Multi-region heterogeneous energy storage joint optimization method based on model predictive control

Abstract

The invention belongs to the technical field of energy storage joint optimization, and provides a multi-region heterogeneous energy storage joint optimization method based on model predictive control, which adopts slow time domain rolling optimization on an upper layer to realize joint scheduling of economy and service life. Fast time domain distributed power distribution and intra-group consistency control are adopted at the lower layer to realize fast frequency response and SOC balance; the upper layer cost function merges the operation loss cost, the time-of-use electricity price income and the degradation cost of unified monetization into the same target, and takes the energy storage unit as the dynamic adjustment basis of the degradation unit price to protect the aging unit; and the lower layer dynamically activates the power type and energy type energy storage units based on a disturbance severity index, formulates a charging and discharging group division and switching strategy, and adopts a distributed consistency algorithm to realize power distribution of the minimum communication overhead in the group so as to balance the SOC, so that the instantaneous frequency performance, the long-term economy and the equipment life are considered under the realizable engineering condition.

Description

Technical Field

[0001] This invention belongs to the field of energy storage joint optimization technology, specifically involving a multi-regional heterogeneous energy storage joint optimization method based on model predictive control. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] With the significant increase in the installed capacity of intermittent renewable energy sources such as wind power and photovoltaics, the power system exhibits the dual characteristics of high output fluctuation, intensified frequency disturbances, and dominance of power electronics. Traditional frequency regulation methods, mainly based on unit inertia and slow-sequence scheduling, are clearly insufficient under rapid disturbance conditions. The system places higher demands on the coordinated ability of short-term high-power rapid response and long-term energy support. When facing a high proportion of renewable energy, the existing frequency regulation framework often requires external large-capacity energy storage to compensate for rapid power gaps. However, it is difficult for a single energy storage medium to achieve the optimal balance between instantaneous power, continuous energy, and lifetime cost.

[0004] Current research on frequency control for heterogeneous energy storage mainly focuses on real-time power allocation and state of charge (SOC) balancing, but neglects the decoupling of economic optimization and physical control, as well as the integration of market participation such as time-of-use pricing arbitrage. Furthermore, existing degradation cost models often use fixed parameters and do not consider dynamic adjustments based on state of health (SOH), leading to overuse of aging units. Simultaneously, grouping strategies are mostly static and cannot dynamically respond to the severity of disturbances, limiting the robustness and economics of the system in high-penetration renewable energy scenarios. These shortcomings pose challenges to day-ahead scheduling and in-depth exploration of the potential of heterogeneous energy storage. Summary of the Invention

[0005] To address the aforementioned problems, this invention proposes a multi-regional heterogeneous energy storage joint optimization method based on model predictive control. At the upper layer, slow-time-domain rolling optimization is employed to achieve joint scheduling considering both economic efficiency and lifespan. At the lower layer, leveraging the multi-timescale characteristics of heterogeneous energy storage systems, fast-time-domain distributed power allocation and intra-group consistency control are used to achieve rapid frequency response and SOC balance. The upper-layer cost function incorporates operating loss costs, time-of-use electricity price revenue, and uniformly monetized degradation costs into a single objective, and uses the SOH of energy storage units as the basis for dynamic adjustment of degradation unit prices to protect aging units. At the lower layer, power-type and energy-type energy storage units are dynamically activated based on disturbance severity indicators, and charging / discharging group division and switching strategies are formulated. A distributed consensus algorithm is used to achieve power allocation with minimal communication overhead within groups to balance SOC, thereby balancing instantaneous frequency performance, long-term economic efficiency, and equipment lifespan under feasible engineering conditions.

[0006] According to some embodiments, the present invention adopts the following technical solution: A joint optimization method for multi-regional heterogeneous energy storage based on model predictive control includes the following steps: Based on the physical parameters of the heterogeneous energy storage system, construct a heterogeneous energy storage system model and a power grid frequency response model; Based on the constructed model, a joint optimization model based on hierarchical distributed model predictive control is constructed. The upper layer of the joint optimization model considers the degradation cost of heterogeneous energy storage and time-of-use electricity price arbitrage, with the goal of maximizing revenue and minimizing cost, in order to delay the degradation problem of heterogeneous energy storage system and actively participate in the energy market to achieve peak-valley arbitrage. The lower layer aims for optimal frequency regulation performance, considering response strategies based on the severity of disturbances and charging / discharging grouping strategies for heterogeneous energy storage systems to reduce the number of charging / discharging cycles of heterogeneous energy storage systems and thus extend their lifespan. At the same time, a distributed consensus-based intra-group SOC recovery strategy is adopted to unify the intra-group SOC, and finally, power commands for each energy storage unit are generated to provide power support for frequency regulation while taking into account economic efficiency.

[0007] As an alternative implementation method, the process of constructing a heterogeneous energy storage system model includes: the heterogeneous energy storage system includes a power-type energy storage system and an energy-type energy storage system, each region is configured with a heterogeneous energy storage unit group, and a first-order damping element is used to describe the equivalent model of the heterogeneous energy storage system in order to characterize its dynamic response delay.

[0008] As an alternative implementation method, the process of constructing the power grid frequency response model includes: based on the heterogeneous energy storage transfer function model, ignoring the constraints of the frequency regulation unit, namely dead zone, slope constraint and upper and lower limit constraints, to represent the frequency response dynamic model of each power system region.

[0009] As an alternative implementation, the time scale of the upper layer is slower than that of the lower layer, and both the upper and lower layers are controlled using model predictive control (MPC).

[0010] As an alternative implementation, both the upper and lower layers utilize independent controllers. The upper controller transmits power commands and dynamic adjustment coefficients to the lower controller. The power commands output by the lower controller then affect the SOC and SOH of the energy storage unit. Changes in various state variables are ultimately fed back to the upper and lower controllers, influencing their subsequent decisions and forming a closed-loop regulation.

[0011] As an alternative implementation, the cost function of the upper-level controller is used as the objective function to generate the command signals allocated to the lower-level controller. The process of establishing the cost function includes: establishing the basic operating loss function of thermal power units and energy storage units, using a quadratic cost model to represent the static operating loss cost, constructing the calculation expression for the incremental operating loss cost, and the cost function also considers the degradation cost of heterogeneous energy storage systems, the unit energy throughput degradation cost of heterogeneous energy storage systems on the same scale, and time-of-use electricity price arbitrage.

[0012] As a further implementation method, the cost function also considers time-of-use (TOU) arbitrage. TOU arbitrage is used to encourage energy storage systems to charge during off-peak hours and discharge during peak hours, thereby enabling active participation in the energy market. Defining the output power of the energy storage system as positive during discharge and negative during charging, the expression considering TOU arbitrage is as follows:

[0013] In the formula λ TOU To predict electricity prices, economic incentives are created: when predicted electricity prices are at their lowest, C TOU When the value is negative, the driver controller formulates a charging plan; when the electricity price is predicted to be at its peak, C TOU When the value is positive, the controller is driven to discharge to obtain the maximum economic return.

[0014] As a further implementation, a frequency offset penalty term is introduced into the upper-level controller to prevent the upper-level controller from sacrificing the frequency support capability of the heterogeneous energy storage system when pursuing economic efficiency. By penalizing the frequency deviation, the optimization is guided towards a stable direction. This penalty term is expressed as:

[0015] In the formula λ fre This is the frequency offset penalty coefficient.

[0016] As a further implementation method, after optimization, the upper-level controller generates the classification power command that the heterogeneous energy storage systems in each region should follow, the common power command that the thermal power units should follow, and the dynamic adjustment coefficient within a set time scale in the future. These are then issued to the lower-level controller as top-level commands, clarifying the reference power that various activated heterogeneous energy storage systems should share in the next scheduling cycle.

[0017] As an alternative implementation method, the process of considering a response strategy based on the severity of the disturbance includes: The most suitable energy storage system is dynamically activated based on real-time frequency disturbance characteristics to optimize response efficiency and reduce losses. The disturbance severity is calculated based on the different response characteristics of heterogeneous energy storage systems to implement a response grouping strategy. The disturbance severity function is expressed as:

[0018] In the formula w 1, w 2 and w 3 represents the weighting coefficients based on historical data, and RoCoF represents the rate of change of frequency. T dru To estimate the duration of the disturbance, this function considers both the amplitude and velocity of the disturbance, avoiding the limitations of a single metric by setting a threshold. S low < S high And dynamically group according to S; If S < S low If the disturbance is determined to be minor, the power storage unit is activated. For the median value S low <S< S high This activates all types of energy storage systems to respond to the frequency deviation. If S> S high If the disturbance is deemed a major disturbance, all available energy storage units will be activated to provide large-capacity continuous support. The threshold is adaptively adjusted based on historical data, forming a real-time feedback loop from disturbance detection to energy storage system activation.

[0019] As an alternative implementation, the charging and discharging grouping strategy of the heterogeneous energy storage system is to equally divide the multiple units of the heterogeneous energy storage system into charging groups and discharging groups, and to prioritize the output power of the charging groups and the discharging groups. To balance reducing the number of charge / discharge cycles and increasing the overall available power of heterogeneous energy storage systems in grouped operation mode, the following coordinated control strategy is formulated: when ≤| KP ESU When | it indicates that a single charge / discharge group can meet the reference power output, only a single group needs to be activated; When | KP ESU |< ≤|2 KP ESUWhen | indicates that a single charge / discharge group cannot meet the reference power output, a short-term switching strategy for charge / discharge states is used to supplement the rated power deficiency of a single group. KP ESU | To set a threshold.

[0020] As an alternative implementation, the process of unifying the SOC within a group using a distributed consensus-based intra-group SOC recovery strategy includes: implementing SOC-based adaptive power allocation within each group, so that each group meets the reference power... At the same time, we try to maintain the balance of SOC within the group and reduce lifetime consumption. We actively promote the SOC of all units in the group to be consistent by using a negative feedback adjustment mechanism based on state deviation. During adaptive allocation, each energy storage unit acts as an intelligent agent, interacting only with neighboring units in the communication network to collaboratively decompose the total power command issued by the lower-level controller according to its own SOC, ultimately ensuring that the power output of each unit is precisely matched with its SOH and SOC.

[0021] Compared with the prior art, the beneficial effects of the present invention are as follows: This invention unifies the degradation costs of three heterogeneous energy storage systems—batteries, supercapacitors, and flywheels—with vastly different physical characteristics into a monetization model based on unit energy throughput. This model innovatively introduces State of Health (SOH) as a dynamic adjustment factor, enabling lifetime costs to adaptively adjust with equipment aging, avoiding the overuse of energy storage units by traditional fixed-parameter models. Furthermore, by incorporating this dynamic degradation cost, operating loss cost, and time-of-use pricing arbitrage revenue into the unified objective function of the upper-level MPC (Multi-Level Marketing) system, this invention allows the controller to weigh immediate benefits such as market arbitrage and frequency regulation services against long-term investments such as extending equipment lifespan within the same economic dimension, thereby making globally optimal economic dispatch decisions and significantly enhancing the overall economic value of the energy storage system throughout its entire lifecycle.

[0022] In the lower-level control, this invention first implements dynamic functional grouping of heterogeneous energy storage based on the severity of disturbances, ensuring precise activation of power-type and energy-type energy storage units. This approach minimizes losses during high-frequency disturbances while providing sufficient energy support for significant disturbances. Building upon this, by dividing homogeneous energy storage units into charge and discharge groups and designing an inter-group switching mechanism based on SOC thresholds, unnecessary charge / discharge state transitions of energy storage units are significantly reduced, effectively decreasing the number of cycles and thus extending the overall collaborative operating life of the energy storage array. This layered strategy across different time scales achieves dual optimization of performance and lifespan. This invention decouples the upper-level slow-time-domain economic planning from the lower-level fast-time-domain physical control through a layered design. A distributed consensus algorithm is innovatively introduced at the lower level. This architecture allows the upper-level controller to focus on economic objectives, while the lower-level controller only needs to output categorized total power commands, without managing the individual behavior of a massive number of units. Specific power allocation tasks are delegated to the local controllers of each energy storage unit, autonomously and collaboratively completed through local communication. This significantly reduces the requirements for the central controller and communication system, solving the communication bottlenecks, computational burdens, and high latency problems faced by centralized control of large-scale energy storage arrays. It ensures the speed and reliability of system control and possesses excellent scalability.

[0023] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0024] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0025] Figure 1 This is a schematic diagram of the method flow of the present invention; Figure 2 This is a regional frequency response framework diagram of a power system with a heterogeneous energy storage system used in this invention; Figure 3 This is an overall flowchart of the present invention; Figure 4 This is a flowchart of the lower-level control of the present invention; Figure 5 This is a diagram showing the SOC load balancing result based on the distributed consensus algorithm of this invention; Figure 6 This is a diagram showing the frequency shift results of the proposed method in three regions. Detailed Implementation

[0026] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0027] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0028] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0029] Where there is no conflict, the embodiments and features described in this application may be combined with each other.

[0030] Example 1 A joint optimization method for multi-regional heterogeneous energy storage based on model predictive control, such as Figure 1 As shown, it includes the following steps: (1) Establish a heterogeneous energy storage system model and a grid frequency response framework; (2) Design the economic optimization objective function of the upper-level MPC; (3) Implement a dynamic functional grouping strategy based on the severity of disturbances; (4) Design the charging and discharging group partitioning of the lower-level MPC and the adaptive power allocation based on SOC; (5) Construct a hierarchical coordination mechanism and conduct simulation verification.

[0031] The heterogeneous energy storage system model is established using the standard three-region interconnected power grid frequency response model (step (1)) as follows: The heterogeneous energy storage systems considered in this invention mainly include two categories: power-type energy storage systems, such as supercapacitors and flywheel energy storage; and energy-type energy storage systems, such as battery energy storage. Power-type energy storage systems have high charge / discharge power but low energy density; in contrast, energy-type energy storage systems have high energy density but low charge / discharge power, and the former have a faster frequency response speed. Therefore, to balance response speed and response duration, heterogeneous energy storage unit groups are configured in each region. Region A's heterogeneous energy storage is allocated as battery energy storage, supercapacitor energy storage, and flywheel energy storage; Region B's is battery energy storage and supercapacitor energy storage; and Region C's is battery energy storage and flywheel energy storage. This configuration ensures balanced response capability across multiple regions and provides a physical basis for subsequent hierarchical control.

[0032] Furthermore, although the physical characteristics of the heterogeneous energy storage systems in step (1) differ significantly, their dynamic response characteristics in the frequency response model are similar. To balance model effectiveness and computational efficiency, a first-order damped element is used to describe the equivalent model of the heterogeneous energy storage system, thus characterizing its dynamic response delay. The model structures of the three energy storage systems are not significantly different; the main difference lies in the model parameters. The main model parameters of the energy storage system include dynamic response time. T Maximum capacityE max Charging and discharging power P The difference in response speed among heterogeneous energy storage systems is due to their dynamic response time. T ESU The response speed comparison of the three energy storage systems is as follows: dynamic response time of supercapacitor energy storage. T ESU,C Flywheel energy storage dynamic response time T ESU,F Battery energy storage dynamic response time T ESU,B These parameter differences reflect the millisecond-level fast response of supercapacitor energy storage systems, the medium power output of flywheel energy storage systems, and the continuous energy support capability of battery energy storage systems.

[0033] Furthermore, the heterogeneous energy storage transfer function model in step (1) can be expressed as:

[0034] In the formula s Let be the Laplace operator. Based on the above transfer function model, neglecting the constraints of the frequency adjustment unit, namely the dead zone, slope constraint, and upper and lower limit constraints, the th... i The frequency response dynamic model for a power system region is as follows:

[0035] In the formula f i and f j The first i The and the first j Frequency deviation in each region P T,i This represents the increase in the output power of the water turbine. P G,i This represents the output power increment of the speed controller. P I,i This is the integral output of the PI controller. P ESU,i,e , S ESU,i,e and E ESU,i,e The first i The first region e The charging and discharging power, SOC, and energy capacity of each energy storage system P D,i For load power disturbance, and These are the reference power increments allocated to the i-th generator and the i-th energy storage system, respectively. H i and D i The first i The inertial constant and damping coefficient of each region, R i This is the generator droop coefficient. β i This is the frequency offset coefficient. T T,i , T G,i and T ESU,i These are the response time constants of the turbine, governor, and energy storage system, respectively. ACE,i It is a region i Regional control error, N For the first i The number of adjacent regions of each region M For the first i The number of energy storage systems in each region.

[0036] Furthermore, the model built in step (1) also needs to satisfy the following inequality constraints:

[0037] In the formula and The first i The maximum and minimum ramping rates of thermal power units within the region. n To control the cycle, and The first i The maximum and minimum adjustable power of thermal power units in each region , , and The first i The first region e The maximum and minimum SOC and maximum and minimum power of each energy storage system.

[0038] Furthermore, step (2) requires defining the cost function of the upper-level controller, which serves as the objective function to generate the command signals allocated to the lower-level controller. The operating costs and regulatory fees associated with power sources within the power system are highly complex. To establish a system economic model in the frequency regulation scenario, the basic operating loss functions for thermal power units and energy storage units are first established. The static operating loss cost using the quadratic cost model can be expressed as:

[0039] In the formula P i It is the first i The output power of the power supply a i , b i and c i It is the first i The static operating cost factor of a power supply. When the output power is... P i Increase to P i + P i When incremental operating loss cost is incurred, the formula for calculating it can be expressed as:

[0040] In the formula P i ( n ( ) is the control cycle n The change in power within the range.

[0041] Furthermore, step (2) considers the degradation cost of heterogeneous energy storage systems. The degradation of energy storage systems during charging and discharging plays a significant role in their frequency regulation performance. This invention considers the unit energy throughput degradation cost of three heterogeneous energy storage systems on the same scale, and its expression is:

[0042] In the formula c d This represents the initial unit throughput cost coefficient of the energy storage system. k d Let SOH be the SOH impact coefficient, which causes the throughput cost of the aging equipment to increase as SOH decreases. In the above formula, SOH for heterogeneous energy storage is defined as:

[0043] In the formula D To control the cycle n The energy absorbed and released within. E life The total energy throughput over the overall lifespan of the energy storage device. D ( n ( ) is the control cycle n The share of damage within.

[0044] Furthermore, step (2) considers time-of-use electricity price arbitrage, which encourages energy storage systems to charge during off-peak hours and discharge during peak hours, thus enabling active participation in the energy market. First, defining the output power of the energy storage system as positive during discharge and negative during charging, the expression for considering time-of-use electricity price arbitrage is:

[0045] In the formula λ TOU To predict electricity prices, this term creates an economic incentive in the optimization objective of MPC: when the predicted electricity price is at a low point, C TOU When the value is negative, the driver controller formulates a charging plan; when the electricity price is predicted to be at its peak, C TOU When the term is positive, the controller is driven to discharge to obtain the maximum economic return. This design makes the heterogeneous energy storage system not only a tool for frequency regulation, but also an intelligent agent that can actively participate in the energy market and realize peak-valley arbitrage, thereby maximizing its comprehensive economic value.

[0046] Furthermore, a frequency offset penalty term is introduced in step (2) to prevent the upper-level controller from sacrificing the frequency support capability of the heterogeneous energy storage system when pursuing economic efficiency. This allows the upper-level MPC to focus on long-term benefits while considering the dynamic frequency constraints in the prediction time domain. By penalizing the frequency deviation, the optimization is guided towards a stable direction. This penalty term can be expressed as:

[0047] In the formula λ fre This is the frequency offset penalty coefficient.

[0048] Furthermore, in step (2), the objective function operates on a slower time scale, such as 15 minutes, with the optimization objectives of minimizing the overall system cost and maximizing profit. A rolling optimization mechanism is used to adjust the weight coefficients to achieve a balance between long-term economic planning and short-term adjustments, avoiding the limitation of traditional fixed-cost models that ignore dynamic degradation. The objective function of the rolling optimization can be expressed as:

[0049] Furthermore, in step (2), the upper layer uses MPC for optimization. This MPC framework predicts future behavior based on a state-space model and solves for the optimal control sequence within a finite time domain. Based on the dynamic model built in step (1), the first... i The state-space model of a power system region is as follows:

[0050] State variables in the formula xi Control variables u i Output variables y i and interference variables w i They can be represented as:

[0051] In the formula A i , B i , C i and F i They are the first i The state matrix, control input matrix, output matrix, and disturbance matrix of each region. A ij It is the first i and the j Interaction matrix between regions.

[0052] Furthermore, the state-space model in step (2) is based on the sampling period. T s The discrete state-space model is obtained as follows:

[0053] In the formula , , , , .

[0054] Based on the above control objectives, the optimization objective function of MPC can be transformed into a standard optimization problem with a quadratic objective function and linear constraints, the expression of which is:

[0055] In the formula Q and P These are the output weighting matrix and the control weighting matrix, respectively. N P To predict the time domain, It is a control period k The target value is dynamically tracked. At the same time, the above optimization problem also needs to satisfy the constraints in step (1).

[0056] Therefore, the optimization objective of the upper-level MPC is within a finite prediction time domain. N P Within, find an optimal control sequence. U ( k )={u ( k ), u ( k +1), ..., u( k + N P -1)}, to minimize the future N P The cumulative cost over each time step. The controller at each sampling time. k Solve this optimization problem, but only take the first element of the optimal control sequence. u ( k This is applied to the system. At the next sampling time, the controller repeats the rolling optimization process described above based on the new system state measurements.

[0057] Furthermore, after solving the upper-level MPC optimization in step (2), the classification power command that the heterogeneous energy storage systems in each region should follow in a future slow timescale is generated. Thermal power units should comply with the common power directive. and dynamic adjustment coefficient α This instruction takes into account multiple objectives such as economy, lifespan, and frequency stability. As a top-level instruction, it is issued to the lower-level controllers, specifying the reference power that various activated heterogeneous energy storage systems should share in the next scheduling cycle.

[0058] Furthermore, step (3) implements a dynamic functional grouping strategy based on disturbance severity, aiming to dynamically activate the most suitable energy storage system according to the real-time frequency disturbance characteristics to optimize response efficiency and reduce losses. As energy-type energy storage units, battery energy storage systems are more suitable for medium- and long-term energy support; supercapacitors and flywheel energy storage systems, as power-type energy storage units, emphasize millisecond-level fast response. Based on the different response characteristics of heterogeneous energy storage systems, the disturbance severity is calculated to implement the response grouping strategy. The disturbance severity function can be expressed as:

[0059] In the formula w 1, w 2 and w 3 represents the weighting coefficients determined based on historical data. RoCoF The rate of change of frequency, T dru This function estimates the duration of a disturbance. It considers both the amplitude and velocity of the disturbance, avoiding the limitations of a single metric. A threshold can be set... S low < S high and according to S Perform dynamic grouping.

[0060] like S <S low If the disturbance is small, the power storage unit is activated, utilizing its low response time and long cycle life to handle high-frequency, small-amplitude fluctuations; for intermediate values... S low < S < S high This activates the common response frequency deviation of all types of energy storage systems. If S > S high If the disturbance is identified as a major disturbance, all available energy storage units will be activated to provide large-capacity continuous support. The threshold of this strategy can be adaptively adjusted through historical data. A real-time feedback loop is formed from disturbance detection to energy storage system activation, which can improve the overall utilization rate of heterogeneous energy storage systems.

[0061] Furthermore, step (4) designs the charging and discharging group division of the lower-level MPC and the adaptive power allocation strategy based on SOC. This layer operates on a fast time scale, such as 0.1 seconds, and is responsible for accurately executing upper-level instructions and maintaining coordination between heterogeneous energy storage units. First, statistical analysis of the disturbance data shows that the positive and negative fluctuation probabilities of frequency deviation are approximately equal. Therefore, multiple units of the three heterogeneous energy storage systems can be equally divided into charging groups and discharging groups. The charging group prioritizes output power, and the discharging group prioritizes output power, avoiding frequent charging / discharging conversions to extend lifespan. Assume that there are a total of 2 energy storage units activated in step (3). K Based on the grouping strategy in the previous step, there are [number] charging groups and [number] discharging groups. K There are 1 minimum unit, and the rated power of each minimum unit is 1. P ESU Then the rated power of all energy storage systems is P ESU,N =2 KP ESU .

[0062] Furthermore, in step (4), in order to balance reducing the number of charge-discharge cycles and improving the overall available power of the heterogeneous energy storage system under the grouped mode, the following coordinated control strategy is formulated: When ≤| KP ESU When | this indicates that a single charge / discharge group can meet the reference power output, only a single group needs to be activated; when | KP ESU |< ≤|2 KP ESUThis indicates that a single charge / discharge group cannot meet the reference power output. In this case, a short-term switching strategy between charge and discharge states is used to supplement the rated power deficiency of a single group. Taking a charging group as an example, if the rated power of the charging group cannot meet the reference power output at this time, the discharging group temporarily participates in the power response output of the charging group to supplement the overall available power.

[0063] Furthermore, step (4) implements SOC-based adaptive power allocation within each group, ensuring that each group meets the reference power requirement. At the same time, it strives to maintain a balanced State of Charge (SOC) within the group and reduce lifespan degradation. For example, if the SOC of the smallest energy storage unit in the discharge group is higher than the group average, then that energy storage unit should be given a higher weight when responding to output power, so that its SOC drops to near the average SOC more quickly. This negative feedback regulation mechanism based on state deviation can actively promote the convergence of the SOC of all units in the group, thereby effectively avoiding overcharging or over-discharging of some units and extending the collaborative working life of the entire energy storage system.

[0064] The reference power output of the energy storage unit within the group is:

[0065] In the formula K This represents the number of energy storage units within the current group.

[0066] Furthermore, in step (4), to achieve this adaptive allocation, a weighted discrete distributed consensus algorithm based on MPC optimization is introduced into the lower-level controller. This algorithm enables each energy storage unit to act as an intelligent agent, interacting only with neighboring units in the communication network to collaboratively distribute the total power command issued by the lower-level MPC controller without a central coordinator. The power allocation is optimally decomposed according to the respective State of Charge (SOC), ultimately ensuring that the power output of each unit precisely matches its State of Health (SOH) and State of Charge (SOC). The specific power allocation task is completely delegated to the local controllers of the three heterogeneous energy storage units. Each local controller executes a distributed consensus algorithm within the group, autonomously calculating the power share that each smallest unit should bear, ensuring that the SOC of the entire group tends towards equilibrium. The closed-loop control system of this algorithm is as follows:

[0067] In the formula x e ( k ) is the first e The energy storage unit in the first k The next iteration is the power allocation state. P εwThis is the weighted consensus matrix, the core of the standard consensus algorithm, which drives the state of each unit to converge towards the weighted average of the states of its neighbors. Its expression is:

[0068] In the formula I It is the identity matrix. ε For discrete iteration step size, L w For weighted Laplace matrix, A for N*N A symmetric matrix whose elements a ij A value of 1 indicates a region. i and j If a communication connection exists, the value is 0; otherwise, it defines how information flows between units. D This is a degree matrix, where the diagonal elements represent the number of neighboring cells connected to a given cell. W This is a weighted matrix, where the diagonal elements represent the current charging and discharging capabilities of the energy storage units. K mpc The state feedback control gain matrix is ​​obtained by solving the MPC optimization problem.

[0069] The MPC problem aims to minimize the deviation of the states of each unit in the future prediction time domain, thereby calculating an optimal virtual control input, such that... x ( k The convergence process of this algorithm is not only stable, but also has the optimal path and the fastest speed. The introduction of this term significantly enhances the robustness and dynamic performance of the algorithm in the face of system disturbances and communication delays.

[0070] In each fast control cycle, the lower-level controller initiates the above iterative process. After a finite number of iterations, the state vector... x ( k It will quickly converge to a steady-state value where all elements are identical. x* x* represents the agreed-upon, SOC-weighted power allocation benchmark per person across the entire group. Subsequently, each energy storage unit... e Based on the convergence value and its own weight, the final reference power command to be executed is calculated.

[0071] Furthermore, the group operation strategy defined in step (4) includes a core inter-group state transition mechanism to ensure that the energy storage system can provide bidirectional power regulation capability for a long time and without interruption, and to protect the energy storage unit from overcharging or over-discharging.

[0072] Specifically, when the average SOC of a charging or discharging group reaches the upper limit of charging or the lower limit of discharging, the controller determines that it has completed the charging or discharging task and immediately executes a switching operation between the charging and discharging groups. At this time, the charging group with a high SOC naturally becomes a discharging group, which can provide effective discharge power support. Conversely, the discharging group with a low SOC naturally becomes a charging group, which can effectively absorb power. This switching strategy ensures that the energy storage system can provide reliable frequency response capabilities.

[0073] Furthermore, in step (5), a hierarchical coordination mechanism is constructed. This mechanism bridges the upper-level economic optimization and the lower-level real-time control, ensuring the consistency of the two-level objectives. The upper-level MPC operates on a slower time scale, performing joint scheduling of economic and lifetime considerations. Its output is the classified power command and preliminary dynamic adjustment coefficients that the lower-level MPC controller should follow in the future. The lower-level MPC operates on a faster time scale. After receiving the power command signal from the upper level, it does not execute it directly. Instead, based on the severity of real-time frequency disturbances (step 3) and the real-time SOC status of the energy storage unit (step 4), it dynamically and in real-time redistributes the total command, ensuring that while meeting the upper-level economic objectives, it can achieve fast and accurate frequency response and intra-group SOC balance. Subsequently, the lower-level controller processes and uploads the actual SOC and SOH status information of the energy storage unit. This information is used to update the state-space model and cost function of the upper-level MPC, thereby enabling the upper-level decision-making to adaptively adjust to the actual state of the lower level. By employing a closed-loop path of 'upper-level optimization instruction issuance - lower-level real-time tracking execution - lower-level state information feedback', the system achieves decoupling and coordination between slow-time-domain economic planning and fast-time-domain instantaneous control, ensuring the robustness and economy of overall operation. Simulation verification uses MATLAB / Simulink, injecting disturbances; this step ensures the overall closed-loop stability of the system.

[0074] As a typical embodiment, the control framework in steps (2) to (4) adopts a hierarchical design, with the upper and lower level controllers operating on different time scales. The upper level controller operates on a slower time scale, aiming to maximize revenue and minimize costs. It considers the degradation costs of heterogeneous energy storage and time-of-use arbitrage to delay the degradation problem of heterogeneous energy storage systems and actively participate in the energy market to achieve peak-valley arbitrage. Finally, it uses MPC optimization to generate reference instructions that the lower level should follow. The lower level controller operates on a faster time scale, aiming for optimal frequency regulation. It considers response strategies based on the severity of disturbances and charge / discharge grouping strategies for heterogeneous energy storage systems to reduce the number of charge / discharge cycles of heterogeneous energy storage systems and thus extend their lifespan. At the same time, it adopts a distributed consensus-based intra-group SOC recovery strategy to unify the intra-group SOC and finally generates power instructions for each energy storage unit to provide power support for frequency regulation while taking into account economic efficiency.

[0075] like Figure 2 As shown, the heterogeneous energy storage system models constructed in step (1) all adopt the equivalent first-order inertial element, and their model structures are identical. The main difference lies in the model parameters to distinguish the response characteristics of the heterogeneous energy storage systems. The regional power grid calculates the regional control error by collecting frequency deviation and interconnection area frequency deviation, and then... Figure 1 The control strategy in the system rationally allocates the reference power signal to thermal power units and hybrid energy storage units to work together to mitigate frequency deviations in the regional power grid.

[0076] like Figure 3 As shown, the hierarchical control process in steps (2) to (4) constitutes a closed-loop system. The upper-level controller transmits the power command and dynamic adjustment coefficient to the lower-level controller, and the power command output by the lower-level controller then affects the SOC and SOH of the energy storage unit. The changes in these state variables will eventually be fed back to the upper and lower-level controllers and affect their subsequent decisions, thus forming a complete closed-loop regulation process.

[0077] like Figure 4 As shown, the charging group partitioning strategy and distributed consensus algorithm flow in step (4) are as follows: The local controller first determines whether a single charging / discharging group can meet the corresponding requirements based on the reference power command signal issued by the lower-level MPC controller. If it cannot, a short-term switching strategy is adopted to temporarily make the charging / discharging groups respond to the reference power command signal simultaneously. Subsequently, according to the distributed consensus algorithm, the local controller automatically allocates power commands to each energy storage unit to unify the SOC of each energy storage unit in the group. After completing a control command, the average SOC of each charging / discharging group is judged. If the average SOC of any group reaches the charging / discharging boundary condition, the charging / discharging group swapping operation is immediately executed to ensure that the energy storage system can provide reliable response capability.

[0078] like Figure 5 As shown, in step (4), a distributed consensus algorithm is used to achieve SOC balancing of the battery energy storage charging groups in region A through local communication and autonomous coordination. Even though the SOC difference within the group is large at the start of frequency regulation, after about 300 seconds of adaptive adjustment, the SOC of unit 3 is basically consistent with the SOC of the other two groups. At the same time, as a charging group, the SOC of the three charging units maintains an overall upward trend, indicating that the charging group is executing the charging power command from the local controller and recovering the frequency deviation by absorbing power.

[0079] like Figure 6 As shown, the simulation in step (5) verifies the given three regions with a step disturbance of 200 seconds. The final frequency shift is less than 0.15Hz, and the disturbance amplitude in the first 300 seconds is greater than that in the subsequent time. This is mainly because the imbalance of SOC within the group in the early stage leads to a decrease in the frequency modulation effect.

[0080] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of one or more computer-usable storage media (including, but not limited to, disk storage, etc.) containing computer-usable program code. CD - ROM It takes the form of a computer program product implemented on (such as optical memory, etc.).

[0081] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0082] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0083] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0084] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made by those skilled in the art without creative effort within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A joint optimization method for multi-regional heterogeneous energy storage based on model predictive control, characterized in that, Includes the following steps: Based on the physical parameters of the heterogeneous energy storage system, construct a heterogeneous energy storage system model and a power grid frequency response model; Based on the constructed model, a joint optimization model based on hierarchical distributed model predictive control is constructed. The upper layer of the joint optimization model considers the degradation cost of heterogeneous energy storage and time-of-use electricity price arbitrage, with the goal of maximizing revenue and minimizing cost, in order to delay the degradation problem of heterogeneous energy storage system and actively participate in the energy market to achieve peak-valley arbitrage. The lower layer aims for optimal frequency regulation performance, considering response strategies based on the severity of disturbances and charging / discharging grouping strategies for heterogeneous energy storage systems to reduce the number of charging / discharging cycles of heterogeneous energy storage systems and thus extend their lifespan. At the same time, a distributed consensus-based intra-group SOC recovery strategy is adopted to unify the intra-group SOC, and finally, power commands for each energy storage unit are generated to provide power support for frequency regulation while taking into account economic efficiency.

2. The multi-regional heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that, The process of constructing a heterogeneous energy storage system model includes: the heterogeneous energy storage system includes power-type energy storage system and energy-type energy storage system, each region is configured with a heterogeneous energy storage unit group, and a first-order damping element is used to describe the equivalent model of the heterogeneous energy storage system in order to characterize its dynamic response delay. The process of constructing the power grid frequency response model includes: based on the heterogeneous energy storage transfer function model, ignoring the constraints of the frequency regulation unit, namely dead zone, slope constraint and upper and lower limit constraint, to represent the dynamic frequency response model of each power system region.

3. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that, The timescale of the upper layer is slower than that of the lower layer, and both the upper and lower layers use model predictive control.

4. The multi-regional heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that the upper layer... Both the upper and lower layers utilize independent controllers, with the upper controller transmitting power commands and dynamic adjustment coefficients to the lower controller. The power commands output by the lower controller then affect the SOC and SOH of the energy storage unit. Changes in various state variables are ultimately fed back to the upper and lower controllers, influencing their subsequent decisions and forming a closed-loop regulation.

5. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 4, characterized in that, The cost function of the upper-level controller serves as the objective function to generate the command signals allocated to the lower-level controller. The process of establishing the cost function includes: establishing the basic operating loss function of thermal power units and energy storage units, using a quadratic cost model to represent the static operating loss cost, constructing the calculation expression for the incremental operating loss cost, and the cost function also considers the degradation cost of heterogeneous energy storage systems, the unit energy throughput degradation cost of heterogeneous energy storage systems on the same scale, and time-of-use electricity price arbitrage. Time-of-use pricing arbitrage is used to encourage energy storage systems to charge during off-peak hours and discharge during peak hours, enabling active participation in the energy market. If we define the output power of an energy storage system as positive during discharge and negative during charging, then the expression for time-of-use pricing arbitrage is as follows: In the formula λ TOU To predict electricity prices, economic incentives are created: when predicted electricity prices are at their lowest, C TOU When the value is negative, the driver controller formulates a charging plan; when the electricity price is predicted to be at its peak, C TOU When the value is positive, the controller is driven to discharge to obtain the maximum economic return.

6. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 5, characterized in that, A frequency offset penalty term is introduced into the upper-level controller to prevent it from sacrificing the frequency support capability of the heterogeneous energy storage system in pursuit of economic efficiency. By penalizing the frequency deviation, the optimization is guided towards stability. This penalty term is expressed as: In the formula λ fre This is the frequency offset penalty coefficient.

7. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 6, characterized in that, After optimization, the upper-level controller generates the classification power command that the heterogeneous energy storage systems in each region should follow, the common power command that the thermal power units should follow, and the dynamic adjustment coefficient within a set time scale in the future. These are then issued as top-level commands to the lower-level controller, clarifying the reference power that various activated heterogeneous energy storage systems should share in the next scheduling cycle.

8. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that, The process of considering response strategies based on the severity of disturbances includes: The most suitable energy storage system is dynamically activated based on real-time frequency disturbance characteristics to optimize response efficiency and reduce losses. The disturbance severity is calculated based on the different response characteristics of heterogeneous energy storage systems to implement a response grouping strategy. The disturbance severity function is expressed as: In the formula w 1, w 2 and w 3 represents the weighting coefficients based on historical data, and RoCoF represents the rate of change of frequency. T dru To estimate the duration of the disturbance, this function considers both the amplitude and velocity of the disturbance, avoiding the limitations of a single metric by setting a threshold. S low < S high And dynamically group according to S; If S < S low If the disturbance is determined to be minor, the power-type energy storage unit is activated. For the median value S low <S< S high This activates all types of energy storage systems to respond to the frequency deviation. If S> S high If the disturbance is deemed a major disturbance, all available energy storage units will be activated to provide large-capacity continuous support. The threshold is adaptively adjusted based on historical data, forming a real-time feedback loop from disturbance detection to energy storage system activation.

9. The multi-regional heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that, The charging and discharging grouping strategy of the heterogeneous energy storage system is to equally divide the multiple units of the heterogeneous energy storage system into charging groups and discharging groups. The charging group is given priority in responding to the output power, and the discharging group is given priority in responding to the output power. To balance reducing the number of charge / discharge cycles and increasing the overall available power of heterogeneous energy storage systems in grouped operation mode, the following coordinated control strategy is formulated: when ≤| KP ESU When | it indicates that a single charge / discharge group can meet the reference power output, only a single group needs to be activated; When | KP ESU |< ≤|2 KP ESU When | indicates that a single charge / discharge group cannot meet the reference power output, a short-term switching strategy for charge / discharge states is used to supplement the rated power deficiency of a single group. KP ESU | To set a threshold.

10. The multi-region heterogeneous energy storage joint optimization method based on model predictive control as described in claim 1, characterized in that, The process of unifying the SOC within a group using a distributed consensus-based intra-group SOC recovery strategy includes: implementing SOC-based adaptive power allocation within each group, ensuring that each group meets the reference power requirement. At the same time, we try to maintain the balance of SOC within the group and reduce lifetime consumption. We actively promote the SOC of all units in the group to be consistent by using a negative feedback adjustment mechanism based on state deviation. During adaptive allocation, each energy storage unit acts as an intelligent agent, interacting only with neighboring units in the communication network to collaboratively decompose the total power command issued by the lower-level controller according to its own SOC, ultimately ensuring that the power output of each unit is precisely matched with its SOH and SOC.

Images