pipeline flow velocity, and the power allocation ratio of battery storage and pumped storage units in each area. The "Real-time Alarm" and "Operation Log" areas centrally display alarm information and adjustment records such as excessively high energy storage temperature, frequency exceeding limits, abnormal heating network temperature, and communication anomalies. This achieves a closed-loop iteration from setting prediction parameters to optimizing algorithm parameters, and then to monitoring execution effects and handling alarms. The system can adaptively adjust and visualize the online operation based on multiple operating conditions such as wind and solar power fluctuations, sudden changes in heat load, and network jitter, ensuring efficient, stable, and safe peak shaving, frequency regulation, and heating scheduling.

[0056] To significantly shorten the system's frequency response time under renewable energy disturbances, improve the discharge / charge efficiency of energy storage units, accurately reproduce the spatiotemporal dynamics of hydraulic-electromechanical and heating networks, and ensure effective control of system frequency stability and heating network temperature deviation under large-scale grid connection and sudden heat load conditions: In step A, based on the classic first-order RC circuit SOC-voltage model, combined with the field-calibrated nonlinear efficiency curve and temperature-dependent internal resistance, temperature, SOC, and power are dynamically coupled into the state equation to adaptively optimize discharge / charge losses and response speed under different operating conditions. This is incorporated into a capacity decay model based on the cumulative cycle number and depth, with the decay rate set as the decision variable. To balance the needs of millisecond-level frequency regulation and minute-level energy dispatch, a hybrid system is constructed using dual discrete step lengths of 0.1s and 60s. The time-domain model is designed with dual-mode state equations for both pumping and power generation. It incorporates hydraulic-electromechanical coupling terms for head variation and a first-order lag pipe water hammer approximation term to capture pressure oscillations and power fluctuations during sudden flow changes. Simultaneously, it dynamically calculates the effective head and outlet capacity based on the elevation difference between the two regions and the pipeline length, coordinating with grid demand. For the heating network, the regional heating network is divided into multi-node pipe segments, establishing heat capacity and thermal resistance equations to simulate the spatiotemporal diffusion of pipeline heat loss, flow velocity, and supply / return water temperature difference. In the dynamic mapping model of the heat pump unit, the MPC is allowed to directly adjust the power output of the battery energy storage unit and the pumped storage unit to meet the heating network's temperature and heat demands. Combining local meteorological data and historical heat consumption curves, the load forecasting model is updated online using the recursive least squares method, and rapid corrections are made for sudden heating demands.

[0057] The scenario involves a two-region interconnected power-heat coupling system where wind power and photovoltaic bases and regional heat load centers are distributed and connected through dual energy storage systems of battery energy storage and pumped hydro storage. In severe cold or when the load fluctuates drastically, the system frequency must be kept stable, the energy storage must be safe and the temperature of the heat network must be balanced. At the same time, the system faces challenges of network jitter and packet loss.

[0058] It should be noted that step B includes, but is not limited to, the following specific working conditions:

[0059] Stable operating condition: The system frequency standard deviation is less than 0.02Hz, the power deviation standard deviation is less than 5% of the rated power, and the communication delay jitter is always less than 50ms within five consecutive minutes. In this case, the prediction window is fixed at 10 minutes. The four optimization weights of frequency, power, energy storage unit SOC and heat network temperature deviation in the controller are allocated according to the established empirical ratio. The ADMM consensus iteration adopts a "lightweight" mode with a relaxation factor of 1 and 5 iterations.

[0060] Rapid fluctuation condition: If the load or new energy output change rate exceeds 20% of the rated power within any minute, or the absolute value of the frequency slope is greater than 0.05Hz / s, it is judged as a rapid fluctuation condition. At this time, the prediction window is shortened to 5 minutes, the frequency deviation weight is increased to 2, and the remaining weights are adjusted proportionally. The ADMM consensus algorithm automatically increases the relaxation factor from 1 to 5 and increases the number of iterations to 8 based on the local residual. If the local MPC solution takes more than 0.2 seconds, the last available solution is automatically adopted to ensure real-time response.

[0061] Peak load condition of heating network: When the return water temperature deviation of the heating network exceeds 3°C for three consecutive minutes and the heat load growth rate exceeds 10% of the rated heat within five minutes, it is identified as peak load condition of heating network. At this time, the weight of heating network temperature deviation is increased to 3 and the weight of frequency deviation is reduced to 0.5. The dual-channel recursive least squares prediction based on weather forecast and historical heat consumption curve is run in parallel. In addition, a lower limit constraint of heat is added to the MPC target to ensure priority response to peak demand of heating network.

[0062] SOC Critical Condition: When the SOC of any energy storage unit falls below 20% or rises above the 90% threshold, it is immediately identified as a SOC critical condition. A SOC trend penalty term with a weight of 10 times that of the normal value is added to the objective function to force the SOC back to the safe range. If the BESS reaches the limit, the peak load is automatically switched from BESS to pumped storage unit and a local alarm and compensation strategy is triggered. The SOC limit exceeding term is incorporated into the ADMM iterative constraint in the form of a slack variable.

[0063] Communication delay fluctuation condition: When the estimated communication delay fluctuates between 100ms and 500ms within a sliding two-minute window and the variance exceeds 50ms. 2 When the condition is identified as a communication delay fluctuation, the network delay estimate is updated online every 30 seconds using the sliding window least squares method, and the current delay is introduced into the MPC prediction equation as a first-order lag mapping term. If the delay exceeds 300ms, the consensus weight is immediately reduced to 0.5 and the number of iterations is reduced to 3 to prioritize local optimization execution.

[0064] For example, the system platform monitored the following data from 08:30 to 08:35 on a certain day: frequency 50.01Hz ± 0.004Hz (5-minute standard deviation 0.004Hz < 0.02Hz), power deviation ± 2.0MW (standard deviation 2.5% < 5%), average communication jitter 30ms, and jitter amplitude < 50ms. This meets the criteria of "frequency standard deviation < 0.02Hz, power deviation < 5%, and average communication jitter < 50ms within 5 consecutive minutes," and is identified as a stable operating condition. The applied strategy is a prediction window of 10 minutes, with weights (frequency: power: SOC: ...). With a temperature ratio of 1:1:0.5:0.5, the ADMM consensus iteration uses a relaxation factor of 1, iterates 5 times, and the controller performs only the minimum number of iterations to release communication bandwidth. All indicators are optimized in a balanced manner to maintain stability at the lowest cost. Within the 08:36-08:37 window, the frequency abruptly changes from 50Hz to 48.5Hz, with a frequency change rate of 1.5Hz / s » 0.05Hz / s. The window is shortened to 5 minutes, the frequency weight is increased to 2, and other weights are simultaneously reduced. The ADMM consensus iteration relaxation factor is dynamically increased to 5, iterates 8 times, and the MPC solution time is 0.2s. The previous solution is retained and executed immediately. Specific examples of peak load conditions, SOC critical conditions, and end-to-end communication delay fluctuation conditions in the heating network are similarly identified through condition recognition and executed according to the aforementioned adaptive application strategies.

[0065] The above operations are limited to fine-grained configurations of prediction window length, optimized weight allocation, dynamic delay compensation, and safety constraints for five typical operating conditions: stable operation, rapid fluctuations, peak load of the heating network, SOC criticality, and communication delay fluctuations. Combined with dynamic mapping of heat pumps and online prediction of meteorological and historical heating data, the support for the accuracy of thermal regulation is further enhanced.

[0066] To ensure rapid tracking and stable compensation of dynamic changes in network latency, in step B, the online latency identification module in the communication latency fluctuation condition uses a recursive least squares method with a forgetting factor of 0.95 to perform weighted estimation of communication latency samples within the most recent 120 seconds. The identified dynamic latency parameters are then mapped to latency compensation coefficients in the MPC prediction equation through a first-order inertial model, accurately compensating for network jitter and latency fluctuations. The Recursive Least Squares (RLS) method with a forgetting factor of 0.95 assigns higher weight to the latest delay samples, enabling a rapid response to recent network conditions while gradually diminishing the impact of past data. When communication delay changes abruptly due to temporary jitter or network packet jitter, the system can accurately capture the new delay level within tens of seconds, providing a reliable basis for subsequent compensation. The online identified delay parameters are then smoothed using a first-order inertial model, effectively filtering between high-frequency noise and low-frequency trends. This ensures that the compensation parameters do not oscillate significantly due to measurement noise or short-term jitter, while retaining sensitivity to actual network delay abrupt changes, thus avoiding over-compensation or failure to keep up with changes. The dilemma of communication delay can be effectively eliminated by introducing real-time updated delay compensation coefficients into the prediction equations of Model Predictive Control (MPC). This allows the control input to act on the field devices in a timely and accurate manner, significantly reducing frequency deviation and heating network temperature error caused by delay. Network jitter and packet loss often disrupt the synchronization between distributed controllers, leading to convergence difficulties or local oscillations. Through dynamic delay identification and inertia compensation, distributed MPC can maintain high synchronization accuracy under different delay conditions, reducing dependence on conservative parameter settings, thereby accelerating algorithm convergence and improving the robustness of overall peak shaving, frequency regulation, and heating control.

[0067] To ensure precise matching between power output and heating network demand within the same optimization framework, thereby maximizing the overall energy utilization efficiency of the system, a high-weight penalty is applied to the state of charge (SOC) of energy storage units to prevent safety hazards and lifespan degradation caused by overcharging or over-discharging. This ensures that the system maintains a safe and reliable operating state throughout the dynamic peak shaving, frequency regulation, and heating coupling process. In step C, the additional electro-thermal coupling term in the objective function uses the square of the difference between the power output of the battery energy storage unit and the pumped storage unit multiplied by their respective electro-thermal conversion efficiencies and the predicted load deviation of the heating network as a penalty, achieving optimal synergy between power output and heat demand. The additional SOC constraint term in the objective function calculates a penalty term based on the square of the absolute value of the SOC deviation from the safe zone when the SOC of any energy storage unit is below 20% or above 90%, with a penalty weight of not less than 10.

[0068] To prioritize and quickly suppress frequency disturbances and improve the dynamic response speed and stability of the system when renewable energy output or load fluctuates drastically, while maintaining a longer forecast horizon under stable operating conditions to balance dispatch efficiency and economy, in step D, in each calculation cycle, the controller first detects the fluctuation amplitude of renewable energy output and load in the most recent minute and compares it with a preset fluctuation threshold. If it exceeds the preset threshold, the controller shortens the time range used by the model prediction proportionally according to the current fluctuation magnitude, raises the priority of frequency deviation in the optimization objectives to the highest level, and correspondingly lowers the priority of frequency deviation, energy storage state of charge deviation, and heating network temperature deviation. If it is below the threshold, the controller automatically restores the prediction time to the original set long cycle and restores the priority of each objective to the initial setting.

[0069] To track and compensate for control command delays caused by network jitter and packet loss in real time, and to ensure that model-predicted control commands can be synchronously issued and executed in a distributed network environment, thereby improving the timeliness, stability, and robustness of system peak shaving, frequency regulation, and heating scheduling, in step E, the delay identification module uses a weighted sliding window least squares method to estimate the communication delay and packet loss over the most recent 120 seconds, and automatically resamples when the packet loss rate exceeds 5%. The obtained delay estimates are smoothed using an exponential weighted average method to obtain dynamic delay parameters. In the model prediction algorithm, these dynamic delay parameters are used as the time constant of the first-order inertial delay model. The control input is compensated for the corresponding delay before being used for optimization to offset the command delays caused by network jitter and packet loss.

[0070] To achieve globally consistent coordination of power generation and thermal power among regional controllers in a distributed architecture and constrained communication environment, ensuring both global power and thermal balance and robustness against packet loss and latency jitter, thereby guaranteeing the real-time performance, stability, and optimality of peak shaving, frequency regulation, and heating scheduling, in step F, the local decision quantities for each region include the active power of generating units, the charging and discharging power of energy storage units, and the thermal power adjustment of the heating network. The incremental active power and thermal power to be coordinated between regions are defined as global consensus variables. Each regional controller introduces a Lagrange dual term and a penalty factor term into its local objective function. The Lagrange dual term is composed of the product of the difference between the local decision quantity and the global consensus variable and the corresponding dual variable. The penalty factor is formed by multiplying the square of the difference between the local decision variable and the global consensus variable by the penalty factor to form an augmented Lagrangian function. The initial value of the penalty factor is within a preset range and is dynamically adjusted during the iteration process according to the ratio of the local residual to the dual residual (when the local residual is more than ten times the dual residual, the penalty factor is doubled; when the dual residual is more than ten times the local residual, the penalty factor is halved. The iteration process includes: solving the corresponding augmented Lagrangian subproblems in parallel for all regions; summarizing the solutions of each subproblem and taking the average to update the global consensus variable; updating the dual variable; the maximum number of iterations is ten, and the process is terminated early when both the local residual and the dual residual are below a preset threshold). Each region controller exchanges local decision variables, global consensus variables, and dual variables through IEC61850 GOOSE messages, with an exchange period of no more than 100ms. The global consensus variable and the dual variable are initially set as zero vectors, and if a communication packet loss occurs or a single message delay exceeds 200ms, the previous valid value is used to continue the iteration.

[0071] This invention achieves precise matching between power output and heating network load by introducing electro-thermal coupling penalties and high-weight SOC constraints within the same optimization framework. Adaptive prediction windows and weight adjustments ensure rapid system response under severe fluctuations and efficient operation under stable conditions. Simultaneously, online delay identification and inertia compensation eliminate the impact of network jitter and packet loss on control synchronization, while the ADMM consensus iteration mechanism ensures global consistency of power and heat across all regions. Unlike existing technologies that rely on fixed time domains and static weights and cannot simultaneously address electro-thermal coordination and delay jitter compensation, this invention balances dynamic performance, operational safety, and communication robustness, significantly improving system energy utilization efficiency, peak shaving and frequency regulation capabilities, heating coordination, and long-term stable operation reliability.

[0072] Example 2. The difference between Example 2 and Example 1 is that this example introduces a peak-shaving and frequency-regulating heating method for multi-element energy storage coupled coal-fired power units.

[0073] In the context of large-scale renewable energy grid integration and centralized heating, power system frequency control and heating network temperature regulation are mutually restrictive, and a single "calculate first, then execute" mode cannot simultaneously ensure real-time coordination between the two. The two area controllers communicate in real-time via IEC61850 GOOSE, and network jitter, packet loss, and latency fluctuations severely threaten the collaborative effect; one-time optimization alone cannot guarantee command synchronization and convergence stability. Furthermore, the output of renewable energy sources changes rapidly, and the dynamic differences between energy storage units and the heating network are significant, with continuous changes in weather, load, and equipment status. Therefore, the following specific improvements are made:

[0074] First, steps A through F are mapped and fed back to each other according to the following cyclical logic:

[0075] Step 1, Feature Encoding: Map the power system frequency, power distribution, energy storage unit SOC, and heating network temperature status data collected in Step A into low-dimensional feature vectors;

[0076] Step 2, Generate initial control commands: Based on the low-dimensional feature vector from Step 1, generate initial control commands in Step B. The initial control commands include battery energy storage discharge / charge power, pumped storage unit output, and heating network heat supply distribution.

[0077] Step 3, Constraint Correction: In step C, the initial control quality is corrected based on electro-thermal coupling and SOC constraints to obtain the corrected control vector;

[0078] Step 4, Adaptive parameter tuning: Step D adaptively adjusts the prediction window length and optimization weights based on the output of new energy sources and load fluctuations, refines the parameters of the corrected control vector, and forms an adaptive control vector;

[0079] Step 5, Delay Compensation: Step E performs dynamic delay identification and compensation on the adaptive control vector, and outputs the compensated control vector;

[0080] Step 6, Distributed Consensus: Step F uses the ADMM consensus algorithm to solve the compensated control vector in parallel iteratively to obtain the final control results for each region;

[0081] Step 7, Response Prediction and Deviation Feedback: Input the final control result into the system model to generate a predicted response, and compare it with the expected frequency, power, SOC and temperature targets. Calculate the deviation vector between the predicted response and the target, and feed it back to the mapping unit as a regularization constraint for the next round of low-dimensional feature vector update.

[0082] Step 8, Cyclic Convergence: Repeat steps 1 to 7 until all deviations converge to the preset threshold.

[0083] In this system, due to the rapid fluctuations in wind and solar power output, measurement noise in the heating network temperature sensor, and model uncertainties caused by the head of the pumped storage unit and the water hammer effect in the pipeline, if only deterministic mapping is used, the generated control commands will fluctuate violently with the input noise, leading to frequent switching of energy storage units and increased equipment wear. In multi-region distributed iteration, traditional control methods suffer from communication delays and packet loss, resulting in incomplete information exchanged between controllers. This makes them prone to getting trapped in local optima, slow iteration convergence, and overfitting historical load / heat load curves, resulting in insufficient generalization performance during seasonal or sudden weather changes and difficulty in maintaining both frequency and temperature steady states.

[0084] Therefore, to further improve the above mechanism, the mapping unit in step 7 specifically includes an encoder, a sampling unit, a decoder, and a distribution regularization module. The encoder extracts features from the state variables collected in step A using a multi-layer neural network and outputs the mean and variance parameters of the posterior distribution of the latent variable space. The sampling unit generates latent data based on the mean and variance parameters output by the encoder and independent and identically distributed random noise through random sampling operations. The decoder maps the latent vector back to the control command space to generate the initial control vector required in step B. The distribution regularization module constrains the posterior distribution parameters by minimizing the divergence between the posterior distribution and the predefined prior distribution.

[0085] By introducing an end-to-end joint optimization mechanism of encoder-sampling unit-decoder-distribution regularization module into the mapping unit, high-dimensional grid frequency, power distribution, energy storage SOC, and heating network temperature state variables are mapped to parameterized latent distributions. Within the latent space, diverse control vectors are generated through random sampling, and distribution regularization minimizes the posterior-prior distribution divergence, thereby suppressing high-frequency jitter caused by measurement noise and model uncertainty. This avoids drastic fluctuations in control commands due to input disturbances caused by uncertainty mapping. Simultaneously, cyclic bias feedback is used to regularize the difference between the predicted response and the target value back to the latent space, achieving online adaptive model updates and significantly improving performance. Distributed MPC effectively prevents traditional multi-region optimization methods from getting stuck in local optima or slow convergence due to incomplete information by improving synchronization and convergence speed under conditions of communication jitter, packet loss, and latency. Through multiple rounds of mapping-correction-decoding iteration, it maintains multi-objective coordinated convergence of frequency regulation, power balance, energy storage safety, and heating network temperature under various operating conditions such as rapid fluctuations in wind and solar power, sudden loads in heating networks, and dynamic differences in equipment. Ultimately, it achieves a significant improvement in system dynamic robustness, energy utilization efficiency, and heat accuracy, providing reliable, efficient, and sustainable technical support for electric-thermal coupling peak-shaving and frequency regulation heating methods in the scenario of large-scale renewable energy grid connection and centralized heating.

[0086] Combining Embodiments 1 and 2, this invention constructs a dual-time-domain, multi-mode power-thermal hybrid time-domain model that balances millisecond-level frequency response and minute-level energy dispatch within the DC-DMPC distributed model predictive control framework. By combining the SOC-voltage dynamic coupling state equation based on a first-order RC circuit with nonlinear efficiency curves, temperature-dependent internal resistance, and capacity decay models, adaptive optimization of battery energy storage and pumped hydro storage discharge / charge losses and response speeds is achieved. Simultaneously, an electro-thermal coupling penalty term and a high-weight SOC constraint are introduced into the MPC objective function to ensure efficient matching of power output with the heating network load and prevent over-limiting of energy storage. This invention also utilizes real-time online delay identification and first-order... Inertial compensation, an encoding-sampling-decoding-regularized cyclic mapping with adaptive prediction windows and weight adjustments, and an ADMM-based global consensus iteration mechanism are used to accurately compensate for network latency jitter and dynamically coordinate battery, energy storage unit, and heat pump heating strategies in typical wind and solar power fluctuations, rapid changes in heat load, and communication jitter and packet loss environments. Through multi-round deviation feedback and loop optimization, the system achieves coordinated convergence of frequency deviation, power balance, SOC, and heating network temperature deviation, significantly improving the system's dynamic robustness, peak shaving, frequency regulation, and heating coordination efficiency. It supports real-time online adaptive model updates, while significantly reducing system operation risks and extending the lifespan of energy storage and heat pump equipment, achieving safe, efficient, and long-term continuous operation.

[0087] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0088] In conclusion, 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 within the protection scope of the present invention.

Claims

1. A method for peak-shaving and frequency-regulating heating using multi-element energy storage coupled coal-fired power units, characterized in that, Includes the following steps: A. Establish a two-region interconnected power-heat system model including battery energy storage units, pumped storage units, and heating network loads; B. Design a distributed model predictive controller based on DC-DMPC, with frequency deviation, power balance, energy storage unit SOC and heating network temperature deviation as optimization objectives; C. Introduce an electro-thermal coupling term and a storage unit SOC constraint into the MPC objective function; D. The controller adaptively adjusts the prediction window length and weighting coefficients based on the output of new energy sources and load fluctuations; E. A network jitter and packet loss dynamic delay identification strategy is adopted, and a delay compensation model is embedded in the prediction model; F. Each area controller coordinates the power and heat among units through the ADMM consensus algorithm and uses the alternating direction multiplier method to decompose and solve the problem.

2. The method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, In step A, based on the classic first-order RC circuit SOC-voltage model, combined with the field-calibrated nonlinear efficiency curve and temperature-dependent internal resistance, temperature, SOC, and power are dynamically coupled and introduced into the state equation to adaptively optimize discharge / charge losses and response speed under different operating conditions. This is then incorporated into a capacity decay model based on the cumulative cycle count and depth, with the decay rate set as the decision variable. To accommodate both millisecond-level frequency regulation and minute-level energy dispatch requirements, a hybrid time-domain model is constructed using dual discrete step lengths of 0.1s and 60s. A dual-mode state equation is designed for both pumping and power generation modes, embedding a hydraulic-electromechanical coupling term based on head variation and a first-order hysteresis loop. The system employs a water hammer approximation term to capture pressure oscillations and power fluctuations during sudden flow changes. It also dynamically calculates the effective head and outlet capacity based on the elevation difference between the two regions and the pipeline length, while coordinating with grid demand. For the heating network, the system divides the regional heating network into multi-node pipe segments, establishing heat capacity and thermal resistance equations to simulate the spatiotemporal diffusion of pipeline heat loss, flow velocity, and supply / return water temperature difference. Furthermore, in the dynamic mapping model of the heat pump unit, the MPC is allowed to directly adjust the power output of the battery energy storage unit and the pumped storage unit to meet the heating network's temperature and heat demands. Combining local meteorological data and historical heat consumption curves, the system uses recursive least squares to update the load forecasting model online and quickly correct for sudden heating demands.

3. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, Step B includes, but is not limited to, the following specific operating conditions: Stable operating condition: The system frequency standard deviation is less than 0.02Hz, the power deviation standard deviation is less than 5% of the rated power, and the communication delay jitter is always less than 50ms within five consecutive minutes. In this case, the prediction window is fixed at 10 minutes. The four optimization weights of frequency, power, energy storage unit SOC and heat network temperature deviation in the controller are allocated according to the established empirical ratio. The ADMM consensus iteration adopts a "lightweight" mode with a relaxation factor of 1 and 5 iterations. Rapid fluctuation condition: If the load or new energy output change rate exceeds 20% of the rated power within any minute, or the absolute value of the frequency slope is greater than 0.05Hz / s, it is judged as a rapid fluctuation condition. At this time, the prediction window is shortened to 5 minutes, the frequency deviation weight is increased to 2, and the remaining weights are adjusted proportionally. The ADMM consensus algorithm automatically increases the relaxation factor from 1 to 5 and increases the number of iterations to 8 based on the local residual. If the local MPC solution takes more than 0.2 seconds, the last available solution is automatically adopted to ensure real-time response. Peak load condition of heating network: When the return water temperature deviation of the heating network exceeds 3°C for three consecutive minutes and the heat load growth rate exceeds 10% of the rated heat within five minutes, it is identified as peak load condition of heating network. At this time, the weight of heating network temperature deviation is increased to 3 and the weight of frequency deviation is reduced to 0.

5. The dual-channel recursive least squares prediction based on weather forecast and historical heat consumption curve is run in parallel. In addition, a lower limit constraint of heat is added to the MPC target to ensure priority response to peak demand of heating network. SOC Critical Condition: When the SOC of any energy storage unit falls below 20% or rises above the 90% threshold, it is immediately identified as a SOC critical condition. A SOC trend penalty term with a weight of 10 times that of the normal value is added to the objective function to force the SOC back to the safe range. If the BESS reaches the limit, the peak load is automatically switched from BESS to pumped storage unit and a local alarm and compensation strategy is triggered. The SOC limit exceeding term is incorporated into the ADMM iterative constraint in the form of a slack variable. Communication delay fluctuation condition: When the estimated communication delay fluctuates between 100ms and 500ms within a sliding two-minute window and the variance exceeds 50ms. 2 When the condition is identified as a communication delay fluctuation, the network delay estimate is updated online every 30 seconds using the sliding window least squares method, and the current delay is introduced into the MPC prediction equation as a first-order lag mapping term. If the delay exceeds 300ms, the consensus weight is immediately reduced to 0.5 and the number of iterations is reduced to 3 to prioritize local optimization execution.

4. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 3, characterized in that, In step B, the online delay identification module in the communication delay fluctuation condition uses the recursive least squares method with a forgetting factor of 0.95 to perform weighted estimation of the communication delay samples in the most recent 120 seconds, and maps the identified dynamic delay parameters to the delay compensation coefficient in the MPC prediction equation through a first-order inertial model, so as to accurately compensate for network jitter and delay fluctuation.

5. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, In step C, the additional electro-thermal coupling term in the objective function multiplies the power output of the battery energy storage unit and the pumped storage unit by their respective electro-thermal conversion efficiencies, and uses the square of the difference between this difference and the predicted load deviation of the heating network as a penalty, thereby achieving optimal synergy between power output and heat demand. The additional SOC constraint term in the objective function calculates a penalty term based on the square of the absolute value of the SOC deviation from the safe zone when the SOC of any energy storage unit is below 20% or above 90%, and the penalty weight is not less than 10.

6. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, In step D, in each calculation cycle, the controller first detects the fluctuation range of new energy output and load in the most recent minute and compares it with the preset fluctuation threshold. If it exceeds the preset threshold, the controller shortens the time range used by the model prediction proportionally according to the current fluctuation size, raises the priority of frequency deviation in the optimization target to the highest level, and correspondingly lowers the priority of frequency deviation, energy storage state of charge deviation and heating network temperature deviation. If it is below the threshold, the controller automatically restores the prediction time to the original set long cycle and restores the priority of each target to the initial setting.

7. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, In step E, the delay identification module uses the weighted sliding window least squares method to estimate the communication delay and packet loss in the last 120 seconds in real time, and automatically resamples when the packet loss rate exceeds 5%. The obtained delay estimates are smoothed using an exponential weighted average method to obtain dynamic delay parameters. In the model prediction algorithm, these dynamic delay parameters are used as the time constant of the first-order inertial delay model. The control input is compensated for the corresponding delay before being used for optimization to offset the instruction delay caused by network jitter and packet loss.

8. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, In step F, the local decision quantities for each region include the active power of the generating units, the charging and discharging power of the energy storage units, and the thermal power adjustment of the heating network. The active power increment and thermal power increment to be coordinated between regions are defined as global consensus variables. Each region controller introduces a Lagrange dual term and a penalty factor term into its local objective function. The Lagrange dual term is composed of the product of the difference between the local decision quantity and the global consensus variable and the corresponding dual variable. The penalty factor term is composed of the square of the difference between the local decision quantity and the global consensus variable multiplied by the penalty factor to form an augmented Lagrange function. The initial value of the penalty factor is within a preset range and is dynamically adjusted during the iteration process according to the ratio of the local residual to the dual residual.

9. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 1, characterized in that, Steps A through F are mapped and fed back to each other according to the following cyclical logic: Step 1, Feature Encoding: Map the power system frequency, power distribution, energy storage unit SOC, and heating network temperature status data collected in Step A into low-dimensional feature vectors; Step 2, Generate initial control commands: Based on the low-dimensional feature vector from Step 1, generate initial control commands in Step B. The initial control commands include battery energy storage discharge / charge power, pumped storage unit output, and heating network heat supply distribution. Step 3, Constraint Correction: In step C, the initial control quality is corrected based on electro-thermal coupling and SOC constraints to obtain the corrected control vector; Step 4, Adaptive parameter tuning: Step D adaptively adjusts the prediction window length and optimization weights based on the output of new energy sources and load fluctuations, refines the parameters of the corrected control vector, and forms an adaptive control vector; Step 5, Delay Compensation: Step E performs dynamic delay identification and compensation on the adaptive control vector, and outputs the compensated control vector; Step 6, Distributed Consensus: Step F uses the ADMM consensus algorithm to solve the compensated control vector in parallel iteratively to obtain the final control results for each region; Step 7, Response Prediction and Deviation Feedback: Input the final control result into the system model to generate a predicted response, and compare it with the expected frequency, power, SOC and temperature targets. Calculate the deviation vector between the predicted response and the target, and feed it back to the mapping unit as a regularization constraint for the next round of low-dimensional feature vector update. Step 8, Cyclic Convergence: Repeat steps 1 to 7 until all deviations converge to the preset threshold.

10. A method for peak-shaving and frequency-regulating heating of a multi-element energy storage coupled coal-fired power unit according to claim 9, characterized in that, Step 7's mapping unit specifically includes an encoder, a sampling unit, a decoder, and a distribution regularization module. The encoder extracts features from the state variables collected in step A using a multi-layer neural network and outputs the mean and variance parameters of the posterior distribution of the latent variable space. The sampling unit generates latent data based on the mean and variance parameters output by the encoder and independent and identically distributed random noise through random sampling operations. The decoder maps the latent vector back to the control command space to generate the initial control vector required in step B. The distribution regularization module constrains the posterior distribution parameters by minimizing the divergence between the posterior distribution and the predefined prior distribution.

Images