Thermal battery energy storage system optimization control method

Abstract

The application discloses a kind of thermal battery energy storage system optimization control method, solve the high energy consumption of thermal battery energy storage system, output temperature and power fluctuation and the problem of serious equipment loss, comprising: initialization each actuator's historical action intensity is zero and configure control parameter;Real-time acquisition optimization control parameter and carry out multi-source information fusion;Estimate and update current system state, based on historical action data dynamically adjust each actuator in optimization target action frequency penalty weight;With the sum of minimum stage cost item and control cost item as the objective function of multi-objective optimization, construct hard constraint condition, solve the objective function, obtain optimal control sequence;Extract the first element in optimal control sequence as the actual control instruction of current time to carry out optimization control.Multiple target dynamic collaborative optimization is realized, the stable output of energy supply temperature and power is guaranteed, the service life of overall equipment is prolonged and maintenance cost is reduced.

Description

Technical Field

[0001] This invention relates to the field of energy storage system control technology, and in particular to an optimized control method for a thermal battery energy storage system. Background Technology

[0002] With the rapid development of renewable energy sources such as wind and solar power, the power grid faces severe supply-demand mismatch and challenges in green electricity consumption due to the intermittent nature and anti-peak-shaving characteristics of wind and solar power output. In some provinces and cities, even negative electricity prices have emerged. Thermal battery energy storage is an emerging long-term energy storage technology that uses an electrothermal / cold conversion method through heat pump units. During periods of surplus electricity, it couples low-grade heat sources from the environment or waste heat from industrial processes to convert low-cost or even negative-price electricity into high-quality thermal and cold energy for storage. This energy is then released during peak energy consumption or high-electricity-price periods, achieving time-shifted energy utilization and value enhancement. It shows broad application prospects in areas such as regional energy supply, industrial waste heat recovery, renewable energy consumption, and deep peak shaving of the power grid.

[0003] For thermal battery energy storage systems, traditional single-loop control is prone to large fluctuations in output temperature and power, leading to overshoot and oscillation. Furthermore, due to the inability to integrate multi-loop information, its adaptability to various operating conditions is poor, severely impacting control quality and threatening the safe and stable operation of the system. In addition, to pursue control precision, actuators such as compressors and valves often operate frequently and with large adjustment ranges, accelerating equipment wear and aging and increasing maintenance costs. Simultaneously, in practical use, the system must simultaneously meet multiple physical constraints such as outlet temperature, output power, and energy storage capacity. There are conflicts between multiple objectives such as system economy, stability, and equipment lifespan. Existing control strategies, to ensure safety, often choose conservative control parameters and objectives, lacking system-wide collaborative optimization capabilities. This makes it difficult to achieve optimal scheduling under dynamic electricity prices, easily leading to high equipment power consumption and energy waste.

[0004] For example, patent CN120538204A describes a control method for a heat pump system based on improved integral model predictive control. This method includes: modeling an air-source heat pump system; constructing the objective function and constraints of an MPC predictive controller; designing an integral feedback controller; introducing an integral limiting mechanism; optimizing the objective function; and adjusting the control input in real time to predict system output data, thereby reducing system prediction errors. However, this invention only addresses a single heat pump unit, does not consider coupled control with an energy storage system, and its control objectives are limited to outlet temperature stability and actuator operating frequency. It fails to comprehensively consider key performance indicators such as energy cost, power delivery guarantee, and energy storage status management, and cannot solve the multi-objective, multi-constraint collaborative optimization challenges faced by thermal battery energy storage systems. Summary of the Invention

[0005] The purpose of this invention is to solve the problems of high energy consumption, large fluctuations in output temperature and power, and severe equipment damage in existing thermal battery energy storage systems. It provides an optimized control method for thermal battery energy storage systems, which realizes multi-objective dynamic collaborative optimization, ensures stable output of energy supply temperature and power, extends the service life of the overall equipment, and reduces maintenance costs.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: An optimized control method for a thermal battery energy storage system includes the following steps: A method for optimizing the control of a thermal battery energy storage system, characterized by comprising: S1: Initialize the historical motion intensity of each actuator to zero and configure the control parameters; S2: Real-time acquisition and optimization of control parameters and fusion of multi-source information; S3: Estimate and update the current system state, and dynamically adjust the action frequency penalty weight of each actuator in the optimization objective based on historical action data; S4: Using the sum of the stage cost term and the control cost term as the objective function for multi-objective optimization, construct hard constraints, solve the objective function, and obtain the optimal control sequence; S5: Extract the first element from the optimal control sequence as the actual control command at the current moment for optimized control.

[0007] The optimized control method for thermal battery energy storage systems provided by this invention constructs an objective function that includes stage cost terms and control cost terms, and solves the optimal control sequence in the prediction time domain. This achieves precise matching of system parameters under varying operating conditions, comprehensive energy efficiency optimization, and rapid response adjustment, effectively improving the intelligence level, control quality, and overall stability of the thermal battery energy storage system.

[0008] Preferably, the dynamic adjustment of the action frequency penalty weight of each actuator in the optimization objective based on historical action data includes: Calculate the over-threshold action amount of the actuator, which is either zero or the larger of the difference between the absolute value of the actuator control increment and the actuator's working determination threshold. Based on the over-threshold motion quantity, the normalized motion intensity of the actuator is calculated; Based on the forgetting factor, the historical action intensity and normalized action intensity of the actuator at the previous moment are recursively updated using the exponential weighted moving average method. Based on the updated historical action intensity, calculate the action frequency penalty weight for each actuator in the current cycle.

[0009] Preferably, the stage cost term is accumulated over the entire prediction time domain, and the stage cost term includes: The energy cost item is obtained based on time-of-use electricity pricing, the predicted system power consumption, and the discrete time step; the delivered power deviation item is obtained based on the under-supply / over-supply penalty weighting coefficient and the under-supply / over-supply slack variable. The temperature stability control item does not generate a penalty when the predicted temperature is within the set value plus or minus the dead zone; otherwise, it accumulates a penalty cost, which increases with the square of the temperature deviation. The temperature-oriented adjustment term is obtained based on the temperature adjustment coefficient and the temperature dynamic reference trajectory. The energy storage state deviation is obtained based on the deviation of the energy storage state from the recommended operating window.

[0010] Preferably, the hard constraints include state equation constraints, control time domain hold constraints, actuator physical limit constraints, actuator rate of change limit constraints, energy storage physical constraints, and relaxation variable non-negativity constraints. The control time domain hold constraints mean that the controller at each time outside the control time domain is equal to the control quantity at the end of the control time domain.

[0011] Preferably, the control cost term is accumulated over the entire control time domain, and the control cost term includes: The control increment smoothing term is the sum of the products of the increment weights of all actuators and the squares of the control increments. The action frequency control term is the sum of the product of the action frequency penalty weights of all actuators and the square of the effective action amount of the actuators exceeding the dead zone. The action frequency penalty weights are updated once at the beginning of the predictive control solution cycle and remain constant throughout the prediction time domain.

[0012] Preferably, the real-time acquisition and optimization control parameters include: Real-time acquisition and preprocessing of system operation data at the current moment; extraction of feature variables for optimization solution. Simultaneously acquire external information for current and future forecast time series, including environmental temperature forecast sequence, waste heat temperature forecast sequence, and time-of-use electricity price sequence; Read the system operation log and extract the historical action intensity and control increment of each actuator at the previous moment; Obtain the current energy state of the system.

[0013] Preferably, step S3 includes: obtaining the optimal estimate of the system state vector at the current moment using a state estimation method based on the collected system operation data, and establishing initial condition constraints; calculating the current energy storage state of the system and normalizing it to the [0,1] interval.

[0014] Preferably, the characteristic variables include state characteristics, process characteristics, statistical characteristics, and derived characteristics. The state characteristics include filtered temperature, motor frequency, pump speed, fan speed, and valve opening. The process characteristics include instantaneous electrical power, instantaneous output heat / cold power, inlet and outlet temperature difference of the heat exchanger, and average temperature of the energy storage medium. The derived characteristics include normalized thermal / cold storage capacity state and system instantaneous performance coefficient.

[0015] Preferably, the configuration control parameters include: setting basic timing parameters for predictive control of the thermal battery system, the basic timing parameters including sampling period, prediction time domain length and control time domain length; obtaining the objective function weights for multi-objective optimization according to the selected operating mode; configuring temperature control dead zone parameters; setting the maximum allowable load deviation of output power; configuring operating window parameters for the thermal battery energy storage system; configuring actuator physical limit and actuator rate of change limit; and configuring action frequency control parameters.

[0016] Preferably, the system operating data includes process parameters and equipment operating information. The process parameters include cold / hot end inlet and outlet temperatures, flow rates, and pressures. The equipment operating information includes motor power and the current position status of each actuator.

[0017] Therefore, the present invention has the following beneficial effects: by combining soft and hard constraints, and designing relaxation variables and dead zones, the control flexibility of the thermal battery energy storage system is enhanced. At the same time, an adaptive action frequency penalty mechanism based on historical action intensity and an energy storage state operation window management strategy are introduced to achieve intelligent protection of the actuator and maintain the adjustment capability of the thermal battery energy storage system. Thus, under the premise of ensuring safety, the economy, stability and reliability of the thermal battery energy storage system under dynamic operating conditions are improved. Attached Figure Description

[0018] Figure 1 This is a flowchart illustrating the overall steps of the optimized control method for the thermal battery energy storage system in Example 1.

[0019] Figure 2 This is a schematic diagram of the architecture of an existing thermal battery energy storage system based on a centrifugal compression heat pump.

[0020] Figure 3 This is a schematic diagram of the architecture for optimized control of the thermal battery energy storage system in Example 2.

[0021] In the diagram: 1. Refrigerant pump; 2. Cold-side control valve; 3. Cold storage tank; 4. Low-temperature heat exchanger; 5. Expander; 6. Regenerator; 7. Compressor; 8. Motor; 9. High-temperature heat exchanger; 10. Hot-side control valve; 11. Fan; 12. Cold storage tank; 13. Cold-end heat exchanger. Detailed Implementation

[0022] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this application will be further described in detail below through embodiments and in conjunction with the accompanying drawings. It should be understood that the specific embodiments described herein are merely one preferred embodiment of this application and are only used to explain this application. They do not limit the scope of protection of this application. All other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this invention. The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments: Example 1: This embodiment provides an optimized control method for a thermal battery energy storage system, such as... Figure 1 As shown, the operation process is as follows: Step 1, initialize the historical action intensity of each actuator to zero and configure the control parameters; Step 2, collect and optimize the control parameters in real time and perform multi-source information fusion; Step 3, estimate and update the current system state, and dynamically adjust the action frequency penalty weight of each actuator in the optimization objective based on historical action data; Step 4, construct hard constraints with the objective function of minimizing the sum of the stage cost term and the control cost term, solve the objective function, and obtain the optimal control sequence; Step 5, extract the first element in the optimal control sequence as the actual control command at the current moment for optimization control.

[0023] like Figure 2 As shown, the thermal battery energy storage system based on a centrifugal compression heat pump includes a refrigerant pump 1, a cold-side control valve 2, a cold storage tank 3, a low-temperature heat exchanger 4, an expander 5, a regenerator 6, a compressor 7, a motor 8, a high-temperature heat exchanger 9, a hot-side control valve 10, a fan 11, a heat storage tank 12, and a cold-end heat exchanger 13. The refrigerant pump input end is connected to the cold-end heat exchanger, the refrigerant pump output end is connected to the first input end of the low-temperature heat exchanger, the first output end of the low-temperature heat exchanger is connected to the input end of the cold storage tank, the cold storage tank output end is connected to the cold-end heat exchanger, and the cold-side control valve is installed on the connecting pipe between the cold storage tank output end and the cold-end heat exchanger. The second output of the low-temperature heat exchanger is connected to the first input of the regenerator. The first output of the regenerator is connected to the compressor. The compressor is connected to both the motor and the first input of the high-temperature heat exchanger. The first output of the high-temperature heat exchanger is connected to the second input of the regenerator. The second input of the regenerator is connected to the expander. The expander is connected to the second input of the low-temperature heat exchanger. The expander is connected to both the compressor and the first output of the high-temperature heat exchanger. The second output of the high-temperature heat exchanger is connected to the first input of the heat storage tank. The second output of the heat storage tank is connected to the fan. The fan is connected to the second input of the high-temperature heat exchanger. A heat-side control valve is installed on the connecting pipe between the fan and the second input of the high-temperature heat exchanger.

[0024] In actual operation, due to the significant time delay characteristics of the cold and heat transfer process in the thermal battery energy storage system, the variable operating condition control of the thermal battery energy storage system faces many challenges. Although traditional PID control based on temperature or power feedback closed-loop regulation and its improved methods can respond quickly and suppress steady-state errors, they still face difficulties in tuning variable operating condition parameters and control response lag when dealing with large inertia, strongly coupled multi-timescale thermal systems such as thermal batteries with multiple loops of mutual coupling of cold storage, heat storage and heat pump units.

[0025] To address the aforementioned issues of high energy consumption, large fluctuations in output temperature and power, and severe equipment damage in thermal battery energy storage systems under multiple physical constraints such as dynamic electricity prices, load fluctuations, energy storage capacity, and equipment lifespan, this embodiment provides a multi-objective optimization control method for thermal battery energy storage systems based on model predictive control. By constructing a comprehensive objective function that includes economic cost, output temperature and power stability, rapid temperature tracking, energy storage health management, and actuator protection, and combining it with a control-oriented linear state-space model identified by the thermal battery energy storage system, the optimal control sequence is solved in a rolling manner within the prediction time domain. This achieves precise matching of system parameters under varying operating conditions, comprehensive energy efficiency optimization, and rapid response adjustment, effectively improving the intelligence level, control quality, and overall stability of the thermal battery energy storage system.

[0026] The following examples and specific application scenarios further illustrate the technical solution and effects of the present invention. The following examples are explanations of the present invention, but the present invention is not limited to the following examples.

[0027] An optimized control method for a thermal battery energy storage system specifically includes the following steps: Step 1: System initialization and parameter configuration.

[0028] System initialization includes: (1) First, set the basic timing parameters for predictive control of the thermal battery energy storage system model, including the sampling period ∆t and the prediction time domain length N. p Control the time domain length N c The prediction time domain length is greater than or equal to the control time domain length.

[0029] (2) Load the objective function weights w1, ..., w7 of the multi-objective optimization according to the selected running mode.

[0030] (3) Initialize the historical motion intensity S of each actuator i (0)=0.

[0031] Parameter configuration includes: (1) Configure temperature control dead zone parameters, including the hot end output temperature dead zone ∆T h and cold end output temperature dead zone ∆Tc .

[0032] (2) Set the maximum allowable load deviation of output power ∆Q, and configure the thermal storage operation window parameters. The operation window parameters include the upper and lower limits of thermal storage capacity operation and the upper and lower limits of cold storage capacity operation.

[0033] (3) Configure the actuator physical limit u min u max Actuator rate of change limit Δu min , Δu max .

[0034] (4) Configure motion frequency control parameters, including smoothing parameter η and motion dead zone threshold ε. i Historical action frequency influence factor β, learning rate λ, executor base frequency penalty weight w 0,i Actuator incremental weight r i The frequency of historical actions affects the upper limit of weight saturation S. max The penalty weighting coefficients w- and w+ for under-supply and over-supply.

[0035] Step 2: Real-time data acquisition and multi-source information fusion.

[0036] The underlying PLC controller collects system operation data in real time at the current moment t through sensors, including process parameters such as cold / hot end inlet and outlet temperatures, flow rate, and pressure, as well as equipment operation information such as motor frequency and the current position status of each actuator.

[0037] The edge controller receives the raw data uploaded from the underlying layer, performs data cleaning, outlier detection and processing, and extracts the feature variables required for optimization.

[0038] The extracted feature variables include: Filtered state features such as temperature, motor frequency, pump speed, fan speed, and valve opening are used for the initialization of the prediction model. Process characteristics used for objective function calculations include instantaneous electric power, instantaneous output heat / cold power, inlet and outlet temperature difference of heat exchanger, and average temperature of energy storage medium. Key statistical characteristics of variables used for anomaly detection include mean, variance, and rate of change. Derived features such as normalized thermal / cold storage capacity status and system instantaneous performance coefficients are used for energy storage status and performance assessment.

[0039] Simultaneously acquire external information in the current and future forecast time domains, including ambient temperature forecast sequences, waste heat source temperature forecast sequences, and time-of-use electricity price sequences. elec .

[0040] Read the system operation log and extract the historical action intensity S of each actuator at the previous moment.i (t-1) and control increment ∆u i (t-1), used to recursively update the action intensity at the current moment (system startup command S). i (0)=0); Read the current energy state of the thermal storage system and the cold storage system for system energy storage state calculation.

[0041] The thermal storage system refers to the loop consisting of a high-temperature heat exchanger 9, a hot-side control valve 10, a fan 11, and a thermal storage tank 12. The cold storage system is a loop consisting of a refrigerant pump 1, a cold-side control valve 2, a cold storage tank 3, and a low-temperature heat exchanger 4. The current energy state of the thermal and cold storage tanks is read for system energy storage state calculation.

[0042] Step 3: System state estimation and prediction initialization.

[0043] Specifically, the following steps are included: (1) Based on the preprocessed measurement data, the measured input and output data of the thermal battery energy storage system, combined with the state-space model obtained by data-driven identification, the optimal estimate of the system state vector x(t) is solved online recursively using the state observation algorithm. And use this state estimate as the initial state in the prediction time domain, that is, let x(t) represents the system state vector used to initialize the prediction model after optimal estimation by the state observer, including key state variables such as hot end outlet temperature, cold end outlet temperature, thermal storage capacity state, cold storage capacity state, and motor frequency of the integrated compression and expansion unit.

[0044] (2) Normalize the current energy storage state (SOC) of the thermal storage system and the cold storage system at time t to the interval [0,1]. The normalization process is as follows: The normalized energy state of a thermal storage system is the ratio of the difference between the actual energy stored in the thermal storage system and the minimum energy storage capacity of the thermal storage system to the difference between the maximum energy storage capacity and the minimum energy storage capacity of the thermal storage system.

[0045] The normalized energy state of a cold storage system is the ratio of the difference between the actual energy stored in the cold storage system and the minimum energy storage capacity of the cold storage system to the difference between the maximum energy storage capacity and the minimum energy storage capacity of the cold storage system.

[0046] (3) Update the action frequency penalty weight of each actuator based on historical action data.

[0047] (3.1) First, calculate the over-threshold action amount of the actuator and take the larger of the following two values: zero, or the difference between the absolute value of the actuator control increment and the action judgment threshold of the actuator.

[0048] The actuator's motion determination threshold is a preset dead zone parameter; minor motions below this threshold are not included in the motion frequency statistics.

[0049] (3.2) Based on this, calculate the normalized motion intensity of the actuator: the normalized motion intensity of the actuator is the ratio of the over-threshold motion quantity to the difference between the physical upper limit and the physical lower limit of the actuator.

[0050] The physical upper limit of an actuator is the maximum position value that the actuator is allowed to operate at, and the physical lower limit of an actuator is the minimum position value that the actuator is allowed to operate at.

[0051] (3.3) The historical motion intensity is updated recursively using the exponentially weighted moving average method: the historical motion intensity S of the actuator i (t) is the product of the forgetting factor and the historical action intensity of the actuator at the previous time step, plus the product of the difference between 1 and the forgetting factor and the normalized action intensity of the actuator at the previous time step. The forgetting factor ranges from 0 to 1, and a larger forgetting factor indicates a greater influence of historical actions.

[0052] (3.4) Based on the updated historical action intensity, calculate the action frequency penalty weight of each actuator in the current cycle: the action frequency penalty weight of the actuator in the current cycle is the sum of the product of the learning rate and the clipped historical action intensity plus 1, and the product of the actuator's base frequency penalty weight.

[0053] The learning rate controls the strength of the influence of historical actions on the weights; the strength of the clipped historical actions is clip(S). i (t),0,S max ), S max The pre-defined upper limit of the influence weight of the action frequency is defined by the clipping function, which restricts the influence weight of historical actions to [0, S]. max Within the interval ], when S i If (t) is less than 0, then take 0, S i (t) is greater than S max Then take S max S i (t) in [0,S max The values ​​between [ ] remain unchanged.

[0054] Step 4: Construct a multi-objective optimization problem.

[0055] Define and construct the objective function for multi-objective optimization of the thermal battery energy storage system. The objective function is the sum of the segment cost term and the control cost term.

[0056] The objective function to be minimized is the sum of the costs at each time step in the prediction time domain, plus the sum of the control costs at each time step in the control time domain.

[0057] Specifically: (1) Stage cost item.

[0058] The stage cost is accumulated over the entire prediction time domain Np, reflecting the economy and quality of the thermal battery energy storage system's operating status. It consists of five sub-items: energy cost, delivered power deviation, temperature stability control, temperature orientation adjustment, and energy storage status deviation.

[0059] That is, the stage cost is the product of the energy cost weight and the energy cost item, plus the product of the delivered power weight and the delivered power deviation penalty item, plus the product of the temperature stability weight and the temperature stability control penalty item, plus the product of the temperature orientation adjustment weight and the temperature orientation adjustment penalty item, plus the product of the energy storage state weight and the energy storage state deviation from the operating window penalty item.

[0060] (1.1) Energy cost item.

[0061] The energy cost item is used to characterize the operating electricity consumption cost of the thermal battery energy storage system under time-of-use pricing.

[0062] The energy cost item is the product of the time-of-use electricity price, the predicted value of system power consumption, and the discrete time step.

[0063] Among them, the time-of-use electricity price is given by electricity market data or electricity price prediction model (a market future electricity price prediction function based on historical electricity price data); the predicted value of system power consumption is calculated by the system energy consumption model based on the current system state and the control input to be optimized.

[0064] The energy cost term represents the electricity cost at each time point in the prediction time domain. By minimizing this term in the objective function, the system's operating electricity cost can be reduced.

[0065] (1.2) Delivery power deviation item.

[0066] The delivered power deviation item measures the degree of deviation between the system's output cooling and heating power and the user's demand settings, ensuring the quality and reliability of external power supply.

[0067] The difference between the power demand setpoint, the allowable power fluctuation margin, and the undersupply slack variable is less than or equal to the predicted system delivery power.

[0068] The sum of the power demand setpoint, the allowable power fluctuation margin, and the oversupply slack variable is greater than or equal to the predicted system delivery power.

[0069] Furthermore, both the undersupply slack variable and the oversupply slack variable are greater than or equal to 0.

[0070] Based on the above, we can define a delivery power deviation term: the delivery power deviation penalty is the product of the under-supply penalty weight coefficient and the square of the supply slack variable, plus the sum of the product of the over-supply penalty weight coefficient and the square of the over-supply slack variable.

[0071] The delivered power deviation penalty represents the thermal or cold power delivered by the system at time t+j as output by the prediction model; the allowable power fluctuation margin reflects the combined effects of equipment inertia, measurement noise, and acceptable fluctuations at the end.

[0072] When the penalty weighting coefficient for undersupply is greater than that for oversupply, the undersupply situation will be subject to greater penalties, and the control strategy will prioritize ensuring sufficient supply; conversely, a certain degree of supply shortage will be allowed in exchange for improvements in other objectives.

[0073] By introducing slack variables, the power delivery constraint is transformed from a hard constraint into a soft constraint, ensuring that the optimization problem still has a feasible solution under extreme conditions. At the same time, the squared penalty form makes the penalty grow faster as the deviation increases.

[0074] (1.3) Temperature stability control item.

[0075] The temperature stability control term is used to constrain the deviation of the output temperature at the cold and hot ends of the system, ensuring the steady-state accuracy of the output temperature.

[0076] First, calculate the effective temperature error of the hot end and the cold end separately, and then define the temperature stability control term based on this. The calculation method for the effective temperature error of the hot end is as follows: First, calculate the absolute value of the difference between the predicted hot-end temperature and the target hot-end temperature, and then subtract the maximum permissible deviation of the hot-end temperature. If the difference is positive, take the positive value as the effective hot-end temperature error. If the difference is zero or negative, the predicted hot-end temperature is within the permissible temperature fluctuation range, the effective hot-end temperature error is zero, and there is no penalty for the hot-end temperature.

[0077] Effective cold junction temperature error: First, calculate the absolute value of the difference between the predicted cold junction temperature and the target cold junction temperature, and then subtract the cold junction temperature dead zone width. If the difference is positive, take the positive value as the effective cold junction temperature error. If the difference is zero or negative, the predicted cold junction temperature is within the allowable temperature fluctuation range, the effective cold junction temperature error is zero, and there is no penalty for the cold junction temperature.

[0078] Based on the effective temperature error between the hot and cold ends, the temperature stability control term is defined and calculated: the temperature dead zone control penalty (temperature stability control term) is the sum of the square of the effective temperature error between the hot and cold ends.

[0079] When the predicted temperature is within the range of the setpoint plus or minus the dead zone, the corresponding error term is zero, and no penalty cost is generated. This is used to effectively avoid the controller from unnecessary frequent operation due to small disturbances near the temperature setpoint. Only when the predicted temperature deviates from the setpoint and exceeds the dead zone range does the penalty cost begin to accumulate, and it increases rapidly with the square of the deviation, ensuring that a sufficiently strong corrective driving force can be obtained after the temperature deviation exceeds the limit.

[0080] (1.4) Temperature-oriented regulation item.

[0081] The temperature-oriented regulation term is used to characterize the tracking error of the cold and hot end output temperatures on the dynamic reference trajectory. It provides the system with an active and directional temperature regulation driving force when the load changes suddenly or the operating conditions switch, thereby accelerating the convergence speed of the temperature towards the target value.

[0082] The temperature-oriented adjustment term is calculated as follows: First, the square of the difference between the predicted cold-end temperature and the dynamic reference trajectory of the cold-end temperature is calculated and multiplied by the cooling direction adjustment coefficient; then, the square of the difference between the predicted hot-end temperature and the dynamic reference trajectory of the hot-end temperature is calculated and multiplied by the heating direction adjustment coefficient; the sum of the two products is the temperature-oriented adjustment term.

[0083] The cooling direction adjustment coefficient and the heating direction adjustment coefficient complement the temperature dead zone control term: the dead zone control term focuses on maintaining the temperature during steady-state operation, while the directional adjustment term focuses on tracking the temperature trajectory during dynamic processes. When the system performs operations that require rapid temperature rise or fall, such as responding to sudden load changes or adjusting energy storage temperature, by setting a dynamic reference trajectory and increasing the corresponding adjustment coefficient, a more proactive and powerful adjustment driving force can be provided for temperature changes than simple dead zone penalty, thereby improving the system's dynamic response capability.

[0084] (1.5) Energy storage state deviation.

[0085] The energy storage state deviation term is used to measure the degree to which the thermal / cold storage capacity deviates from the recommended operating window, avoiding the system from operating at extreme boundaries close to full charge or full discharge for a long time, and reserving sufficient bidirectional adjustment margin for subsequent scheduling.

[0086] The deviation of the energy storage state of thermal and cold storage systems from the recommended operating window is calculated. Taking the thermal storage system as an example, if the predicted thermal energy state is lower than the lower limit of the thermal storage operating window, the deviation is the difference between the lower limit and the predicted thermal energy state; if the predicted thermal energy state is higher than the upper limit, the deviation is the difference between the predicted thermal energy state and the upper limit; if the predicted thermal energy state is between the upper and lower limits of the operating window, the deviation is zero. The calculation method for cold storage systems is the same as that for thermal storage systems.

[0087] Based on the energy storage state deviation of thermal and cold storage systems, the energy storage state deviation term is defined as follows: the total energy storage state deviation is the product of the thermal storage deviation penalty coefficient and the square of the thermal storage system operating window deviation, plus the product of the cold storage deviation penalty coefficient and the square of the cold storage system operating window deviation at that moment.

[0088] The penalty coefficient for thermal and cold storage systems deviating from the optimal energy storage operating window is mainly used to maintain the energy storage state within the preset operating window, avoid the system being in an extreme state of overcharging or over-discharging for a long time, ensure a certain system regulation margin, allow the system to maintain buffering capacity under uncertain loads, and ensure the safety of the energy storage system.

[0089] (2) Control cost item.

[0090] The control cost is accumulated over the entire control time domain Nc, reflecting the rationality of the actuator's actions and the degree of equipment protection. It consists of two sub-items: a control increment smoothing term and an action frequency control term.

[0091] That is, the control cost is the product of the action frequency weight and the action frequency penalty term, plus the product of the control increment weight and the control increment smoothing penalty term.

[0092] (2.1) Control the incremental smoothing term.

[0093] The control increment smoothing term is used to constrain the single-step motion amplitude of each actuator, suppress large jumps in control quantity, reduce mechanical shock and component wear, and ensure the smooth operation of the actuator.

[0094] Control increment smoothing term: For each actuator, calculate the product of the actuator's increment weight and the square of the actuator's control increment; sum the calculation results of all actuators to obtain the control increment smoothing term.

[0095] For each actuator, a larger incremental weight indicates stronger suppression of the actuator's actuation amplitude (valve opening, rotational speed, etc.). The incremental weight of each actuator is set differently according to the characteristics of the equipment. A larger weight can be set for valves with fast response speed but sensitive mechanical life, while a smaller weight can be set for pumps with strong inertia but good durability. This item suppresses large jumps in the actuator between adjacent sampling times by penalizing the square of the control increment, thereby protecting the equipment, reducing mechanical wear, and reducing system noise.

[0096] (2.2) Action frequency control item.

[0097] The motion frequency control item is used to measure the cumulative motion load of each actuator in the recent historical period. By adaptively updating the frequency penalty weight based on the historical motion intensity, actuators that have been moving frequently recently will be subject to greater penalties, while actuators that have been moving less recently will have more flexible response space. This achieves automatic balancing of the motion load of each actuator and extends the overall service life of the equipment.

[0098] For the control increment of the actuator at a certain moment, first calculate its smoothed absolute value (by adding the square of the control increment to the square of a very small smoothing parameter, and then taking the arithmetic square root to ensure that it is mathematically differentiable at the zero point).

[0099] Subtract the actuator's dead zone threshold from the smoothed absolute value; if the difference is positive, take the positive value as the effective motion amount exceeding the dead zone; if the difference is zero or negative, take zero as the effective motion amount. Minor motions smaller than the dead zone threshold are considered normal adjustments and are not included in the frequency penalty.

[0100] The motion frequency control term is as follows: For each actuator, calculate the motion frequency penalty weight of the actuator and the square of the effective motion amount of the actuator exceeding the dead zone; the sum of the calculation results of all actuators is the motion frequency control term.

[0101] The actuator's action frequency penalty weight is updated based on the historical action intensity. This weight is updated once at the beginning of each predictive control solution cycle and remains constant throughout the entire prediction time domain.

[0102] The motion frequency control term is used to suppress the motion frequency of the adaptive actuator. For actuators with frequent recent motions, the historical motion intensity S is used. i A larger value for (t) leads to a greater penalty weight for the actuator's motion frequency. During optimization, the actuator's motion will incur a greater penalty cost, making it more likely to remain stationary or reduce its motion amplitude in subsequent periods. Conversely, actuators with fewer recent motions have relatively smaller penalty weights and can respond more flexibly when adjustments are needed. By applying motion frequency penalties, the system can automatically balance the motion load of each actuator, preventing individual actuators from prematurely failing due to continuous high-frequency motion and effectively extending the overall equipment lifespan.

[0103] The objective function adjusts the relative importance of different optimization objectives through various weight coefficients. The weight coefficients are summed to 1, and their values ​​are configured according to the user-selected energy supply mode (high-efficiency mode, constant temperature mode, or temperature regulation mode).

[0104] Step 5: Set hard constraints.

[0105] In this embodiment, the constructed multi-objective optimization problem needs to satisfy the following hard constraints: (1) State equation constraints.

[0106] Predicting the evolution of the system state in the time domain requires following the dynamic equations described by the identified linear state-space model (inputs are motor frequency, fan speed, pump speed, and valve opening; outputs are motor power, output power, output temperature, and energy storage state). The predicted value of the system state at the next moment is calculated by the dedicated prediction model for the thermal battery energy storage system (based on the control-oriented linear state-space model identified from the physical system) based on the current system state, control input, and external disturbance input (including ambient temperature, waste heat source temperature, etc.).

[0107] This control-guided linear state-space prediction model is not only used to construct hard constraints for the state equations, but also to provide the required state and output prediction sequences for each cost term in the stage objective function in the fourth step, such as the predicted power value in the energy cost term, the predicted output power value in the delivered power deviation term, the predicted temperature value in the temperature stability term and the temperature orientation adjustment term, and the predicted energy storage state value in the energy storage state deviation term. It is the basis for realizing multi-objective rolling optimization solution.

[0108] (2) Control the constraints outside the time domain.

[0109] The optimization variables are limited to Nc control actions in the control time domain, which is used to reduce the number of optimization decision variables, reduce computational complexity and improve solution efficiency.

[0110] The control quantity at each time point outside the control time domain is equal to the control quantity at the end of the control time domain.

[0111] (3) Actuator physical limit constraint.

[0112] The lower physical limit of the actuator is less than or equal to the upper physical limit of the actuator at each time. The lower physical limit vector and upper physical limit vector of each actuator include the frequency range of the frequency converter, the valve opening range, the pump speed range, etc.

[0113] (4) Actuator change rate limit constraint.

[0114] The control increment must be within the range of the actuator's allowable rate of change, that is: the minimum allowable change of the actuator in a single step is less than or equal to the control increment at each time step is less than or equal to the maximum allowable change of the actuator in a single step step.

[0115] The minimum and maximum allowable changes of each actuator within a single time step are used to prevent rapid changes in control commands from causing mechanical shocks to the actuators or leading to system oscillations.

[0116] (5) Physical constraints of energy storage.

[0117] The state of charge of energy storage must be within the physical safety boundary, that is, the lower limit of physical safety of energy storage is less than or equal to the state of charge of energy storage at each time and is less than or equal to the upper limit of physical safety of energy storage.

[0118] Unlike the soft constraints of the energy storage operation window, the physical limit constraint is a hard boundary that must not be violated. Exceeding this boundary may lead to damage to the energy storage medium or a safety accident.

[0119] (6) Relaxed variable nonnegativity constraint.

[0120] The slack variables introduced in the soft constraint of power delivery are only used to characterize the degree of insufficient or excessive power delivery, and must be non-negative, that is, the slack variable for insufficient delivery is greater than or equal to zero at each time step, and the slack variable for excessive delivery is greater than or equal to zero at each time step.

[0121] Step 6: Optimize the solution and instruction execution.

[0122] The edge controller invokes the numerical optimization solver Gurobi to solve the multi-objective optimization problem constructed in step four, while satisfying all the constraints set in step five, to obtain the optimal control sequence U. * (t)=[u * (t),...,u * t+Nc-1), based on the rolling time-domain principle of predictive control, only the first element u in the optimal control sequence is extracted. * (t) serves as the actual control command at the current moment, which is then sent from the edge controller to the underlying PLC controller. The underlying controller drives each actuator to perform actions according to the command settings, completing specific control operations such as motor frequency adjustment, fan speed adjustment, pump speed adjustment, and valve opening adjustment.

[0123] Step 7: Rolling Iteration and Closed-Loop Feedback.

[0124] As time progresses to t+1, the system returns to the second step to re-collect measurement data, update the system state estimate and various prediction information, and reconstruct and solve the optimization problem based on the new initial state. Through the iterative mechanism of the rolling time domain, a complete closed-loop predictive control is formed, encompassing data uploading, optimization decision-making, command issuance, rapid execution, and feedback updates. This enables the system to continuously adapt to changes in operating conditions and constantly modify the control strategy to cope with the effects of model uncertainties and external disturbances.

[0125] Unless otherwise specified, all systems mentioned in this embodiment are thermal battery energy storage systems.

[0126] The optimized control method for the thermal battery energy storage system provided in this embodiment has the following beneficial effects: (1) Multi-objective dynamic collaborative optimization, taking into account both economy and energy quality.

[0127] This invention integrates multiple objectives, including time-of-use pricing response, power delivery quality, temperature control accuracy, energy storage health management, and equipment protection, into a predictive control optimization framework. A weighted objective function is used to achieve dynamic coordination among these multiple objectives. Unlike traditional single-objective or single-loop control strategies, this invention comprehensively considers future electricity price trends and system evolution within the prediction time domain, automatically planning the optimal control sequence. This significantly reduces energy time shift and operating costs while ensuring stable output of energy supply temperature and power.

[0128] (2) Enhance the dynamic response capability of the system and suppress overshoot and oscillation.

[0129] To address the control lag and oscillation overshoot issues caused by the large inertia and strong coupling characteristics of thermal batteries, a high-precision dynamic prediction model is combined with temperature and power dead-zone design to achieve forward-looking optimization and fine-tuning of the system. The model predicts system behavior in advance, and the dead-zone is used to avoid overreacting to minor disturbances, reducing temperature and power output overshoot and fluctuation amplitude, thus improving system control accuracy and stability. Furthermore, a temperature-oriented adjustment term is introduced to further enhance the system's dynamic tracking capability.

[0130] (3) Robust control strategy with soft and hard constraints to ensure feasibility under extreme conditions.

[0131] A flexible constraint handling strategy combining hard and soft constraints is adopted. Hard constraints are used to strictly guarantee the inviolable safety constraints such as the physical safety boundary of energy storage and the physical limit of actuators. For performance constraints that allow moderate deviations, such as power delivery targets and energy storage operation windows, slack variables and penalized soft constraints are introduced to effectively avoid the dilemma that traditional hard constraint control may not be able to solve under extreme conditions such as load changes and equipment failures. This ensures that the optimization problem always has a feasible solution, and the larger the deviation, the stronger the penalty is, which takes into account the tracking accuracy of the control target and the robustness of the system through the square penalty form.

[0132] (4) Adaptive actuator protection mechanism to extend equipment service life.

[0133] An adaptive frequency penalty mechanism based on historical motion intensity is proposed. This mechanism uses an exponentially weighted moving average method to track the historical motion frequency of each actuator in real time and dynamically adjusts its penalty weight in the optimization objective. Actuators with frequent recent motions will receive stronger motion suppression in subsequent optimizations, while actuators with fewer recent motions will maintain higher adjustment flexibility. Combined with the constraint of motion amplitude by a control increment smoothing term, this achieves intelligent balancing of the motion load of each actuator, effectively preventing premature damage to individual actuators due to continuous high-frequency, large-amplitude motions, significantly extending the overall equipment lifespan and reducing maintenance costs.

[0134] Example 2: Based on Embodiment 1, this embodiment provides an optimized control system for a thermal battery energy storage system, used to implement the optimized control method for the thermal battery energy storage system in Embodiment 1. It includes a bottom control layer, an edge control layer, and an upper control layer. It is a multi-level, multi-objective optimization control system that can simultaneously consider indicators such as energy cost, delivered power quality, temperature control accuracy, energy storage status, and equipment protection. It can achieve real-time online coordinated optimization of multiple performance indicators of the system during operation.

[0135] like Figure 3 As shown, the entire optimization control system adopts a hierarchical architecture design, consisting of three levels: the bottom control layer, the edge control layer, and the upper optimization decision layer.

[0136] (1) Bottom control layer.

[0137] The underlying control layer is based on a PLC controller. It uses sensors to collect parameters such as motor power, inlet and outlet temperature, pressure, flow rate, and actuator status of heat pumps and energy storage devices in real time. It drives actuators such as motors, fans, pumps, and valves with millisecond-level response speed to achieve rapid tracking and execution of control commands.

[0138] (2) Edge control layer.

[0139] As a data hub and coordination node between the lower-level control and the upper-level optimization decision-making, the edge control layer performs preprocessing such as cleaning and feature extraction on the real-time operation data uploaded by the lower-level PLC, and integrates multi-source information such as operation logs (historical sequence of actuator actions, thermal / cold storage capacity status) and external data (ambient temperature, waste heat source temperature, electricity price at different times) and uploads the processed data to the upper-level optimization decision-making layer. On the other hand, it receives the optimal control sequence issued by the upper-level optimization decision-making layer and issues optimization instructions such as motor frequency, circulating pump speed, fan speed, and valve opening setpoints to the lower-level PLC for execution.

[0140] (3) Upper control layer.

[0141] The upper-level optimization decision layer is deployed on a local server. It takes the control-oriented predictive model based on the physical system identification as the core. Under the constraints of actuator physical and rate of change limits, energy storage physical limits, and non-negativity of relaxation variables, it constructs a multi-objective optimization problem in the prediction time domain with energy cost, delivered power quality, temperature quality, temperature-oriented regulation rate, energy storage state management, and actuator amplitude / frequency as control objectives. It then uses the optimization solver Gurobi to perform rolling optimization and sends optimization instructions for motor frequency, circulating pump speed, fan speed, and valve opening setpoints to the lower-level PLC for execution, forming a complete closed-loop predictive control.

[0142] The optimized control system provided in this embodiment can reduce the online computational load and meet real-time control requirements. Specifically, it adopts a hierarchical control architecture. The upper-level optimization decision control layer uses the Gurobi solver to quickly solve multi-objective optimization problems in the prediction time domain, comprehensively considering the system's dynamic characteristics, multiple constraints, and future disturbance predictions, and outputs the globally optimal control sequence. The lower-level execution layer uses a PLC controller as its core, achieving fast and accurate tracking of control commands with a millisecond-level response cycle. The two layers are functionally separated and complementary, with the upper layer ensuring the global optimality of decision-making and the lower layer ensuring real-time execution, effectively solving the real-time control problem of thermal battery energy storage systems and achieving the best balance between optimization depth and response speed.

[0143] The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any way. Other variations and modifications are possible without departing from the technical solutions described in the claims.

Claims

1. An optimized control method for a thermal battery energy storage system, characterized in that, include: S1: Initialize the historical motion intensity of each actuator to zero and configure the control parameters; S2: Real-time acquisition and optimization of control parameters and fusion of multi-source information; S3: Estimate and update the current system state, and dynamically adjust the action frequency penalty weight of each actuator in the optimization objective based on historical action data; S4: Using the sum of the stage cost term and the control cost term as the objective function for multi-objective optimization, construct hard constraints, solve the objective function, and obtain the optimal control sequence; S5: Extract the first element from the optimal control sequence as the actual control command at the current moment for optimized control.

2. The optimized control method for a thermal battery energy storage system according to claim 1, characterized in that, The dynamic adjustment of the action frequency penalty weight of each actuator in the optimization objective based on historical action data includes: Calculate the over-threshold action amount of the actuator, which is either zero or the larger of the difference between the absolute value of the actuator control increment and the actuator's working determination threshold. Based on the over-threshold motion quantity, the normalized motion intensity of the actuator is calculated; Based on the forgetting factor, the historical action intensity and normalized action intensity of the actuator at the previous moment are recursively updated using the exponential weighted moving average method. Based on the updated historical action intensity, calculate the action frequency penalty weight for each actuator in the current cycle.

3. The optimized control method for a thermal battery energy storage system according to claim 1, characterized in that, The stage cost term is accumulated over the entire prediction time domain, and the stage cost term includes: The energy cost item is obtained based on time-of-use electricity pricing, predicted system power consumption, and discrete time step; the delivered power deviation item is obtained based on the under-supply / over-supply penalty weighting coefficient and under-supply / over-supply slack variable. The temperature stability control item does not generate a penalty when the predicted temperature is within the set value plus or minus the dead zone; otherwise, it accumulates a penalty cost, which increases with the square of the temperature deviation. The temperature-oriented adjustment term is obtained based on the temperature adjustment coefficient and the temperature dynamic reference trajectory. The energy storage state deviation is obtained based on the deviation of the energy storage state from the recommended operating window.

4. The optimized control method for a thermal battery energy storage system according to claim 1, 2, or 3, characterized in that, The hard constraints include state equation constraints, control time domain hold constraints, actuator physical limit constraints, actuator rate of change limit constraints, energy storage physical constraints, and relaxation variable non-negativity constraints. The control time domain hold constraints mean that the controller at each time outside the control time domain is equal to the control quantity at the end of the control time domain.

5. The optimized control method for a thermal battery energy storage system according to claim 1 or 3, characterized in that, The control cost term is accumulated over the entire control time domain, and the control cost term includes: The control increment smoothing term is the sum of the products of the increment weights of all actuators and the squares of the control increments. The action frequency control term is the sum of the product of the action frequency penalty weights of all actuators and the square of the effective action amount of the actuators exceeding the dead zone. The action frequency penalty weights are updated once at the beginning of the predictive control solution cycle and remain constant throughout the prediction time domain.

6. The optimized control method for a thermal battery energy storage system according to claim 1, 2, or 3, characterized in that, The real-time acquisition and optimization control parameters include: Real-time acquisition and preprocessing of system operation data at the current moment; extraction of feature variables for optimization solution. Simultaneously acquire external information for current and future forecast time series, including environmental temperature forecast sequence, waste heat temperature forecast sequence, and time-of-use electricity price sequence; Read the system operation log and extract the historical action intensity and control increment of each actuator at the previous moment; Obtain the current energy state of the system.

7. The optimized control method for a thermal battery energy storage system according to claim 2, characterized in that, S3 includes: obtaining the optimal estimate of the system state vector at the current moment using a state estimation method based on the collected system operation data, and establishing initial condition constraints; calculating the current energy storage state of the system and normalizing it to the [0,1] interval.

8. The optimized control method for a thermal battery energy storage system according to claim 1, 2, 3, or 7, characterized in that, The characteristic variables include state characteristics, process characteristics, statistical characteristics, and derived characteristics. The state characteristics include filtered temperature, motor frequency, pump speed, fan speed, and valve opening. The process characteristics include instantaneous electrical power, instantaneous output heat / cold power, inlet and outlet temperature difference of the heat exchanger, and average temperature of the energy storage medium. The derived characteristics include normalized thermal / cold storage capacity state and system instantaneous performance coefficient.

9. The optimized control method for a thermal battery energy storage system according to claim 1, 2, 3, or 7, characterized in that, The configuration control parameters include: setting the basic timing parameters of the predictive control of the thermal battery system, including the sampling period, prediction time domain length, and control time domain length; obtaining the objective function weights for multi-objective optimization based on the selected operating mode; configuring the temperature control dead zone parameters; setting the maximum allowable load deviation of the output power; configuring the operating window parameters of the thermal battery energy storage system; configuring the actuator physical limit and actuator rate of change limit; and configuring the action frequency control parameters.

10. The optimized control method for a thermal battery energy storage system according to claim 6, characterized in that, The system operation data includes process parameters and equipment operation information. The process parameters include cold / hot end inlet and outlet temperatures, flow rates, and pressures. The equipment operation information includes motor power and the current position status of each actuator.

Images