Hydrogen and air co-regulated pumped storage coupled compressed air energy storage method and system

Abstract

The application discloses a water-air synergic regulation pumped storage coupled compressed air energy storage method and system, relates to the technical field of energy storage, and comprises the following steps: after collecting the response data set of water and air double medium, renewable energy input power prediction and power grid load prediction are performed by constructing a double medium response characteristic vector, and an external demand tensor in a regulation window is constructed; the external demand tensor and the double medium response characteristic vector are fused, an adjustment intention tensor is output to a strategy generator, and a water-air joint output strategy matrix is output; joint scheduling instruction sets are configured according to the water-air joint output strategy matrix, the joint scheduling instruction sets are issued to an execution terminal, and synergic energy storage control management is performed. The application solves the technical problems of insufficient synergic regulation capability of water-air double energy storage medium and difficulty in adapting to power grid load and renewable energy fluctuation in the prior art, and achieves the technical effects of improving the response speed of the energy storage system and the overall energy utilization efficiency.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of energy storage, in particular to a pumped hydro storage coupled with compressed air energy storage method and system with water-air coordinated regulation. BACKGROUND

[0002] In the energy storage system represented by pumped hydro storage or compressed air energy storage, the potential energy of water and the pressure energy of air are usually operated and controlled as independent energy carriers, and each uses a fixed or weakly coupled scheduling mode. When the power grid load changes rapidly or the output of renewable energy such as wind power and photovoltaic power presents strong fluctuation characteristics, the differences in response speed, regulation inertia and energy conversion characteristics of different media easily lead to regulation lag or unreasonable energy allocation, making it difficult to achieve coordinated complementary output, and limiting the adaptability of the energy storage system to power grid fluctuations and the overall operation efficiency. SUMMARY

[0003] The present application provides a pumped hydro storage coupled with compressed air energy storage method and system with water-air coordinated regulation, which is used to solve the technical problems of insufficient water-air dual energy storage medium coordination regulation capability and difficulty in adapting to power grid load and renewable energy fluctuations in the prior art.

[0004] In view of the above problems, the present application provides a pumped hydro storage coupled with compressed air energy storage method and system with water-air coordinated regulation.

[0005] In a first aspect of the present application, a pumped hydro storage coupled with compressed air energy storage method with water-air coordinated regulation is provided, the method comprising:

[0006] After collecting the response data set of the water and air dual medium, a multi-modal time series encoding network is used to construct a dual medium response characteristic vector, the response data set including the water potential energy change rate, the inertia parameter of the pumped water / discharged water process, the air cavity pressure decay characteristic and the dynamic response time delay of the compression / expansion process; renewable energy input power prediction and power grid load prediction are performed, and the predicted results are used to construct an external demand tensor within the regulation window; the external demand tensor and the dual medium response characteristic vector are fused to output a regulation intention tensor; the regulation intention tensor is sent to a strategy generator to output a water-air joint output strategy matrix, the water-air joint output strategy matrix being used to represent the optimal output proportion and coupling strength coefficient of the air expander and the water turbine within the current regulation window; a joint scheduling instruction set is configured according to the water-air joint output strategy matrix, and the joint scheduling instruction set is sent to an execution terminal to perform coordinated energy storage control management.

[0007] In a second aspect of the present application, a pumped hydro storage coupled with compressed air energy storage system with water-air coordinated regulation is provided, the system comprising:

[0008] The characteristic vector construction module is used for constructing a double medium response characteristic vector through a multi-modal time sequence coding network after collecting a response data set of the water body and the air double medium, the response data set including a water body potential energy change rate, an inertial parameter of a water pumping / water releasing process, an air cavity pressure attenuation characteristic and a dynamic response time delay of a compression / expansion process; the prediction module is used for performing renewable energy input power prediction and power grid load prediction, and constructing an external demand tensor in an adjustment window by using a prediction result; the vector fusion module is used for fusing the external demand tensor and the double medium response characteristic vector, and outputting an adjustment intention tensor; the strategy matrix acquisition module is used for sending the adjustment intention tensor to a strategy generator, and outputting a water-air joint output strategy matrix, the water-air joint output strategy matrix being used for representing optimal output proportions and coupling intensity coefficients of the air expander and the water turbine in the current adjustment window; and the control management module is used for configuring a joint scheduling instruction set according to the water-air joint output strategy matrix, and issuing the joint scheduling instruction set to an execution terminal to perform cooperative energy storage control management.

[0009] One or more technical solutions provided in the present application have at least the following technical effects or advantages:

[0010] After collecting a response data set of the water body and the air double medium, the present application constructs a double medium response characteristic vector through a multi-modal time sequence coding network, the response data set including a water body potential energy change rate, an inertial parameter of a water pumping / water releasing process, an air cavity pressure attenuation characteristic and a dynamic response time delay of a compression / expansion process; performs renewable energy input power prediction and power grid load prediction, and constructs an external demand tensor in an adjustment window by using a prediction result; fuses the external demand tensor and the double medium response characteristic vector, and outputs an adjustment intention tensor; sends the adjustment intention tensor to a strategy generator, and outputs a water-air joint output strategy matrix, the water-air joint output strategy matrix being used for representing optimal output proportions and coupling intensity coefficients of the air expander and the water turbine in the current adjustment window; configures a joint scheduling instruction set according to the water-air joint output strategy matrix, and issues the joint scheduling instruction set to an execution terminal to perform cooperative energy storage control management. The present application solves the technical problems of insufficient water-air double energy storage medium cooperative adjustment capability and difficulty in adapting to power grid load and renewable energy fluctuation in the prior art, and achieves the technical effect of improving energy storage system response speed and overall energy utilization efficiency through cooperative adjustment control of water-air joint output. BRIEF DESCRIPTION OF DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed to be used in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without any creative effort on the basis of these drawings.

[0012] Figure 1 A schematic diagram of the pumped-storage coupled compressed-air energy storage method for water-air coordinated regulation provided in the embodiments of this application;

[0013] Figure 2 This is a schematic diagram of a pumped-storage energy storage system coupled with compressed air energy storage, which provides a water-air coordinated regulation system according to an embodiment of this application.

[0014] Figure labeling: Feature vector construction module 11, prediction module 12, vector fusion module 13, policy matrix acquisition module 14, control management module 15. Detailed Implementation

[0015] This application provides a pumped-storage energy storage coupled with compressed air energy storage method and system with water-air coordinated regulation. It addresses the technical problems of insufficient coordinated regulation capability of water and air dual energy storage media in the prior art, and difficulty in adapting to grid load and renewable energy fluctuations. Through coordinated regulation and control of water and air joint output, it achieves the technical effect of improving the response speed and overall energy utilization efficiency of the energy storage system.

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of this application, and not all of them. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.

[0017] It should be noted that any variation of the terms "comprising" and "having" is intended to cover non-exclusive inclusion, for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units that are explicitly listed, but may include other steps or modules that are not explicitly listed or that are inherent to such processes, methods, products, or devices.

[0018] Example 1, as Figure 1 As shown, this application provides a method for pumped-storage coupled with compressed-air energy storage with coordinated water and air regulation, the method comprising:

[0019] Step S100: After collecting the response datasets of water and air as dual media, a dual-media response characteristic vector is constructed through a multimodal temporal coding network. The response dataset includes the water potential energy change rate, inertial parameters of the pumping / draining process, air cavity pressure decay characteristics, and dynamic response delay of the compression / expansion process.

[0020] In the embodiments of the present application, when collecting the response data set of the water body and air dual medium, the operating state information of the water body and air medium is synchronously obtained at the corresponding positions of the pumped storage operation loop and the compressed air energy storage operation loop, the water level and water head change process of the upper and lower reservoirs is continuously recorded within the same adjustment time scale, and the water body potential energy change rate is formed according to the water level change relationship at adjacent sampling time points. At the same time, during the switching and power adjustment process of the pumping and discharging conditions, the speed change, guide vane opening change and output power change process of the water pump or water turbine are continuously recorded, and the inertia parameters of the pumping / discharging process for representing the dynamic lag characteristics of the pumping / discharging process are extracted. In the air energy storage process, the change of the air cavity pressure with time is continuously sampled to form the air cavity pressure decay characteristics reflecting the change law of the pressure with the storage or release process, and under the compression or expansion adjustment action, the time difference between the adjustment instruction application time and the key operating state reaching the response criterion time is recorded to form the dynamic response time delay of the compression / expansion process for representing the lag characteristics of the compression / expansion process, so as to obtain the response data set of the water body and air dual medium containing the water body potential energy change rate, the inertia parameters of the pumping / discharging process, the air cavity pressure decay characteristics and the dynamic response time delay of the compression / expansion process.

[0021] In constructing the dual medium response characteristic vector, the response data set of the water body and air dual medium is organized as a multivariate time sequence input sequence in a unified time sequence, so that the various response data form a corresponding relationship on the same time axis, and then the multivariate time sequence input sequence is input into the multimodal time sequence coding network. The water body response information corresponding to the water body potential energy change rate and the inertia parameters of the pumping / discharging process, and the air response information corresponding to the air cavity pressure decay characteristics and the dynamic response time delay of the compression / expansion process are jointly time sequence coded. Through continuous representation and overall convergence of the response characteristics at each time point in the time dimension, the dynamic response behaviors of the water body and air two media are mapped into a unified feature space, and finally the dual medium response characteristic vector which simultaneously represents the water body potential energy change characteristics and the air pressure change characteristics is formed.

[0022] Step S200: Perform renewable energy input power prediction and power grid load prediction, and use the prediction results to construct an external demand tensor in the adjustment window.

[0023] In the embodiments of the present application, when performing renewable energy input power prediction, first, wind power output data and photovoltaic output data collected in historical operation stages are used to construct a renewable energy historical output time sequence sample set, and the historical output data is arranged according to a unified time resolution, so that each training sample contains renewable energy output observation values of consecutive historical time points; then the renewable energy historical output time sequence sample set is input into a renewable energy input power prediction model for training, the renewable energy input power prediction model is composed of an input layer, a time sequence feature extraction layer and an output layer, wherein the input layer is used to receive the renewable energy historical output time sequence sample, the time sequence feature extraction layer recursively represents the change trend and periodic characteristics of renewable energy output in the time dimension based on a recurrent neural network structure, and the output layer is used to generate renewable energy input power output values corresponding to the prediction time; in the training process, the error between the prediction output value and the actual output observation value is used as the optimization target to iteratively update the model parameters, and after the model training is completed, the corresponding renewable energy input power prediction results of each time point in the future adjustment window are output.

[0024] When performing grid load prediction, a grid historical load time sequence sample set is constructed based on grid load data collected in historical operation stages, and the grid historical load data is arranged according to the same time resolution as the renewable energy input power prediction, so that each training sample contains grid load observation values of consecutive historical time points. Then the grid historical load time sequence sample set is input into a grid load prediction model for training, the grid load prediction model is composed of an input layer, a time sequence feature extraction layer and an output layer, wherein the input layer is used to receive the grid historical load time sequence sample, the time sequence feature extraction layer recursively represents the change trend and periodic characteristics of the grid load in the time dimension based on a recurrent neural network structure, and the output layer is used to generate grid load output values corresponding to the prediction time; in the training process, the error between the prediction output value and the actual load observation value is used as the optimization target to iteratively update the model parameters, and after the model training is completed, the corresponding grid load prediction results of each time point in the future adjustment window are output.

[0025] When performing renewable energy input power prediction, wind power output data and photovoltaic output data of consecutive time points are input into the renewable energy input power prediction model which has been constructed, the renewable energy input power prediction model performs time sequence analysis on the historical output change relationship, and outputs the corresponding renewable energy input power prediction results of each time point in the future adjustment window. When performing grid load prediction, grid historical load data of consecutive time points is input into the grid load prediction model which has been constructed, the grid load prediction model performs time sequence analysis on the historical load change relationship, and outputs the corresponding grid load prediction results of each time point in the future adjustment window.

[0026] After obtaining the renewable energy input power prediction result output by the renewable energy input power prediction model and the power grid load prediction result output by the power grid load prediction model, the two types of prediction results are aligned according to a unified time index, so that each prediction time in the adjustment window corresponds to a renewable energy input power prediction value and a power grid load prediction value at the same time, and the prediction results of each prediction time are combined and arranged in the time dimension, thereby constructing an external demand tensor in the adjustment window, which is used to uniformly represent the comprehensive external adjustment demand formed by the change of renewable energy power supply and the change of power grid load demand.

[0027] Step S300: fuse the external demand tensor and the dual-medium response characteristic vector to output an adjustment intention tensor.

[0028] Further, the method provided by the application embodiment further comprises:

[0029] According to the inertia parameter of the pumping / discharging process, a water body inertia time constant is constructed, short-period and long-period time bases are constructed by using the water body inertia time constant and the dynamic response time delay of the compression / expansion process respectively, and the external demand tensor is multi-scale decomposed to generate a short-period disturbance feature sub-tensor and a long-period trend sub-tensor, the short-period disturbance feature sub-tensor is used to represent the rapid power gap caused by renewable energy fluctuating input in the adjustment window, and the long-period trend sub-tensor represents the steady-state adjustment demand corresponding to the slow load trend; a mutual feedback relationship between the short-period disturbance feature sub-tensor, the long-period trend sub-tensor and the dual-medium vector characteristic vector is established by using bidirectional gate feedback; a physical constraint vector of water-air is established, the physical constraint vector includes a head-cavity equivalent pressure boundary consistency condition, a coupled power conservation condition and an air-water interface response threshold condition; the multi-scale modeled external demand sub-tensor is projected and modulated by using the physical constraint vector; a local sensitivity map of the water body potential energy change rate and the air cavity pressure decay rate is calculated based on the dual-medium response characteristic vector, a medium sensitivity factor is established, and the dynamic compression and amplification processing of the external demand tensor expression is driven by attention using the medium sensitivity factor; based on the sequentially executed mutual feedback processing, projection modulation and dynamic compression and amplification processing, an adjustment intention tensor is constructed.

[0030] In the embodiment of the present application, when the external demand tensor is fused with the double-medium response characteristic vector, first, the water body inertia time constant is constructed according to the inertia parameter of the pumping / discharging process. In this process, first, the power adjustment process of multiple pumping and discharging conditions in the adjustment window is selected, for each power adjustment process, the adjustment instruction issuing time and the corresponding power change sequence are recorded, the time when the power change first enters the preset stable interval is determined, and the time interval between the adjustment instruction issuing time and the time when the stable interval is entered is taken as the inertia time corresponding to the pumping or discharging process. Subsequently, statistical processing is performed on all the inertia times obtained in the adjustment window, and the water body inertia time constant is calculated in an arithmetic average manner, so as to obtain the water body inertia time constant representing the overall dynamic response characteristic of the water body. At the same time, the dynamic response time delay of the compression / expansion process is directly read as the lag time scale of the air medium, and the short-period time base is constructed with the time scale corresponding to the water body inertia time constant, and the long-period time base is constructed with the time scale corresponding to the dynamic response time delay of the compression / expansion process. The short-period time base and the long-period time base do not directly represent the inherent response speed of the water body or the air medium, but are used to characterize the change characteristics of the external adjustment demand at different time scales; the water body inertia time constant is used to determine the minimum adjustment time scale of the system disturbance formed by the renewable energy fluctuating input, and the dynamic response time delay of the compression / expansion process is used to determine the time span that can be smoothed and absorbed in the load trend adjustment.

[0031] Next, when the short-period time base and the long-period time base are used to perform multi-scale decomposition on the external demand tensor, first, the external demand tensor is expanded according to the time index into a continuous time sequence in the adjustment window, then the continuous time sequence is segmented at equal intervals with the time span corresponding to the short-period time base as the segmentation length, the change amount between adjacent time points in each segment is extracted, and the components with large change amounts are collected to form a short-period disturbance characteristic sub-tensor, so that the short-period disturbance characteristic sub-tensor is used to represent the rapid power gap caused by the renewable energy fluctuating input in the adjustment window. At the same time, the continuous time sequence is segmented with the time span corresponding to the long-period time base as the segmentation length, the overall trend in each segment is extracted, and the trend components of each segment are collected to form a long-period trend sub-tensor, so that the long-period trend sub-tensor is used to represent the steady-state adjustment demand corresponding to the slow change of the load.

[0032] After that, the mutual feedback relationship between the short-period perturbation feature sub-tensor, the long-period trend sub-tensor and the dual-medium response characteristic vector is established by using the bidirectional gated feedback. In this process, the forward and reverse gating modulation processes are set for the short-period perturbation feature sub-tensor and the long-period trend sub-tensor respectively, and the feature information transmission strength under different time scales is adaptively adjusted according to the local change characteristics of the water body potential energy change rate and the air cavity pressure decay characteristics in the dual-medium response characteristic vector. At the same time, the dual-medium dependent factor composed of the water body potential energy change rate and the air cavity pressure decay characteristics is introduced, and the correlation between the short-period perturbation feature sub-tensor and the long-period trend sub-tensor in the dual-medium response characteristic space is analyzed to form the cross-period interaction weight expression, and the forward and reverse gating results are updated bidirectionally based on the interaction weight. Finally, the short-period perturbation feature sub-tensor and the long-period trend sub-tensor updated by the bidirectional mutual feedback are jointly encoded to obtain the mutual feedback enhanced representation tensor for representing the synergistic effect of demand characteristics under different time scales under the constraint of dual-medium response.

[0033] After the construction of the mutual feedback enhanced representation tensor, the physical constraint vector of water-air is established to constrain the physical feasibility of multi-scale demand characteristics. In this process, first, the water head change range corresponding to the water body operation and the pressure cavity pressure change range corresponding to the air energy storage operation are read within the adjustment window, and they are mapped to the same physical quantity dimension to form the water head-pressure cavity equivalent pressure boundary consistency condition for limiting the matching relationship between water head and equivalent pressure. Then, the power distribution relationship between the water turbine output power and the air expander output power under the joint operation state is read to construct the coupled power conservation condition for constraining the energy balance relationship between the water body output power and the air output power. At the same time, the allowable response interval of the air-water interface under the joint action of water level change and pressure change is read to construct the air-water interface response threshold condition for limiting the dynamic response amplitude of the air-water interface. Finally, the above three types of conditions are combined according to the unified dimension to form the physical constraint vector of water-air.

[0034] Then, the external demand sub-tensor after multi-scale modeling is projected using the physical constraint vector. In this process, the mapping rule for constraining demand expression is constructed according to the water head-pressure cavity equivalent pressure boundary consistency condition, the coupled power conservation condition and the air-water interface response threshold condition, and the short-period perturbation feature sub-tensor and the long-period trend sub-tensor are input into the mapping rule to adjust their feature components, so that the two types of sub-tensors are limited within the feasible range that meets the physical constraint conditions of water-air on the basis of maintaining the original time scale characteristic structure, thereby obtaining the multi-scale external demand expression that meets the joint operation physical constraint requirements.

[0035] Then, based on the dual-medium response characteristic vector, the local sensitivity map of the water potential energy change rate and the air cavity pressure decay rate is calculated to characterize the response sensitivity of the water medium and the air medium to the external demand change. In this process, first, a continuous local time interval is selected in the adjustment window, and the corresponding water potential energy change rate sequence, air cavity pressure decay characteristic sequence and projected and modulated external demand sub-tensor change sequence in the time interval are extracted synchronously. Then, the change amplitudes between the water potential energy change rate and the external demand change, and between the air cavity pressure decay rate and the external demand change are compared at each time, respectively, to form a local sensitivity map reflecting the response strength difference of the water medium and the air medium to the external demand change. The local sensitivity map is normalized in the adjustment window to obtain the medium sensitivity factor.

[0036] Then, the dynamic compression and amplification processing of the external demand tensor expression under the attention driving is performed using the medium sensitivity factor. In this process, first, the medium sensitivity factor is aligned with the projected and modulated short-period disturbance feature sub-tensor and long-period trend sub-tensor according to the time index, so that the medium sensitivity factor at each time corresponds to the external demand feature at the same time. Then, the medium sensitivity value corresponding to the water potential energy change rate and the medium sensitivity value corresponding to the air cavity pressure decay rate are read at each time, respectively, and normalized processing is performed on the two, so that the sum of the sensitivity values corresponding to the water medium and the air medium at the same time is a preset constant, thereby generating the corresponding attention weight coefficient. The medium with a larger medium sensitivity value is assigned a higher weight coefficient, and the medium with a smaller medium sensitivity value is assigned a lower weight coefficient. On this basis, the feature components in the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are weighted at each time using the attention weight coefficient, the feature component corresponding to the higher weight coefficient is subjected to amplitude enhancement processing, so that its proportion in the external demand expression is increased, and the feature component corresponding to the lower weight coefficient is subjected to amplitude compression processing, so that its proportion in the external demand expression is reduced, thereby forming the multi-scale expression of the external demand after the dynamic compression and amplification processing.

[0037] Finally, the adjustment intention tensor is constructed based on the sequentially executed mutual feedback processing, projection modulation and dynamic compression amplification processing. In this process, first, the short-period disturbance feature sub-tensor and the long-period trend sub-tensor after the dynamic compression amplification processing are spliced in the time dimension, and are arranged in fusion with the double-medium response characteristic vector at the same time index, so that the modulation results of external demand on the short-period and long-period time scales form a one-to-one correspondence with the dynamic response capability of the water body and air double medium. Subsequently, the fused results are organized according to the time sequence of the adjustment window to form a unified tensor structure containing short-period disturbance adjustment demand, long-period steady-state adjustment demand, medium response sensitivity weight and physical constraint modulation result, and finally the adjustment intention tensor is output, which represents the comprehensive matching relationship between the external demand characteristics and the response capability of the water body and air double medium in the adjustment window.

[0038] Further, the method provided by the application embodiment further comprises:

[0039] The forward gating unit and the reverse gating unit are respectively established for the short-period disturbance feature sub-tensor and the long-period trend sub-tensor, the gating unit adjusts the gating weight through the local gradient change rate of the double-medium response characteristic vector, which is used to modulate the information transmission strength of different period characteristics; the double-medium dependent factor is generated according to the water body potential change rate and the air cavity pressure decay characteristic, the double-medium dependent factor is executed for interactive attention analysis, the correlation score of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor in the double-medium response characteristic space is calculated to generate a cross-period interaction weight matrix; the output of the forward gating unit and the reverse gating unit is updated by bidirectional mutual feedback based on the cross-period interaction weight matrix; the short-period disturbance feature sub-tensor and the long-period trend sub-tensor after bidirectional mutual feedback update are stacked and encoded to output a mutual feedback enhanced representation tensor.

[0040] In the embodiments of the present application, when the forward gating unit and the reverse gating unit are respectively established for the short-period disturbance feature sub-tensor and the long-period trend sub-tensor, at each time point, the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are aligned with the time index of the dual-medium response characteristic vector, and the feature vectors at the same time point are respectively taken out as the input of the gating calculation. Then, the local gradient change rate of the dual-medium response characteristic vector is calculated, specifically, the difference between the dual-medium response characteristic vectors at adjacent two time points is obtained to get the change amount, and the absolute value of the change amount is taken as the local gradient change rate. Then, the local gradient change rate is normalized to the interval of 0 to 1, and the normalized result is taken as the gating weight. The gating weight of the forward gating unit is used for element-wise multiplication weighting of the feature vector of the short-period disturbance feature sub-tensor or the long-period trend sub-tensor at the time point, and the gating weight of the reverse gating unit is used for element-wise multiplication weighting of the feature vector of the dual-medium response characteristic vector at the time point, so as to obtain the output of the forward gating unit and the output of the reverse gating unit, and to realize the modulation of the transmission intensity of the feature information of different periods.

[0041] When the dual-medium dependent factor is generated according to the water body potential energy change rate and the air cavity pressure decay characteristic, at each time point, the water body potential energy change rate value and the air cavity pressure decay characteristic value are respectively taken out from the dual-medium response characteristic vector, and the absolute values of the two are taken to eliminate the influence of the positive and negative directions. Then, the sum of the two is calculated, and the water body dependent component is obtained by dividing the absolute value of the water body potential energy change rate by the sum, and the air dependent component is obtained by dividing the absolute value of the air cavity pressure decay characteristic by the sum, so as to form the dual-medium dependent factor composed of the water body dependent component and the air dependent component at each time point, so that the dual-medium dependent factor directly represents the relative action proportion of the water medium and the air medium at the time point.

[0042] When the dual-medium dependent factor is executed for interactive attention analysis and a cross-period interaction weight matrix is generated, at each time point, the feature vectors of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are respectively taken out, and the inner product of the two is calculated as the correlation score. Then, the water body dependent component and the air dependent component in the dual-medium dependent factor are used to weight the correlation score, specifically, the correlation score is multiplied by the water body dependent component and the air dependent component to get two weighted scores, and the two weighted scores are added to get the final correlation score at the time point; then, the final correlation scores of the time points in the adjustment window are arranged in time sequence to form the cross-period interaction weight matrix, so that the cross-period interaction weight matrix is used to represent the cross-period correlation strength of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor in the dual-medium response characteristic space.

[0043] When the outputs of the forward gating unit and the backward gating unit are updated by bidirectional mutual feedback based on the cross-period interaction weight matrix, the weight value of the cross-period interaction weight matrix is read at each time point, and the weight value is used to perform element-wise multiplication weighting on the short-period disturbance feature sub-tensor vector and the long-period trend sub-tensor vector output by the forward gating unit, respectively. Then, the weighted short-period disturbance feature sub-tensor vector and the weighted long-period trend sub-tensor vector are element-wise added to obtain a period modulation vector, and the period modulation vector and the double-medium response characteristic vector output by the backward gating unit are element-wise added to form an update vector, where the element-wise addition process corresponds to an additive residual path, thereby completing bidirectional mutual feedback update, so that the updated short-period disturbance feature sub-tensor and the long-period trend sub-tensor can represent the coupling behavior characteristics of the water-gas double medium under different adjustment time scales.

[0044] When the short-period disturbance feature sub-tensor and the long-period trend sub-tensor updated by bidirectional mutual feedback are stacked and encoded to output a mutual feedback enhanced representation tensor, the updated short-period disturbance feature sub-tensor vector and the updated long-period trend sub-tensor vector are spliced in the feature dimension at each time point to form a stacked vector. Then, the arithmetic mean of the stacked vector in the feature dimension is calculated to obtain an encoded vector of a unified dimension, and the encoded vectors of each time point in the adjustment window are arranged in time sequence, and finally the mutual feedback enhanced representation tensor is output.

[0045] Further, the method provided by the application embodiment further comprises:

[0046] The bidirectional mutual feedback update constructs an additive residual path to jointly update the period modulation vector generated by the forward gating unit and the backward gating unit and the double-medium response characteristic vector, so that the short-period disturbance feature sub-tensor and the long-period trend sub-tensor represent the coupling behavior characteristics of the water-gas double medium under different adjustment time scales after being updated.

[0047] In the application embodiment, when performing bidirectional mutual feedback update, first, the output results of the forward gating unit and the backward gating unit are respectively obtained in the adjustment window according to a unified time index, wherein the period modulation vector output by the forward gating unit is used to represent the modulation result of the short-period disturbance feature sub-tensor or the long-period trend sub-tensor at the current time point, and the period modulation vector output by the backward gating unit is used to represent the feedback modulation result of the double-medium response characteristic vector at the current time point, and the period modulation vector and the double-medium response characteristic vector at the corresponding time point are aligned in dimension to ensure that each vector has the condition of element-wise joint processing at the same time point.

[0048] In the joint update of the additive residual path, the periodic modulation vector and the double medium response characteristic vector are updated by element-wise addition, that is, at each time point, the periodic modulation vector generated by the forward gating unit is added to the double medium response characteristic vector at the corresponding time point element by element to obtain an intermediate result superimposed with periodic modulation information, and then the intermediate result is added to the periodic modulation vector generated by the reverse gating unit element by element to form the double medium response characteristic vector after joint update, so that the modulation information from the short-period disturbance feature sub-tensor and the long-period trend sub-tensor is introduced on the basis of retaining the original dynamic characteristics of the double medium response characteristic vector, and the residual update of the double medium response characteristic vector is realized.

[0049] After the joint update of the double medium response characteristic vector is completed, the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are updated by feedback using the updated double medium response characteristic vector. In this process, at each time point, the updated double medium response characteristic vector is added to the corresponding time point characteristic vector of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor element by element, so that the short-period disturbance feature sub-tensor is updated to fuse the dynamic response information of the water potential energy change rate and the air cavity pressure decay characteristic of the double medium, and the long-period trend sub-tensor is updated to fuse the response characteristics of the water and air media to the steady-state regulation requirement, so that the updated short-period disturbance feature sub-tensor and the updated long-period trend sub-tensor respectively represent the coupling behavior characteristics of the water-air double medium at different regulation time scales.

[0050] Further, in the method provided by the application embodiment, the projection modulation of the external demand sub-tensor after the multi-scale modeling by using the physical constraint vector further comprises:

[0051] A projection operator is constructed based on the physical constraint vector; the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are respectively input into the projection operator, and the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are adjusted to the feasible region satisfying the constraint according to the linear or nonlinear mapping rule defined by the projection operator.

[0052] In the embodiments of the present application, when constructing the projection operator based on the physical constraint vector, first, the head-cavity equivalent pressure boundary consistency condition, the coupled power conservation condition and the gas-water interface response threshold condition are respectively converted into feasible region constraint parameters for the short-period perturbation characteristic sub-tensor and the long-period trend sub-tensor, specifically including the allowed lower limit and the allowed upper limit of the characteristic component corresponding to the head-cavity equivalent pressure boundary consistency condition, the allowed upper limit of the power combination corresponding to the coupled power conservation condition, and the allowed upper limit of the amplitude change and the allowed upper limit of the change rate corresponding to the gas-water interface response threshold condition. Subsequently, the above constraint parameters are written into the processing flow of correcting the characteristic components one by one, so that the projection operator can perform three types of correction operations of boundary truncation, scaling and rate limiting when inputting any characteristic component, thereby forming a complete projection operator.

[0053] When the short-period perturbation characteristic sub-tensor is input into the projection operator, first, the characteristic values of the short-period perturbation characteristic sub-tensor are read at each time according to the time sequence of the adjustment window, and boundary correction operations are performed on each characteristic value according to the head-cavity equivalent pressure boundary consistency condition, for example, when the power demand corresponding to the short-period perturbation characteristic sub-tensor at a certain time is higher than the allowed equivalent upper limit of the head and the cavity pressure, the power demand value is adjusted to the upper limit value, and when it is lower than the allowed lower limit, it is adjusted to the lower limit value. Subsequently, at the same time, the power demands corresponding to the short-period perturbation characteristic sub-tensor and the long-period trend sub-tensor after boundary correction are summed and compared with the power upper limit given by the coupled power conservation condition, when the sum exceeds the power upper limit, the characteristic values of the short-period perturbation characteristic sub-tensor are reduced in proportion, so that the sum of the adjusted short-period perturbation power demand and the long-period trend power demand is exactly equal to the power upper limit. Finally, according to the gas-water interface response threshold condition, the change amount of the short-period perturbation characteristic sub-tensor between adjacent two time points is limited, for example, when the power change amplitude between a certain time and the previous time exceeds the maximum change amplitude allowed by the gas-water interface, the change amplitude is adjusted to the maximum change amplitude, thereby obtaining a short-period perturbation characteristic sub-tensor that is smooth in the time dimension and meets the interface response capability.

[0054] In the process of inputting the long-period trend sub-tensor into the projection operator, the projection modulation operation is performed at each time instant in the same processing order as the short-period disturbance characteristic sub-tensor. First, the eigenvalues of the long-period trend sub-tensor are corrected by the upper and lower limits according to the head-cavity equivalent pressure boundary consistency condition, to ensure that the long-term power demand level does not exceed the equivalent energy carrying range of the water body and the air medium. Then, combined with the coupled power conservation condition, the power demands of the long-period trend sub-tensor and the short-period disturbance characteristic sub-tensor at the same time instant are coordinated. When the sum exceeds the allowed power upper limit, the eigenvalues of the long-period trend sub-tensor are proportionally adjusted to ensure the consistency of the steady-state regulation demand and the rapid regulation demand in power allocation. Finally, according to the air-water interface response threshold condition, the change amplitude of the long-period trend sub-tensor between consecutive time instants is limited to make the change process consistent with the stable response ability of the air-water interface on a long time scale, and the long-period trend sub-tensor that meets the physical constraint requirements is obtained.

[0055] Through the above step-by-step projection modulation process, the short-period disturbance characteristic sub-tensor and the long-period trend sub-tensor are respectively mapped into the feasible region that meets the head-cavity equivalent pressure boundary consistency condition, the coupled power conservation condition, and the air-water interface response threshold condition while maintaining their respective time scale semantic characteristics, forming a multi-scale external demand expression that meets the physical constraint requirements of the combined operation of the water and air media.

[0056] Step S400: Send the adjustment intention tensor to the strategy generator to output a water-air joint output strategy matrix, which is used to represent the optimal output proportion and coupling strength coefficient of the air expander and the water turbine in the current adjustment window.

[0057] In the process of sending the adjustment intention tensor to the strategy generator to output the water-air joint output strategy matrix in the embodiments of the present application, first, the strategy generator receives the adjustment intention tensor and analyzes the characteristic information in the adjustment intention tensor representing the short-period disturbance regulation demand and the long-period steady-state regulation demand. The adjustment characteristics at different time scales are separated to form corresponding periodic sub-feature representations. Then, combined with the double-medium response characteristic vector and the mutual feedback enhancement representation tensor formed in the adjustment intention tensor, the cooperative output relationship of the water turbine and the air expander in the combined operation state is analyzed, and the joint output coupling coefficient reflecting the cooperative action strength of the water turbine and the air expander in the current adjustment window is calculated.

[0058] On this basis, the periodic sub-feature representation and the joint output coupling coefficient are fused, and the output result is uniformly scaled by normalization constraint to form a water-air joint output strategy matrix describing the output power allocation proportion and coupling strength of the air expander and the water turbine at each adjustment time instant.

[0059] Further, the method provided by the application embodiment further comprises the following steps:

[0060] The strategy generator comprises a feature encoding layer, a coupling weight calculation layer, and an output fusion layer. The feature encoding layer is configured to receive the adjustment intention tensor, perform channel separation coding on the short-period disturbance and the long-period trend, and form a periodized sub-feature representation. The coupling weight calculation layer is configured to calculate the joint output coupling coefficient of the hydraulic turbine and the air expander by using the mutual feedback enhancement representation tensor of the bimedium response characteristic vector and the adjustment intention tensor. The output fusion layer is configured to integrate the joint output coupling coefficient and the periodized sub-feature representation to output the water-air joint output strategy matrix through nonlinear weighting and normalization operations.

[0061] In the application embodiment, the strategy generator comprises a feature encoding layer, a coupling weight calculation layer, and an output fusion layer. When the feature encoding layer performs channel separation coding on the adjustment intention tensor to form a periodized sub-feature representation, first, the adjustment intention tensor is input into the feature encoding layer in the time sequence of the adjustment window, and the channel data corresponding to the short-period disturbance is extracted as a short-period disturbance channel sequence, and the channel data corresponding to the long-period trend is extracted as a long-period trend channel sequence in the channel dimension. Then, linear transformation is performed on the short-period disturbance channel sequence and the long-period trend channel sequence at each time point, specifically, the channel vector at the time point is multiplied by a preset weight matrix to obtain a corresponding encoding vector. Finally, the short-period disturbance encoding vectors at each time point in the adjustment window are arranged in the time sequence to form a periodized sub-feature representation of the short-period disturbance, and the long-period trend encoding vectors at each time point are arranged in the time sequence to form a periodized sub-feature representation of the long-period trend, thereby forming the periodized sub-feature representation.

[0062] When the coupling weight calculation layer calculates the coupling strength coefficient of the hydraulic turbine and the air expander, first, the bimedium response characteristic vector and the mutual feedback enhancement representation tensor formed in the adjustment intention tensor are read at each adjustment time, and the two are aligned in the feature dimension. Then, the inner product between the bimedium response characteristic vector and the mutual feedback enhancement representation tensor is calculated to obtain a correlation value reflecting the degree of collaborative response of the water body and the air medium. The correlation value is normalized in the adjustment window to map it to a preset range, thereby obtaining the coupling strength coefficient corresponding to each adjustment time.

[0063] Finally, when generating the water-gas combined output strategy matrix by integrating the output fusion layer, first, the periodicized sub-feature representation of short-period disturbance and the periodicized sub-feature representation of long-period trend output by the feature encoding layer are read at each adjustment time, respectively, and the sum operation is performed on the respective periodicized sub-feature representations in the feature dimension. The numerical values of each feature component at the same time are added to obtain the short-period adjustment intensity value representing the strength of the rapid adjustment demand and the long-period adjustment intensity value representing the strength of the steady-state adjustment demand, respectively. Then, the coupling strength coefficient corresponding to the adjustment time is introduced to perform weighted processing on the short-period adjustment intensity value and the long-period adjustment intensity value. The short-period adjustment intensity value is multiplied by the coupling strength coefficient to obtain the weighted adjustment intensity of the air expander side, and the long-period adjustment intensity value is multiplied by a minus coupling strength coefficient to obtain the weighted adjustment intensity of the water turbine side, thereby establishing the adjustment correlation that the short-period disturbance feature is mainly borne by the air expander and the long-period trend feature is mainly borne by the water turbine.

[0064] After completing the weighted processing, the weighted adjustment intensity of the air expander side and the weighted adjustment intensity of the water turbine side are respectively subjected to nonlinear transformation, and the exponential values of the two are taken to enhance the discrimination between different adjustment intensities. Then, the normalized processing is performed on the results after the exponential transformation, so that the sum of the air expander output proportion and the water turbine output proportion is equal to the preset power distribution total amount constraint.

[0065] Finally, the air expander output proportion, the water turbine output proportion and the corresponding coupling strength coefficient corresponding to each adjustment time in the adjustment window are combined and arranged in time sequence to form a water-gas combined output strategy matrix. The water-gas combined output strategy matrix is used to represent the optimal output proportion of the air expander and the water turbine and the coupling strength coefficient thereof in the current adjustment window.

[0066] Step S500: According to the water-gas combined output strategy matrix, a joint scheduling instruction set is configured, and the joint scheduling instruction set is issued to an execution terminal to perform cooperative energy storage control management.

[0067] In the embodiments of the present application, when the joint scheduling instruction set is configured according to the water-gas combined output strategy matrix, the water-gas combined output strategy matrix is analyzed in the adjustment window, and the air expander output proportion, the water turbine output proportion and the coupling strength coefficient corresponding to each adjustment time are read according to the time index. Based on the total adjustment power demand at the corresponding time in the adjustment window, the proportional decomposition method is used to generate the air expander target output power and the water turbine target output power. The total adjustment power demand is multiplied by the air expander output proportion and the water turbine output proportion, respectively, to obtain the initial target output power of the two types of equipment at the adjustment time.

[0068] After introducing the coupling strength coefficient, the target output powers of the air expander and the water turbine are cooperatively corrected. In this process, the difference between the target output powers of the two types of equipment is adjusted according to the numerical value of the coupling strength coefficient. When the coupling strength coefficient is large, the mutual constraint between the target output powers of the two types of equipment is increased, so that the air expander and the water turbine tend to be consistent in the power change direction and the change amplitude. When the coupling strength coefficient is small, the above constraint is weakened, so that the air expander and the water turbine adjust the power independently according to the respective adjustment proportion, thereby realizing the cooperative power correction based on the coupling strength coefficient. After the cooperative correction of the target output powers is completed, the air expander correction target output power, the water turbine correction target output power and the corresponding adjustment time sequence information corresponding to each adjustment time are combined to form a joint scheduling instruction set containing the power set value and the time constraint relationship.

[0069] Finally, the joint scheduling instruction set is sent to the execution terminal, and the joint scheduling instructions are sent one by one in the time sequence of the adjustment window, so that the execution terminal adjusts the operating conditions of the air expander according to the air expander correction target output power, and adjusts the operating conditions of the water turbine according to the water turbine correction target output power, and simultaneously synchronously constrains the power change process of the two types of equipment according to the coupling strength coefficient, thereby realizing the cooperative adjustment of the air pressure energy and the water potential energy in the charging and discharging processes within the adjustment window, and completing the cooperative energy storage control management.

[0070] Further, the method provided by the application embodiment further comprises:

[0071] The execution terminal collects the operating state data of the water turbine and the air expander, and the operating state data includes the water head change rate, the pump speed, the cavity pressure, the output power and the response delay. The operating state data is compared with the adjustment intention tensor synchronously, and a control error vector is established. The control error vector is used for control compensation management.

[0072] In the application embodiment, in the cooperative energy storage control process, the execution terminal first continuously collects the operating states of the water turbine and the air expander. The collected operating state data includes the water head change rate for reflecting the water body operating condition change speed, the pump speed for reflecting the mechanical operating state of the water turbine or the pump group, the cavity pressure for reflecting the air energy storage operating condition, the output power for reflecting the actual energy release or absorption level, and the response delay for reflecting the time lag between the control instruction and the actual execution. The above operating state data is recorded according to a unified sampling period, thereby forming an operating state data sequence consistent with the time scale of the adjustment window.

[0073] After that, the running state data is synchronized and aligned with the adjustment intention tensor according to the time index, so that the actual running state at each sampling time is corresponded to the target adjustment requirement at the corresponding time in the adjustment intention tensor. Then, the head change rate, pump speed, cavity pressure, output power and response delay are compared item by item, the differences between each running state data and the corresponding target value in the adjustment intention tensor are calculated, and the differences are combined according to the preset order to form a control error vector including head change rate deviation, speed deviation, pressure deviation, power deviation and response delay deviation.

[0074] Finally, control compensation management is performed according to the control error vector. In this process, error decomposition is performed on the control error vector to distinguish the influence direction of each error component on the adjustment deviation of the hydraulic turbine and the air expander. Then, according to the change of the control error vector, the output proportion coefficient and the coupling strength coefficient in the water-air combined output strategy matrix are iteratively adjusted step by step, so that the updated output proportion coefficient and the coupling strength coefficient can continuously reduce the overall amplitude of the control error vector, thereby realizing control adaptive optimization management in the collaborative adjustment process of the hydraulic turbine and the air expander.

[0075] Further, the method provided by the application embodiment further comprises:

[0076] After error decomposition of the control error vector, the output proportion coefficient and the coupling strength coefficient are iteratively adjusted in the joint output strategy matrix space using gradient descent to perform control adaptive optimization management.

[0077] In the application embodiment, when performing control adaptive optimization management, first, the error decomposition is performed on the control error vector, and the head change rate error, pump speed error, cavity pressure error, output power error and response delay error contained therein are corresponded to the adjustment deviation sources of the hydraulic turbine and the air expander, so as to clarify the influence direction of each error component on the joint adjustment effect.

[0078] Then, taking the control error vector as the optimization basis, the output proportion coefficient in the joint output strategy matrix for describing the output distribution relationship of the air expander and the hydraulic turbine and the coupling strength coefficient for describing the cooperation degree of the two are taken as adjustable variables, and the above coefficients are iteratively updated in the gradient descent manner, that is, the adjustment direction of the coefficients is determined according to the change trend of the control error vector, and the output proportion coefficient and the coupling strength coefficient are slightly corrected in each iteration, so that the running state corresponding to the modified joint output strategy matrix in the next adjustment period is closer to the adjustment intention requirement.

[0079] The output proportion coefficient and the coupling strength coefficient are updated through continuous iteration, so that the overall amplitude of the control error vector is gradually reduced, thereby realizing adaptive optimization management of the water turbine and the air expander during combined operation.

[0080] Further, the method provided by the application embodiment further comprises:

[0081] It is judged whether the control error vector meets a preset deviation threshold, and if so, a control abnormality early warning is established, and a shutdown process and a warning reporting management are simultaneously performed.

[0082] In the application embodiment, firstly, the real-time generated control error vector is judged, and the head change rate error, the pump speed error, the cavity pressure error, the output power error and the response delay error contained in the control error vector are compared with the corresponding preset deviation threshold respectively, and when any error component in the control error vector reaches or exceeds its preset deviation threshold, it is judged that the current running state meets the abnormal condition.

[0083] After the abnormality is judged, a control abnormality early warning is established, and the early warning is taken as an abnormal state identifier. At the same time, a shutdown process is simultaneously performed, a shutdown instruction is issued to an execution terminal, so that the water turbine and the air expander enter a safe shutdown state, and a warning reporting management is performed, and control abnormality early warning information is sent to a monitoring unit for recording and alarming, so that timely shutdown and abnormal prompt are realized when the control deviation is out of limit.

[0084] In the application embodiment, as described above, the application embodiment has at least the following technical effects:

[0085] After the response data set of the water body and the air medium is collected, a multi-modal time sequence coding network is used to construct a double-medium response characteristic vector, the response data set includes the water body potential energy change rate, the inertia parameter of the pumping / water releasing process, the air cavity pressure decay characteristic and the dynamic response time delay of the compression / expansion process; renewable energy input power prediction and power grid load prediction are performed, and an external demand tensor in the adjustment window is constructed using the prediction results; the external demand tensor and the double-medium response characteristic vector are fused to output an adjustment intention tensor; the adjustment intention tensor is sent to a strategy generator to output a water-air joint output strategy matrix, the water-air joint output strategy matrix is used to represent the optimal output proportion and coupling strength coefficient of the air expander and the water turbine in the current adjustment window; a joint scheduling instruction set is configured according to the water-air joint output strategy matrix, and the joint scheduling instruction set is issued to an execution terminal to perform cooperative energy storage control management. The application solves the technical problems of insufficient water-air double energy storage medium cooperative adjustment capability and difficulty in adapting to power grid load and renewable energy fluctuation in the prior art, and through cooperative adjustment control of water-air joint output, the technical effects of improving the response speed and overall energy utilization efficiency of the energy storage system are achieved.

[0086] Embodiment two, based on the same inventive concept as the pumped hydro storage coupled with compressed air energy storage method of water-air coordinated regulation in the preceding embodiments, as Figure 2 As shown in the accompanying drawings, the present application provides a water-air coordinated regulation pumped hydro storage coupled with compressed air energy storage system, and the system and method embodiments in the present application are based on the same inventive concept. The system comprises:

[0087] A characteristic vector construction module 11 is configured to construct a dual-medium response characteristic vector through a multi-modal time series encoding network after collecting a response data set of the water body and air medium, the response data set including a water body potential energy change rate, an inertia parameter of a pumping / discharging process, an air cavity pressure decay characteristic, and a dynamic response time delay of a compression / expansion process; a prediction module 12 is configured to perform renewable energy input power prediction and power grid load prediction, and construct an external demand tensor within a regulation window using the prediction results; a vector fusion module 13 is configured to fuse the external demand tensor and the dual-medium response characteristic vector, and output a regulation intention tensor; a strategy matrix acquisition module 14 is configured to send the regulation intention tensor to a strategy generator, and output a water-air joint output strategy matrix, the water-air joint output strategy matrix being used to represent an optimal output proportion and coupling strength coefficient of an air expander and a water turbine within a current regulation window; and a control management module 15 is configured to configure a joint scheduling instruction set according to the water-air joint output strategy matrix, and send the joint scheduling instruction set to an execution terminal to perform coordinated energy storage control management.

[0088] Further, the system is also used to implement the following functions:

[0089] The inertia time constant of the water body is constructed according to the inertia parameter of the pumping / discharging process, the short-period time base and the long-period time base are respectively constructed by using the inertia time constant of the water body and the dynamic response time delay of the compression / expansion process, and the external demand tensor is multi-scale decomposed to generate a short-period disturbance feature sub-tensor and a long-period trend sub-tensor, the short-period disturbance feature sub-tensor is used to represent the rapid power gap caused by the renewable energy fluctuating input in the adjustment window, and the long-period trend sub-tensor represents the steady-state adjustment demand corresponding to the slow trend of the load; a mutual feedback relationship between the short-period disturbance feature sub-tensor, the long-period trend sub-tensor and the double-medium vector characteristic vector is established by using the bidirectional gate feedback; a physical constraint vector of water-air is established, the physical constraint vector includes a head-cavity equivalent pressure boundary consistency condition, a coupled power conservation condition and an air-water interface response threshold condition; the external demand sub-tensor after multi-scale modeling is projected and modulated by using the physical constraint vector; a local sensitivity map of the water body potential energy change rate and the air cavity pressure decay rate is calculated based on the double-medium response characteristic vector, a medium sensitivity factor is established, and the dynamic compression and amplification processing of the external demand tensor expression is driven by the medium sensitivity factor; and the adjustment intention tensor is constructed based on the sequentially executed mutual feedback processing, projection modulation and dynamic compression and amplification processing.

[0090] Further, the system is also used to implement the following functions:

[0091] The forward gate unit and the reverse gate unit are respectively established for the short-period disturbance feature sub-tensor and the long-period trend sub-tensor, the gate unit adjusts the gate weight through the local gradient change rate of the double-medium response characteristic vector, and is used to modulate the information transmission intensity of different period characteristics; the double-medium dependent factor is generated according to the water body potential energy change rate and the air cavity pressure decay characteristic, the double-medium dependent factor is subjected to interactive attention analysis, the cross-period interaction weight matrix is generated by calculating the correlation score of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor in the double-medium response characteristic space; the outputs of the forward gate unit and the reverse gate unit are bidirectionally fed back and updated based on the cross-period interaction weight matrix; the short-period disturbance feature sub-tensor and the long-period trend sub-tensor after bidirectional feedback update are stacked and encoded to output the mutual feedback enhanced representation tensor.

[0092] Further, the system is also used to implement the following functions:

[0093] The bidirectional feedback update constructs an additive residual path to jointly update the period modulation vectors generated by the forward gate unit and the reverse gate unit and the double-medium response characteristic vector, so that the short-period disturbance feature sub-tensor and the long-period trend sub-tensor represent the coupling behavior characteristics of the water-air double medium under different adjustment time scales after update.

[0094] Further, the system is also used to realize the following functions:

[0095] A projection operator is constructed based on the physical constraint vector; the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are respectively input into the projection operator, and the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are adjusted to the feasible region satisfying the constraint according to the linear or nonlinear mapping rules defined by the projection operator.

[0096] Further, the system is also used to realize the following functions:

[0097] The policy generator comprises a feature encoding layer, a coupling weight calculation layer and an output fusion layer, wherein the feature encoding layer is used to receive the adjustment intention tensor, channel separation coding is performed on the short-period disturbance and the long-period trend to form a periodized sub-feature representation; the coupling weight calculation layer is used to calculate the joint output coupling coefficient of the water turbine and the air expander by using the mutual feedback enhanced representation tensor of the bimedium response characteristic vector and the adjustment intention tensor; and the output fusion layer is used to integrate and output the joint output coupling coefficient, the periodized sub-feature representation, and a water-air joint output policy matrix through nonlinear weighting and normalization operations.

[0098] Further, the system is also used to realize the following functions:

[0099] The running state data of the water turbine and the air expander are collected by the execution terminal, and the running state data includes the water head change rate, the pump rotating speed, the cavity pressure, the output power and the response delay; the running state data and the adjustment intention tensor are compared synchronously to establish a control error vector; and the control error vector is used for control compensation management.

[0100] Further, the system is also used to realize the following functions:

[0101] It is judged whether the control error vector satisfies a preset deviation threshold value, if yes, a control abnormality early warning is established, and a shutdown processing and a warning reporting management are synchronously executed.

[0102] Further, the system is also used to realize the following functions:

[0103] After the control error vector is decomposed, the gradient descent is used to iteratively adjust the output proportion coefficient and the coupling strength coefficient in the joint output policy matrix space, and control adaptive optimization management is executed.

[0104] It should be noted that the above-mentioned order of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. And the above describes a specific embodiment of the present application. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0105] The above is only a preferred embodiment of the present application, and does not limit the present application in any form. Although the present application has been disclosed as above with a preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the technical solution of the present application, and any modification, equivalent change and modification made to the above embodiments according to the technical essence of the present application are still within the scope of the technical solution of the present application.

Claims

1. A pumped hydro energy storage coupled compressed air energy storage method with water-air coordinated regulation, characterized in that, The method comprises: After collecting the response data set of the water and air dual medium, a dual medium response characteristic vector is constructed by a multi-modal time sequence coding network, the response data set comprising a water body potential energy change rate, an inertial parameter of a water pumping / discharging process, an air cavity pressure decay characteristic and a dynamic response time delay of a compression / expansion process; Performing renewable energy input power prediction and power grid load prediction, and constructing an external demand tensor within an adjustment window using the prediction results; Fusing the external demand tensor with the dual medium response characteristic vector to output an adjustment intention tensor; Sending the adjustment intention tensor to a strategy generator to output a water-gas combined output strategy matrix, the water-gas combined output strategy matrix being used to represent the optimal output proportion and coupling strength coefficient of the air expander and the water turbine within the current adjustment window; According to the water-gas combined output strategy matrix, a joint scheduling instruction set is configured, and the joint scheduling instruction set is issued to an execution terminal to perform cooperative energy storage control management; Fusing the external demand tensor with the dual medium response characteristic vector to output an adjustment intention tensor, comprising: According to the inertial parameter of the water pumping / discharging process, a water body inertial time constant is constructed, and the water body inertial time constant and the dynamic response time delay of the compression / expansion process are used to construct a short-period time base and a long-period time base, respectively, and the external demand tensor is multi-scale decomposed to generate a short-period disturbance feature sub-tensor and a long-period trend sub-tensor, the short-period disturbance feature sub-tensor being used to represent a rapid power gap caused by renewable energy fluctuating input within the adjustment window, and the long-period trend sub-tensor representing a steady-state adjustment demand corresponding to a slow load trend; A mutual feedback relationship between the short-period disturbance feature sub-tensor, the long-period trend sub-tensor and the dual medium vector characteristic vector is established by using a bidirectional gate feedback; A hydraulic-air physical constraint vector is established, the physical constraint vector comprising a water head-air cavity equivalent pressure boundary consistency condition, a coupling power conservation condition and an air-water interface response threshold condition; The external demand sub-tensor after multi-scale modeling is projected and modulated by using the physical constraint vector; Based on the dual medium response characteristic vector, a local sensitivity map of the water body potential energy change rate and the air cavity pressure decay rate is calculated, a medium sensitivity factor is established, and dynamic compression and amplification processing of the external demand tensor expression is performed under attention driving by using the medium sensitivity factor; Based on sequentially executed mutual feedback processing, projection modulation and dynamic compression and amplification processing, an adjustment intention tensor is constructed.

2. The water-air cogregated pumped hydro coupled compressed air energy storage method of claim 1, wherein, The mutual feedback relationship between the short-period disturbance feature sub-tensor, the long-period trend sub-tensor and the dual medium vector characteristic vector is established by using a bidirectional gate feedback, comprising: A forward gate unit and a reverse gate unit are respectively established for the short-period disturbance feature sub-tensor and the long-period trend sub-tensor, the gate unit adjusting gate weights through a local gradient change rate of the dual medium response characteristic vector, and being used to modulate the information transmission intensity of different period characteristics; A bimedium-dependent factor is generated according to the water body potential energy change rate and the air cavity pressure decay characteristics, the bimedium-dependent factor is subjected to interactive attention analysis, a cross-period interactive weight matrix is generated by calculating the correlation scores of the short-period disturbance feature sub-tensor and the long-period trend sub-tensor in the bimedium response characteristic space; The outputs of the forward gate unit and the reverse gate unit are subjected to bidirectional mutual feedback update based on the cross-period interactive weight matrix; The short-period disturbance feature sub-tensor and the long-period trend sub-tensor subjected to bidirectional mutual feedback update are stacked and encoded, and a mutual feedback enhanced representation tensor is output.

3. The water-air cogregated pumped hydro coupled compressed air energy storage method of claim 2, wherein, The bidirectional mutual feedback update updates the period modulation vector generated by the forward gate unit and the reverse gate unit and the bimedium response characteristic vector jointly through the construction of an additive residual path, so that the short-period disturbance feature sub-tensor and the long-period trend sub-tensor represent the coupling behavior characteristics of the water-air bimedium under different adjustment time scales after update.

4. The hydropneumatic co-regulated pumped storage coupled compressed air energy storage method of claim 1, wherein, The external demand sub-tensor subjected to multi-scale modeling by using the physical constraint vector is projected and modulated, including: A projection operator is constructed based on the physical constraint vector; The short-period disturbance feature sub-tensor and the long-period trend sub-tensor are respectively input into the projection operator, and the short-period disturbance feature sub-tensor and the long-period trend sub-tensor are adjusted to the feasible region that satisfies the constraint according to the linear or nonlinear mapping rules defined by the projection operator.

5. The hydropneumatic co-regulated pumped storage coupled compressed air energy storage method of claim 1, wherein, The adjustment intention tensor is sent to a policy generator, and a water-air joint output strategy matrix is output, including: The policy generator includes a feature encoding layer, a coupling weight calculation layer, and an output fusion layer, wherein the feature encoding layer is used to receive the adjustment intention tensor, channel separation coding is performed on the short-period disturbance and the long-period trend to form a periodized sub-feature representation; The coupling weight calculation layer is used to calculate the joint output coupling coefficient of the water turbine and the air expander by using the mutual feedback enhanced representation tensor of the bimedium response characteristic vector and the adjustment intention tensor; The output fusion layer is used to integrate the joint output coupling coefficient, the periodized sub-feature representation, and output the water-air joint output strategy matrix through nonlinear weighting and normalization operations.

6. The hydropneumatic co-regulated pumped storage coupled compressed air energy storage method of claim 1, wherein, Cooperative energy storage control management is performed, including: The running state data of the water turbine and the air expander are collected by using an execution terminal, and the running state data includes the water head change rate, the pump rotating speed, the cavity pressure, the output power, and the response delay; The running state data and the adjustment intention tensor are compared synchronously to establish a control error vector; Control compensation management is performed according to the control error vector.

7. The hydropneumatic co-regulated pumped storage coupled compressed air energy storage method of claim 6, wherein, It is judged whether the control error vector satisfies a preset deviation threshold value, if yes, a control abnormality early warning is established, and synchronous shutdown processing and early warning reporting management are performed.

8. The hydropneumatic co-regulated pumped storage coupled compressed air energy storage method of claim 6, wherein, Control compensation management is performed according to the control error vector, including: After error decomposition of the control error vector, the output proportion coefficient and the coupling strength coefficient are iteratively adjusted in the joint output strategy matrix space by using gradient descent to perform control adaptive optimization management.

9. A pumped hydro coupled compressed air energy storage system with water-air coordinated regulation, characterized in that, The system is used to perform the water-air cooperative adjustment pumped storage energy coupling compressed air energy storage method according to any one of claims 1-8, and the system includes: A characteristic vector construction module is configured to construct a dual-medium response characteristic vector through a multi-modal time series encoding network after collecting a response data set of the water body and air dual medium, the response data set including a water body potential energy change rate, an inertial parameter of a water pumping / discharging process, an air cavity pressure decay characteristic, and a dynamic response time delay of a compression / expansion process; A prediction module is configured to perform renewable energy input power prediction and power grid load prediction, and construct an external demand tensor within an adjustment window using the prediction results; A vector fusion module is configured to fuse the external demand tensor and the dual-medium response characteristic vector, and output an adjustment intention tensor; A strategy matrix acquisition module is configured to send the adjustment intention tensor to a strategy generator, and output a water-air joint output strategy matrix, the water-air joint output strategy matrix being used to represent an optimal output proportion and a coupling strength coefficient of an air expander and a water turbine within a current adjustment window; A control management module is configured to configure a joint scheduling instruction set according to the water-air joint output strategy matrix, and send the joint scheduling instruction set to an execution terminal to perform cooperative energy storage control management.

Images