Energy coordination control method and system for energy storage tunnel

Abstract

The application provides an energy coordination control method and system for an energy storage tunnel, and the method comprises the following steps: acquiring spatial layout information, electrical connection mode and thermal-physical parameters of each energy storage unit of the energy storage tunnel; constructing a structure diagram of thermal-electric-space coupling of the energy storage tunnel based on each energy storage unit; collecting real-time operation data of each energy storage unit, combining the structure diagram to perform state propagation and trend prediction, and generating structure-aware state embedding and a global trend vector; generating a preliminary power scheduling strategy according to the state embedding and the global trend vector; and performing constraint fusion on the preliminary power scheduling strategy to generate a scheduling instruction. The application significantly improves the operation safety, response speed and scheduling accuracy of the energy storage tunnel under complex and variable working conditions.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of energy storage tunnels, and particularly relates to an energy coordination control method and system for energy storage tunnels. BACKGROUND

[0002] As a new type of energy infrastructure form, energy storage tunnels have been widely used in recent years in the development of urban underground space, grid-connected regulation of new energy, and multi-station coordinated energy management. Energy storage tunnels are usually arranged in underground or specific geological structures to accommodate a large number of energy storage devices (such as lithium battery modules, compressed air energy storage units, and flow batteries) in a linear or segmented form, realizing large-capacity and high-density energy deployment. Such tunnels not only undertake energy storage and release, but also involve the coordinated operation of multiple systems such as temperature control, ventilation, signal communication, and emergency protection. With the expansion of scale and diversification of equipment types, the existing technology has exposed obvious bottlenecks in operation: lack of unified modeling of multi-physical coupling relationships such as heat diffusion, electrical connection, and spatial distribution in energy storage tunnels, leading to scheduling strategies that cannot truly reflect physical structure characteristics, and easily causing risks such as local overheating and unbalanced power distribution; state perception is mostly dependent on single-point measurement, without considering the thermal, electrical, and spatial influences between nodes in the calculation, resulting in lagging or one-sided state judgment; static rules or single-objective optimization are generally used in strategy generation, lacking a dynamic allocation mechanism combined with trend prediction, making it difficult to respond to dynamic working conditions such as load fluctuations and heat risk accumulation in a timely manner; in the final execution link, existing methods rarely consider the fusion of preliminary strategies with device physical execution constraints, power resolution, and climbing ability, and direct issuance of strategies can easily cause execution deviations or even device protection actions, affecting the stable operation of the system.

[0003] Therefore, there is an urgent need for a method that can start from the multi-physical structure of an energy storage tunnel, combine real-time state propagation and trend prediction, generate a multi-step scheduling scheme that meets the physical coupling constraints, and convert it into directly executable control instructions, thereby improving the safety, response speed, and scheduling accuracy of energy storage tunnels under complex working conditions. SUMMARY

[0004] The purpose of the present application is to design an energy coordination control method and system for energy storage tunnels to solve the above technical problems.

[0005] To achieve the above purpose, the present application provides an energy coordination control method for an energy storage tunnel, which comprises:

[0006] obtaining the spatial layout information, electrical connection mode, and thermal-physical parameters of each energy storage unit in the energy storage tunnel;

[0007] constructing a thermal-electric-spatial coupling structure diagram of the energy storage tunnel based on each energy storage unit;

[0008] Collect real-time operation data of each energy storage unit, combine the structure diagram to perform state propagation and trend prediction, generate structure-aware state embedding and global trend vector;

[0009] According to the state embedding and global trend vector, combine the physical boundary parameters of the structure diagram, the spatial layout information, the electrical connection mode, and the edge weight matrix of the thermal-physical parameters to construct coupling constraints, and generate a preliminary power scheduling strategy;

[0010] Perform constraint fusion on the preliminary power scheduling strategy to generate a scheduling instruction.

[0011] Further, the spatial layout information is the center point coordinates of each energy storage unit; the electrical connection mode is the series / parallel relationship, cable routing and cross-section impedance of each energy storage unit; and the thermal-physical parameters are the steady-state and transient temperature rise curves under different operating conditions, to inversely deduce the thermal resistance parameters between modules.

[0012] Further, the construction process of the thermal-electric-space coupled structure diagram includes:

[0013] Obtain all energy storage units, and abstract each energy storage unit of the energy storage tunnel as a node; wherein the attributes of each node include corresponding spatial layout information, electrical connection mode, and thermal-physical parameters;

[0014] Based on the corresponding spatial layout information, electrical connection mode, and thermal-physical parameters, construct three types of thermal diffusion edges to obtain the edge weights of the spatial layout information, electrical connection mode, and thermal-physical parameters, i.e. the corresponding spatial, electrical, and thermal weights, which fall between

[0015] Combine all nodes and spatial, electrical, and thermal weights to generate a thermal-electric-space coupled structure diagram.

[0016] Further, the process of performing state propagation and trend prediction includes:

[0017] Obtain the structure diagram and real-time operation data of each energy storage unit, wherein the real-time operation data includes SOC, voltage, current, and temperature of each energy storage unit;

[0018] Based on the real-time operation data, fuse neighbor information based on the weighted adjacency matrix of the structure diagram, thereby performing state propagation to generate an updated state vector of the node fused with neighbor information, which is directly used as structure-aware state embedding;

[0019] Input the state embedding sequence of the last sampling period into a pre-constructed prediction model to generate a multi-step prediction result of all nodes;

[0020] ​The multi-step prediction matrix corresponding to the multi-step prediction results of all nodes is processed by a dimension reduction algorithm to generate a global trend vector.

[0021] Further, the loss function of the pre-constructed prediction model is a multi-step prediction loss function with physical constraints.

[0022] The physical constraints are used to force nodes with strong physical coupling to have consistent future state change trends. The strength of the physical coupling depends on the size of the sum of the corresponding spatial, electrical and thermal weights.

[0023] Further, the process of constructing a coupling constraint according to the state embedding and the global trend vector, combining the physical boundary parameters of the structure diagram, the spatial layout information, the electrical connection mode and the edge weight matrix of the thermal physical parameters to generate a preliminary power scheduling strategy includes:

[0024] The state embedding and the global trend vector are spliced to form an enhanced state vector.

[0025] The power upper and lower limits of each node are limited according to the physical boundary parameters of the structure diagram, and the coupling constraint is constructed using the weight matrix of the spatial, electrical and thermal weights.

[0026] Based on the coupling constraint, the state embedding and the global trend vector are combined to construct an optimization objective function, so that the global power matches the predicted load, and a preliminary power scheduling strategy is output, which is used for the node to output the preliminary charging and discharging power instruction in the future period.

[0027] Further, the optimization objective function is represented as:

[0028] ;

[0029] wherein, is a trend deviation cost, which is composed of the square sum of the difference between the predicted total load and the sum of the allocated power, to ensure that the global power follows the load prediction in the global trend vector ; is the current preliminary charging and discharging power instruction; is the current global cost; , is the comprehensive physical coupling coefficient of the node and , , and is the fusion coefficient of the spatial layout information, the electrical connection mode and the thermal physical parameter relationship, , and is the spatial, electrical and thermal weight; is the node a thermal risk coefficient of the node i; a safety temperature rise threshold of the node i; a future temperature rise rate in trend prediction; a safety temperature rise threshold of the node i; a physical constraint penalty coefficient for balancing the priority of trend following and physical safety; a preliminary charge-discharge power instruction of the node i, a preliminary charge-discharge power instruction of the node j; an edge set; a node set;

[0030] wherein, a thermal delay inhibition term for ensuring that the power distribution between node pairs with strong physical coupling tends to be consistent, and structurally inhibiting the thermal and electrical impact risk of high-coupling areas;

[0031] a spatial interference smoothing term for performing second-order difference punishment on the power change of spatially adjacent nodes, and reducing the possibility of spatially concentrated overheating.

[0032] Further, the process of performing constraint fusion on the preliminary power scheduling strategy to generate scheduling instructions comprises:

[0033] First, the preliminary power scheduling strategy is subjected to ramping and envelope constraints to obtain a final executable sequence;

[0034] Subsequently, the final executable sequence is subjected to resolution quantization by rounding to the nearest whole step to generate final scheduling instructions.

[0035] Further, the process of subjecting the preliminary power scheduling strategy to ramping and envelope constraints to obtain a final executable sequence comprises:

[0036] The power of the executed power scheduling strategy in the previous cycle is collected as a constraint;

[0037] Based on the constraint, an intermediate sequence is generated in the preliminary power scheduling strategy by using a time-by-time truncation method;

[0038] According to the intermediate sequence, ramping is applied, the upper limit of the ramping uses a self-adaptive quantization value based on fluctuation for different nodes, and then power resolution quantization is performed to obtain a final executable sequence.

[0039] In a second aspect of the present application, an energy coordination control system for an energy storage tunnel is provided, and the system comprises:

[0040] A data acquisition module is configured to acquire spatial layout information, electrical connection mode and thermal physical parameters of each energy storage unit of the energy storage tunnel.​

[0041] a structure modeling module configured to construct a structure graph of thermal-electric-space coupling of the energy storage tunnel based on each of the energy storage units;

[0042] a state prediction module configured to collect real-time operation data of each of the energy storage units, perform state propagation and trend prediction in combination with the structure graph, and generate a structure-aware state embedding and a global trend vector;

[0043] a strategy generation module configured to construct coupling constraints according to the state embedding and the global trend vector in combination with physical boundary parameters of the structure graph, the spatial layout information, the electrical connection mode, and edge weight matrices of thermal-physical parameters, and generate a preliminary power scheduling strategy;

[0044] an execution scheduling module configured to perform constraint fusion on the preliminary power scheduling strategy, and generate a scheduling instruction.

[0045] The present application has at least the following beneficial technical effects:

[0046] The present application proposes an energy coordination control method and system covering the whole process, aiming at the multi-physical coupling characteristics of the energy storage tunnel and the deficiencies of existing scheduling control in structure modeling, state prediction, strategy generation, and execution implementation. The method first constructs a coupling structure graph reflecting the thermal, electrical, and spatial relationships of the energy storage units, unifies the fixed physical properties of the nodes and the normalized weights of the three types of relationships, and provides structure constraints for subsequent calculations. On this basis, in combination with real-time operation data, the structure-aware state propagation mechanism is used to fuse the neighborhood influence, and the trend prediction model is used to output the system state evolution results for multiple future periods, forming the structure-aware node state embedding and the global trend vector. Then, the trend information and the structure constraints are used to jointly optimize the multi-step preliminary power scheduling strategy, and the thermal delay suppression term and the spatial mutual interference smoothing term are introduced for the energy storage tunnel scenario, so that the strategy can meet the predicted load demand while actively avoiding thermal risks and coupling impacts. Finally, the preliminary strategy is fused with the device execution layer parameters, and based on the ramping limit, power resolution, and station-level consistency requirements, the execution instruction that can be directly issued to each energy storage unit is generated.

[0047] The four links form an organic closed loop, from physical structure to state modeling to strategy generation and execution output, all based on the actual operation characteristics of the energy storage tunnel, and cooperate with each other to realize structure-aware, trend-driven, physically constrained, and engineering-landable energy coordination control, significantly improving the operation safety, response speed, and scheduling accuracy of the energy storage tunnel under complex and variable working conditions. BRIEF DESCRIPTION OF DRAWINGS

[0048] The application is further illustrated by the accompanying drawings, but the embodiments in the drawings do not constitute any limitation to the application, and other embodiments can be obtained by those skilled in the art without creative labor on the basis of the following drawings.

[0049] Figure 1 Flow chart of the energy coordination control method for the energy storage tunnel of the application.

[0050] Figure 2 Frame chart of the energy coordination control system for the energy storage tunnel of the application. DETAILED DESCRIPTION

[0051] The embodiments of the application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, in which the same or similar reference signs represent the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the drawings are exemplary and are only used to explain the application, and cannot be understood as a limitation to the application.

[0052] In one or more embodiments, as shown in Figure 1 The energy coordination control method for the energy storage tunnel is disclosed, and the method comprises the following steps:

[0053] S1, obtaining the spatial layout information, electrical connection mode and thermal-physical parameters of each energy storage unit of the energy storage tunnel.

[0054] Specifically, the data of this step mainly comes from three parts: first, the spatial layout information of the energy storage cabin, which comes from the construction drawings and the on-site three-dimensional laser scanning measurement, the scanning resolution can reach millimeter level, and the three-dimensional coordinate file (such as.pts or.las format) is output, and the center point coordinates of each energy storage unit are used to calculate the spatial adjacency weight; second, the electrical connection mode, which comes from the system electrical design document and the on-site wiring acceptance record, including the series / parallel relationship of each energy storage unit, the cable direction and the cross-section impedance, which can be verified by the electrical completion drawing and the test instrument (such as the precision resistance measuring instrument); third, the thermal-physical parameters, which come from the module specification table (thermal capacity, thermal conductivity, etc.) provided by the energy storage unit manufacturer and the on-site thermal imager measurement data, and the steady-state and transient temperature rise curves under different operating conditions are selected for measurement, so as to back-propagate the thermal resistance parameters between the modules. In addition, the fixed operating boundary parameters (maximum charge and discharge power, maximum allowable temperature rise) of the equipment also need to be collected, and these data can be directly derived from the static configuration file of the BMS and PCS.

[0055] S2, constructing a thermal-electric-space coupling structure chart of the energy storage tunnel based on each energy storage unit.

[0056] Specifically, this step aims to construct a thermal-electric-space coupling structure chart of the energy storage tunnel , as the constraint carrier of all subsequent state propagation and control policy generation. Since this is the first step, there is no pre-step input, so the raw data is directly obtained from the engineering deployment and operation management system for processing. The structure diagram not only describes the physical connection between the energy storage units, but also contains the fixed physical properties of each node and the numerical weights of the three types of relationships, which will directly participate in the propagation of node influence in subsequent calculations. In order to ensure that different physical quantities are comparable in the same model, the weights will be normalized when constructing.

[0057] Further, after the data collection is completed, each energy storage unit in the energy storage tunnel is abstracted as a node , the attributes of the node include the position coordinates , heat capacity , electrical interface position and direction, maximum power , etc. In order to construct the heat diffusion edge , it is necessary to first calculate the thermal resistance between nodes according to the physical contact surface and distance, for example, if two cabins are connected through the same metal bracket, then the thermal conductivity and cross-sectional area of the bracket material are measured, and

[0058] ;

[0059] where is the center distance between the cabins. The weight of the electrical connection edge can be obtained from the length of the connecting cable and the impedance per unit length

[0060] ;

[0061] For example, for a copper cable with a length of m and an impedance of mΩ / m, mΩ = 2 mΩ. The spatial adjacency edge is directly calculated using the three-dimensional distance .

[0062] After combining the three types of edges and their weights into a multi-layer weighted adjacency matrix, the coupled structure diagram can be represented as

[0063] ;

[0064] where, , , ​are the normalized thermal, electrical, and spatial weights, respectively. The normalization is done by dividing each weight value by the maximum value of the same type of weight, so that all weight values fall between 0 and 1, ensuring that different physical relationships can be directly compared and combined in subsequent calculations. For example, if the maximum thermal resistance measured in the tunnel is 10 K / W, then a pair of nodes with a thermal resistance of 5 K / W has a normalized weight of 0.5. The benefit of this treatment is that the three types of relationships can participate in neighborhood influence propagation in the same algorithm with a unified scale, without the need for additional unit conversion.

[0065] During modeling, special attention should be paid to data consistency. For example, if the module coordinates provided by the BMS deviate from the construction drawings, the coordinate information needs to be corrected through on-site scanning results; during thermal resistance measurement, it is necessary to ensure that the modules are compared under the same operating power to eliminate the influence of load fluctuations on thermal parameter estimation.

[0066] The output of this step is the thermal-electric-spatial coupling graph , where each node is attached with fixed physical attributes, and each type of edge stores the corresponding normalized weight matrix. This graph not only contains topological information, but also contains physical constraint information for subsequent state propagation modeling and control strategy generation. In the next step, this graph will be used as input to participate in state propagation calculation, so that the state of each node automatically includes thermal, electrical, and spatial influences from its physical neighbors.

[0067] S3, collect real-time operating data of each energy storage unit, combine the structure graph to perform state propagation and trend prediction, and generate structure-aware state embedding and global trend vector.

[0068] Specifically, the goal of this process is to obtain an expression that not only contains the current global state distribution, but also reflects the future short-term evolution trend, providing dynamic information under physical constraints for subsequent control strategy generation. Due to the strong coupling characteristics of thermal diffusion, current distribution, and spatial proximity in the energy storage tunnel, this step not only performs neighborhood fusion of node states at a single time, but also introduces coupling information into the time modeling process, ensuring that the prediction results are consistent with the physical structure.

[0069] The input part consists of two types of data: one is the fixed attributes of the nodes bound in the structure graph, such as the position coordinates , thermal capacity , maximum power , and the normalized weight matrix of the three types of edges , ; the other is the real-time state vector , where,​​​​ SOCt-1, Tt-1, Vt-1, It-1,

[0070] Further, in the state propagation phase, the state vector of each node not only retains its real-time value, but also fuses neighbor information through a weighted adjacency matrix. To fit the physical characteristics of the energy storage tunnel scenario, we introduce a "thermal-electric- spatial joint attenuation factor" when aggregating the neighborhood. This not only considers the normalized weights of different physical relationships, but also introduces a time decay term to suppress the influence of outdated information:

[0071] ;

[0072] where, is the neighbor set of node (determined by 's , , three types of edges); , , are the normalized weights of the three types of physical relationships (output from step S2); is the time difference between the latest data update between node and (obtained from the acquisition system timestamp); is the overall weight coefficient of neighborhood information; , , are the fusion coefficients of the three types of physical relationships; and are the real-time state vectors of the node (collected by BMS, PCS, including SOC, voltage, current, temperature, etc.).

[0073] In the trend modeling phase, the extended state sequence of the last sampling periods is input into the time modeling module. Considering the existence of local overheating risk and load mutation in the operation of the energy storage tunnel, this patent introduces a prediction model optimization method with physical constraint regularization term. A structural constraint term is added to the multi-step prediction loss function, forcing the change rate of the predicted state to be consistent with the neighborhood smoothness of :

[0074] ;

[0075] in, The error between the predicted value and the actual measured value (such as mean square error, calculated based on measured data from BMS and PCS). For nodes In the future Predicted state for each sampling period; Weights for regularization terms; for The set of edges. This constraint term makes physically coupled nodes tend to change in a consistent manner in the prediction, thereby suppressing physically unreasonable sharp changes, which is particularly critical for the thermal safety and current balance of adjacent modules in the tunnel.

[0076] The multi-step prediction results of all nodes form a "high-dimensional prediction matrix" (each row corresponds to the future τ-step state of a node). By using dimensionality reduction algorithms (such as PCA, principal component analysis) to reduce the dimensionality of this matrix, the "principal components" that best represent the global trend are extracted, which are the system trend vectors.

[0077] Through the above two steps of calculation, two outputs can be obtained: one is the structure-aware node state embedding. The first is the updated state vector of the node after incorporating neighbor information; it combines the real-time state at the current moment with the fused neighborhood information. The second is the global trend vector. It is obtained by dimensionality reduction (e.g., PCA) of the multi-step prediction matrix of all nodes, and is used for global trend judgment in subsequent policy generation.

[0078] S4. Based on the state embedding and global trend vector, and combined with the physical boundary parameters of the structure graph, the spatial layout information, the electrical connection method, and the edge weight matrix of the thermophysical parameters, a coupling constraint is constructed to generate a preliminary power scheduling strategy.

[0079] Specifically, during execution, the node state is first embedded. With global trend vector The trend components of the corresponding nodes are concatenated to form an enhanced state vector. The vector contains:

[0080] The node's current structure awareness state (output from step two) Provides information including SOC, temperature, voltage, current, and neighborhood fusion information.

[0081] Node future state trend (by Extraction includes future temperature rise rate prediction, SOC change trend, power demand direction, etc.

[0082] Next, according to Physical boundary parameters in (e.g.) , thermal capacity ) define the upper and lower power bounds of each node, and construct coupling constraints with three types of edge weight matrices. For example:

[0083] If is large, it means that the two nodes are strongly coupled thermally, and they should not be discharged at the same time in the same cycle; If is large, the difference in power change rate between the two nodes should be avoided to prevent current shock;

[0084] If is large, the spatially adjacent nodes should not be subjected to peak thermal load at the same time.

[0085] Further, in order to combine these requirements with the trend-driven power target, this step introduces two innovative constraint terms in the optimization objective:

[0086] (1) Thermal delay suppression term: based on the future temperature rise rate obtained in step two, additional power reduction priority is allocated to high-risk nodes; (2) Spatial interference smoothing term: the second-order difference penalty is applied to the power change of spatially adjacent nodes to reduce the possibility of spatially concentrated overheating.

[0087] Further, the optimization objective function is designed as follows:

[0088] ;

[0089] Wherein:

[0090]

[0091] : trend deviation cost, composed of the square sum of the difference between the predicted total load and the sum of the allocated power, to ensure that the global power follows the load prediction in ;

[0092] : the comprehensive physical coupling coefficient of node and (combining the weight in step one and the fusion coefficient in step two);

[0093] : the thermal risk coefficient of node , calculated from the temperature rise trend prediction result in step two, for example, the higher the future temperature rise rate, , the larger the;

[0094] : the future temperature rise rate of node in trend prediction;

[0095] ​ : safety temperature rise threshold;

[0096] , : physical constraint penalty coefficient, used to balance the priority between trend following and physical safety.

[0097] The first constraint term guarantees the power allocation between nodes with strong physical coupling tends to be consistent, structurally suppressing the risk of thermal and electrical impact in high-coupling areas. The second constraint term directly introduces the temperature rise trend of step S3 into the scheduling optimization, realizing trend-driven active thermal safety control, which is particularly critical in long strip closed spaces such as energy storage tunnels, as local overheating can quickly spread along the tunnel.

[0098] Further, in the solving process, first, a baseline power vector is allocated according to the trend prediction, so that the global power matches the predicted load; then, iterative correction is performed using the above optimization objective (quadratic programming or gradient descent method with constraints can be used), gradually eliminating the violation of physical constraints, to obtain a preliminary scheduling strategy that satisfies the thermal-electric-space constraints .

[0099] The output part is the future periods of , where represents the preliminary charging and discharging power instructions of node in the future periods. These instructions will be fused with the final execution constraints in the next step to form the scheduling plan that can be directly issued to the controller.

[0100] S5, performing execution constraint fusion on the preliminary power scheduling strategy to generate scheduling instructions.

[0101] Specifically, the operation focuses on the engineering executable level: based on the power resolution of the device and the ramping ability of the transformer, the slope and quantization processing are performed, and the control frame is formed to avoid introducing new modeling or prediction links again.

[0102] Further, first, the ramp and envelope constraints are applied, and the constraints are derived from the sustainable ramping ability of the device side and the executed power of the previous period, and the intermediate sequence is generated in a time-by-time truncation manner. The upper limit of the ramping uses an adaptive quantization value based on volatility , which is calculated by the controller before entering this step using the moving median absolute deviation of and superimposed with a preset lower limit, without the need for additional external data. Then, the power resolution quantization is performed to obtain the final executable sequence , ensure the granularity of the PCS / EMS instructions. The core calculation is as follows:

[0103] ;

[0104] : node The preliminary power in the future period (part of the output of step three).

[0105] : node The power in the last executed period, from the execution cache of the last control period (locally saved by the controller).

[0106] : node The upper limit of the ramp, calculated by the controller according to the fluctuation degree and plus a safety lower limit, used to limit the single-period power change amplitude; the value is determined by the station-level configuration and the statistical quantity of , not dependent on external sensors.

[0107] : Standard clipping operation to limit the input to the interval Subsequent resolution quantization and final scheduling instruction generation:

[0108]

[0109] ;

[0110] : Power instruction resolution step, pre-configured by the PCS / EMS parameter table (fixed constant read by the controller).

[0111] : Round to the nearest integer step, ensuring consistency with the device's minimum step.

[0112] To align the "execution layer" positioning, the following examples are based only on the output of the last step and the device parameters known to the controller (the actual power executed in the last period, the ramp upper limit generation rule, the instruction resolution , station-level target window, etc.), without introducing new modeling quantities. The example uses a small-scale fragment of three nodes , two future periods ( ).

[0113] (1) Start alignment ( ): The field is in a hot start, and the controller caches the actual record of the last period: , , ​​If it is a cold start, replace the three values with the baseline power of the station start-stop process (such as "grid-connected open-loop power"), and the rest of the process remains unchanged.

[0114] (2) Uphill upper limit setting and clipping: the controller calculates the sliding median absolute deviation (provided by the local statistical library) for each node in the window of , and takes the greater value with the preset lower limit as . For example:

[0115] Node: ; the statistical fluctuation is 7, the preset lower limit is 5, and the greater value is taken as . The is

[0116] ;

[0117] Again for the second period: the previous period has not been executed, and is used as the reference according to the "forward assumption": .

[0118] Node: ; the fluctuation is 4, the preset lower limit is 6, and the greater value is taken as . There is , .

[0119] Node: ; the fluctuation is 9, the preset lower limit is 3, and the greater value is taken as . There is , . The above only uses and the upper limit of the station in the last period to generate , which is aligned with the scheme and can be directly reproduced.

[0120] (3) Quantization and jitter suppression:

[0121] The resolution of the station PCS command is set to . Direct quantization gives:

[0122] : ; .

[0123] : ; .

[0124] : ; .

[0125] If a node quantizes to "45↔50↔45" in the next cycle, the controller enables the "hold-update" threshold for this node (e.g. only update when the difference between the unquantized value and the executed value exceeds ). This can suppress the execution stage oscillation caused by small jitter. For example If a node quantizes to "45↔50↔45" in the next cycle, the controller enables the "hold-update" threshold for this node (e.g. only update when the difference between the unquantized value and the executed value exceeds ). This can suppress the execution stage oscillation caused by small jitter. For example

[0126] (4) Instruction frame assembly:

[0127] Pack each into a bus frame: {addr=i, ts=t+τ, cmd=P_exec,i(t+τ), crc=check}. The frame examples for the three nodes in are:

[0128] addr=A, ts=t+1, cmd=45, crc=…;

[0129] addr=B, ts=t+1, cmd=25, crc=…;

[0130] addr=C, ts=t+1, cmd=-5, crc=…;

[0131] Enqueue in the order of , and the lower machine polls and sends. This process only depends on the and the bus protocol description file, and is implemented transparently.

[0132] (5) Station-level consistency check and secondary quantization (only for low-sensitive nodes):

[0133] The station-level window is , and the target total power is . The sum after the above quantization is , and the deviation is . If the allowed following error in the station is set to , it is directly issued; if the target is changed to , the deviation is , and secondary quantization is needed. According to the "low-sensitive first" rule, start fine-tuning from the node with "large power margin, large slope space (P_slope ), and far from the boundary" in the current cycle. For example, at this time still has a margin from the upper limit and is the largest, first increase to ; if it is still , increase to , (two steps, each adding ); if still different, then go back up to up to . The entire secondary quantization uses only the already obtained , , and resolution .

[0134] In one or more embodiments, as shown in Figure 2 , an energy coordination control system for an energy storage tunnel is disclosed, the system comprising:

[0135] a data acquisition module 101 configured to acquire spatial layout information, electrical connection mode, and thermal-physical parameters of each energy storage unit of the energy storage tunnel;

[0136] a structure modeling module 102 configured to construct a thermal-electric-space coupled structure diagram of the energy storage tunnel based on each of the energy storage units;

[0137] a state prediction module 103 configured to acquire real-time operation data of each energy storage unit, combine the structure diagram to perform state propagation and trend prediction, and generate a structure-aware state embedding and a global trend vector;

[0138] a strategy generation module 104 configured to construct coupling constraints according to the state embedding and the global trend vector, combine physical boundary parameters of the structure diagram, edge weight matrices of the spatial layout information, the electrical connection mode, and the thermal-physical parameters, and generate a preliminary power scheduling strategy;

[0139] an execution scheduling module 105 configured to perform constraint fusion on the preliminary power scheduling strategy to generate a scheduling instruction.

[0140] It is worth noting that the specific workflow of the energy coordination control system for the energy storage tunnel provided by the embodiments of the present application is the same as the workflow of the energy coordination control method for the energy storage tunnel described in the above embodiments, and will not be repeated here.

[0141] The embodiments of the present application also provide an energy coordination control device for an energy storage tunnel, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the steps of the above-mentioned energy coordination control method for the energy storage tunnel embodiments when executing the computer program, such as the steps S1-S5 described in Figure 1 ; or, the processor implements the functions of each module in each system embodiment when executing the computer program.

[0142] For example, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present application. The one or more modules can be a series of computer program instruction segments capable of completing a specific function, which are used to describe the execution process of the computer program in the energy coordination control device for energy storage tunnel.

[0143] The energy coordination control device for energy storage tunnel can be a computing device such as a desktop computer, a notebook computer, a palm computer, and a cloud server. The energy coordination control device for energy storage tunnel can include, but is not limited to, a processor, a memory. Those skilled in the art can understand that the energy coordination control device for energy storage tunnel can also include input / output devices, network access devices, buses, etc.

[0144] The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASAC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can be any conventional processor, etc. The processor is the control center of the energy coordination control device for energy storage tunnel, and connects various parts of the energy coordination control device for energy storage tunnel through various interfaces and lines.

[0145] The memory can be used to store the computer program and / or modules, and the processor realizes various functions of the energy coordination control device for energy storage tunnel by running or executing the computer program and / or modules stored in the memory, and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required by a function, etc.; the data storage area can store data created according to the running of the air conditioner controller, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory such as a hard disk, a memory, a plug-in hard disk, a smart memory card (SMC), a secure digital (SD) card, a flash card, at least one disk storage device, a flash memory device, or other volatile solid-state storage devices.

[0146] If the module for integrating the energy coordination control device of the energy storage tunnel is implemented in the form of a software function unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium. When the processor executes the computer program, the steps of the above-mentioned various method embodiments can be implemented. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer-readable medium can include any entity or device that can carry the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc.

[0147] A person of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium. When the program is executed, it can include the processes of the above-mentioned various method embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0148] The above is the preferred embodiment of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principles of the present application, a number of improvements and refinements can be made, which are also considered within the scope of protection of the present application.

Claims

1. A method for energy coordinated control of an energy storage tunnel, characterized in that, The method comprises: acquiring spatial layout information, electrical connection mode and thermal-physical parameters of each energy storage unit of the energy storage tunnel; constructing a thermal-electric-space coupled structure diagram of the energy storage tunnel based on each energy storage unit; collecting real-time operation data of each energy storage unit, combining the structure diagram to perform state propagation and trend prediction, generating structure-aware state embedding and global trend vector; constructing coupling constraints according to the state embedding and global trend vector, combining the physical boundary parameters of the structure diagram, the edge weight matrix of the spatial layout information, the electrical connection mode and the thermal-physical parameters, and generating a preliminary power scheduling strategy; performing execution constraint fusion on the preliminary power scheduling strategy to generate a scheduling instruction; wherein the construction process of the thermal-electric-space coupled structure diagram comprises: acquiring all energy storage units and abstracting each energy storage unit of the energy storage tunnel as a node; wherein the attributes of each node include corresponding spatial layout information, electrical connection mode and thermal-physical parameters; Based on the corresponding spatial layout information, electrical connection mode and thermophysical parameters, three types of heat diffusion edges are constructed to obtain the edge weights of the spatial layout information, electrical connection mode and thermophysical parameters, i.e. the corresponding spatial, electrical and thermal weights, which fall between 0 and 1. combining all nodes and spatial, electrical and thermal weights to generate a thermal-electric-space coupled structure diagram; the process of performing state propagation and trend prediction comprises: acquiring the structure diagram and real-time operation data of each energy storage unit, the real-time operation data including SOC, voltage, current and temperature of each energy storage unit; based on the real-time operation data, performing weighted adjacency matrix fusion of neighbor information on the structure diagram to propagate the state, generate an updated state vector of the node fused with neighbor information, and directly use it as structure-aware state embedding; The recent The state embedding sequence of the last sampling period is input into a pre-constructed prediction model to generate multi-step prediction results of all nodes. generate a global trend vector by performing dimensionality reduction algorithm processing on the multi-step prediction matrix corresponding to the multi-step prediction results of all nodes; the loss function of the pre-constructed prediction model is a multi-step prediction loss function with physical constraints; wherein the physical constraints are used to force the nodes with strong physical coupling to have consistent future state change trends; wherein the strength of the physical coupling depends on the size of the sum of the corresponding spatial, electrical and thermal weights.

2. The energy-coordinated control method for an energy storage tunnel according to claim 1, wherein, The spatial layout information is the center point coordinates of each energy storage unit; the electrical connection mode is the series / parallel relationship, cable layout and cross-section impedance of each energy storage unit; and the thermal-physical parameters are the steady-state and transient temperature rise curves under different operating conditions, which are used to back-propagate the thermal resistance parameters between modules.

3. The energy-coordinated control method for an energy storage tunnel according to claim 1, wherein, The process of constructing coupling constraints according to the state embedding and global trend vector, combining the physical boundary parameters of the structure diagram, the edge weight matrix of the spatial layout information, the electrical connection mode and the thermal-physical parameters, and generating a preliminary power scheduling strategy, comprises: splicing the state embedding and global trend vector to form an enhanced state vector; limiting the power upper and lower limits of each node according to the physical boundary parameters of the structure diagram, and constructing coupling constraints using the weight matrix of the spatial, electrical and thermal weights; Based on the coupling constraint, an optimization objective function is constructed in combination with the state embedding and the global trend vector to match the global power with the predicted load, and a preliminary power scheduling strategy is output for the node to match the preliminary charging and discharging power instructions in the future period.

4. The energy-coordinated control method for an energy storage tunnel according to claim 3, wherein, the optimization objective function is represented as: ； wherein, is the trend deviation cost, composed of the square sum of the difference between the predicted total load and the sum of the allocated power, ensuring that the global power follows the global trend vector of the load prediction; is the current preliminary charge-discharge power instruction; is the current global cost; , is the comprehensive physical coupling coefficient of the node and , , and is the fusion coefficient of the spatial layout information, the electrical connection mode and the thermal-physical parameter relationship, , and is the spatial, electrical and thermal weight; is the thermal risk coefficient of the node ; is the future temperature rise rate of the node in the trend prediction; is the safe temperature rise threshold; , is the physical constraint penalty coefficient, used to balance the priority of trend following and physical safety; is the preliminary charge-discharge power instruction of the node i, is the preliminary charge-discharge power instruction of the node j; is the edge set; is the node set; wherein, is a thermal delay suppression term, used to ensure that the power distribution between pairs of nodes with strong physical coupling tends to be uniform, structurally suppressing the risk of thermal and electrical impact in the high coupling region; Spatial interference smoothing term, for second order difference penalty on power variation of spatial neighbors, to reduce the possibility of spatial hot spots.

5. The energy-coordinated control method for an energy storage tunnel according to claim 1, wherein, the process of performing execution constraint fusion on the preliminary power scheduling strategy to generate a scheduling instruction comprises: firstly, imposing ramping and envelope constraints on the preliminary power scheduling strategy to obtain a final executable sequence; The final executable sequence is then subjected to resolution quantization, generating final scheduling instructions by rounding to the nearest whole step.

6. The energy-coordinated control method for an energy storage tunnel according to claim 5, wherein, The process of imposing the ramping and envelope constraints on the preliminary power scheduling strategy to obtain the final executable sequence includes: Collecting the power of the executed power scheduling strategy in the previous period as a constraint; Based on the constraint, an intermediate sequence is generated in the preliminary power scheduling strategy using the hour-by-hour truncation method; According to the intermediate sequence, the ramping is imposed, the upper limit of the ramping uses the self-adaptive quantization value based on fluctuation for different nodes, and then the power resolution quantization is performed to obtain the final executable sequence.

7. A system for performing the energy-coordinated control method for an energy storage tunnel as claimed in claim 1, characterized in that, The system includes: A data acquisition module for acquiring spatial layout information, electrical connection mode, and thermal-physical parameters of each energy storage unit of the energy storage tunnel; A structure modeling module for constructing a thermal-electric-space coupled structure diagram of the energy storage tunnel based on each energy storage unit; A state prediction module for collecting real-time operation data of each energy storage unit, combining the structure diagram to perform state propagation and trend prediction, generating structure-aware state embedding and global trend vector; A strategy generation module for constructing coupling constraints according to the state embedding and global trend vector, combining the physical boundary parameters of the structure diagram, the edge weight matrix of the spatial layout information, the electrical connection mode, and the thermal-physical parameters to generate a preliminary power scheduling strategy; An execution scheduling module for performing execution constraint fusion on the preliminary power scheduling strategy to generate scheduling instructions.

Images