Method and system for managing energy of off-network energy storage equipment

Through task segmentation and satellite-assisted computing, deep learning neural networks and linear planning models are used to optimize task allocation, solving the problem of insufficient computing capabilities of off-grid energy storage equipment, and achieving efficient and stable energy management.

CN118940899APending Publication Date: 2024-11-12天能新能源(湖州)有限公司
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411020766.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-12

AI Technical Summary

Technical Problem

Due to insufficient computing power, existing off-grid energy storage equipment leads to long calculation time for energy management, large energy consumption, low utilization rate, and is prone to out of control when load demand surges.

Method used

Through task segmentation and satellite-assisted computing, deep learning neural networks and linear programming models are used to adjust the segmentation ratio vectors, optimize task allocation and resource allocation, reduce total energy consumption, and improve computing power.

Benefits of technology

It effectively reduces the energy consumption of off-grid energy storage equipment, improves energy utilization, reduces calculation pressure, avoids the risk of system out of control, and improves processing and response capabilities.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118940899A_ABST
    Figure CN118940899A_ABST
Patent Text Reader

Abstract

The invention provides an off-network energy storage equipment energy management method. The method comprises the following steps: S1, acquiring a general task; reading a total task set stored in the off-network energy storage equipment, wherein the total task set comprises a plurality of tasks; step S2, task segmentation; segmenting each task into sub-tasks according to a segmentation ratio vector, wherein the segmentation ratio vector reflects the proportion of segmenting each task to energy storage equipment for processing and segmenting each task to satellite for processing; s3, calculating energy consumption; the total energy consumption E under the segmentation ratio vector is calculated, the total energy consumption E comprises satellite total energy consumption E1 and energy storage equipment total energy consumption E2, and the satellite total energy consumption E1 is the sum of satellite subtask processing total energy consumption E11 and satellite subtask transmission total energy consumption E12; s4, adjusting calculation; the segmentation ratio vectors are changed, and corresponding total energy consumption E under different segmentation ratio vectors is calculated; s5, selecting the minimum total energy consumption; the minimum total energy consumption Emin is selected from all the calculated total energy consumption Es, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of off-grid energy storage, and in particular to a method and system for energy management of off-grid energy storage equipment. Background Art

[0002] Energy storage systems are devices or technologies used to store energy. They can store energy when it is generated and release it when needed. Off-grid energy storage systems are energy storage systems that are disconnected from the power grid and the network at the same time. They are mostly deployed in remote areas where communications are not covered, such as mountainous areas, deserts, and islands. They are used in conjunction with batteries through new energy generation methods such as photovoltaic modules. Since they cannot be connected to the network, all data processing and computing needs of off-grid energy storage systems are concentrated on local equipment.

[0003] The existing off-grid energy storage system usually uses small industrial control computers for energy management. Small industrial control computers are compact computers designed for industrial control and can operate reliably in harsh industrial environments. Small industrial control computers used in off-grid energy storage local equipment mostly use CPUs with low main frequency and small cache capacity.

[0004] Due to the lack of high-speed CPU support with strong computing power and the large energy management computing requirements of off-grid energy storage equipment, the energy management computing time of off-grid energy storage equipment in the existing technology is long, the energy consumption is large, and the energy utilization rate is low. In special circumstances where the load demand surges, the computing power of the local off-grid energy storage equipment is likely to reach its limit, and the computing power is insufficient to handle all computing tasks, resulting in the system facing performance bottlenecks or even the risk of loss of control. Summary of the invention

[0005] In view of the above-mentioned defects of the prior art, the present invention provides a method and system for energy management of off-grid energy storage equipment, which reduces energy waste, improves energy utilization, and solves the problem of insufficient computing power of local off-grid energy storage equipment.

[0006] In order to achieve the above object, the technical solution adopted by the present invention is:

[0007] A method for energy management of an off-grid energy storage device, comprising:

[0008] Step S1: acquiring a total task; reading a total task set stored in an off-grid energy storage device, wherein the total task set includes a plurality of tasks;

[0009] Step S2: task segmentation; segmenting each of the tasks into subtasks according to a segmentation ratio vector, wherein the segmentation ratio vector reflects the ratio of each of the tasks segmented to the energy storage device for processing and to the satellite for processing;

[0010] Step S3: energy consumption calculation; calculating the total energy consumption E under the split ratio vector, wherein the total energy consumption E includes the total energy consumption E1 of the satellite and the total energy consumption E2 of the energy storage device, and the total energy consumption E1 of the satellite is the sum of the total energy consumption E11 of the satellite subtask processing and the total energy consumption E12 of the satellite subtask transmission;

[0011] Step S4: adjusting the calculation; changing the split ratio vector, and calculating the total energy consumption E corresponding to different split ratio vectors;

[0012] Step S5: selecting the minimum total energy consumption; selecting the minimum total energy consumption Emin from all the calculated total energy consumptions E, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector.

[0013] The energy management of off-grid energy storage devices requires the completion of complex computing tasks, including but not limited to battery state estimation, power station operation diagnosis, intelligent strategy scheduling and log alarm analysis. Therefore, the method first obtains the computing tasks stored in the off-grid energy storage device. The acquisition and storage of computing tasks in the present invention are prior art. These tasks that need to be calculated are stored in a total task set, which contains multiple tasks, and a single task is represented in the form of a tuple.

[0014] Through the initialization algorithm and adjustment calculation, a split ratio vector is generated. The split ratio vector represents the ratio of task splitting. Therefore, the split ratio of each task is between 0 and 1, and the split ratio vector is normalized within [0, 1]. The total task set contains multiple tasks. These tasks are split into energy storage device processing and satellite processing according to the split ratio vector. The split tasks are subtasks, that is, each task is divided into two subtasks, one for energy storage device processing and the other for satellite processing. The total energy consumption under the split ratio (the sum of off-grid energy storage device energy consumption and satellite energy consumption) is obtained from the initial split ratio vector.

[0015] In order to generate the optimal computing task splitting scheme and obtain the minimum total energy consumption, the present invention adjusts the calculation to obtain a more optimal split ratio vector and obtains a more optimal total energy consumption. The adjustment algorithm includes but is not limited to a deep learning neural network and a linear programming model.

[0016] The present invention divides energy management into two sub-problems: task division and resource allocation. For the task division sub-problem, a split ratio vector is generated by adjusting the calculation. For the resource allocation problem, a linear programming method is used to obtain the minimum total energy consumption under the split ratio vector based on the split ratio vector obtained from the adjustment calculation and the computing resources and bandwidth resources allocated by the satellite to the sub-tasks of each task, and then use this minimum total energy consumption to adjust the calculation to obtain a better split ratio vector.

[0017] In summary, this method overcomes the shortcomings of insufficient computing power of off-grid energy storage equipment by splitting computing tasks and using satellites. It can reduce the energy consumption of the off-grid energy storage system itself, improve energy utilization, optimize energy management, and have better application prospects.

[0018] Preferably, in step S4, a deep learning neural network is used to automatically change the segmentation ratio vector.

[0019] Deep learning neural networks are a type of algorithm inspired by the structure of the human brain. They use multi-layer artificial neural networks to simulate the way the human brain processes information. Their application requires data preprocessing, selection of appropriate network architecture, adjustment of hyperparameters, training and evaluation of models, etc. Common types of deep learning neural networks include deep deterministic policy gradients and convolutional neural networks. Using deep learning neural networks to automatically adjust segmentation vectors is more efficient than other algorithm models, and can automatically learn complex features from data without the need for manual adjustment of data.

[0020] Preferably, the deep learning neural network is a double-delay deep deterministic policy gradient network.

[0021] The double-delayed deep deterministic policy gradient network is an advanced deep reinforcement learning algorithm, which is an improvement on the original deep deterministic policy gradient algorithm. The advantages of the double-delayed deep deterministic policy gradient network are that it improves the stability of the algorithm, reduces the over-estimation of value estimation by using a double critic network, introduces exploration noise to enhance the exploration ability of the strategy, and achieves better performance in a variety of complex environments through smoothing the target strategy and delaying the update mechanism.

[0022] Preferably, in step S4, the deep learning neural network is initialized, and the initialization includes defining the state, action and reward of the deep learning neural network; the state includes necessary data for the task; the action is to generate a segmentation ratio vector; the reward is to set a reward function that is negatively correlated with the total energy consumption value E, which is used to train the deep learning neural network to generate a better segmentation ratio vector.

[0023] Deep learning neural networks require an initialization state before they can start learning. Specifically, these initial states include but are not limited to the necessary data for each task in the computing task set, such as the size of the task, the computing resources required to process each bit of the task, the maximum tolerable delay of the task, the ratio of the size of the processed task to the original task size, etc. The initial state also includes the computing resources and bandwidth allocated by the LEO satellite to the specific computing tasks in the computing task set. Specifically, the agent selects an action based on the initial state, that is, generates a split ratio vector. After executing the action, the agent receives a reward, which is a reward function that is negatively correlated with the total energy consumption E, that is, if the total energy consumption E corresponding to the split ratio vector is greater than the total energy consumption E corresponding to the previously generated split ratio vector, the reward function will be smaller. In order to achieve the goal of maximizing the long-term cumulative reward, the next action of the agent will generate a better split ratio vector to obtain a smaller total energy consumption E.

[0024] Preferably, the necessary data of each task includes the task size, the computing resources required to process each bit of the task, the maximum tolerable delay of the task, and the ratio of the processed task size to the original task size.

[0025] The necessary data of each task in the total task set provides support for subsequent total energy consumption calculation, setting reward function and training deep learning neural network.

[0026] Preferably, in step S4, the minimum total energy consumption E corresponding to the allocation of different bandwidths and computing resources under each segmentation ratio vector is calculated according to a linear programming model.

[0027] The split ratio vector and the bandwidth and computing resources allocated to each subtask processed by the energy storage device and the satellite are brought into the linear programming model. Different allocation schemes result in different total energy consumption E. The minimum total energy consumption E under the split ratio vector is obtained through the linear programming model and the constraints. The linear programming model and the constraints are set by the present invention. The maximum reward function can be obtained with this minimum total energy consumption E, so the optimal split ratio vector in this adjustment calculation can be obtained.

[0028] Preferably, the satellite is a low-orbit earth satellite.

[0029] Low-orbit earth satellites generally refer to satellites operating in orbits approximately 2,000 kilometers above the earth's surface. Their main advantages include: low communication latency, fast signal transmission speed, and low latency; high data transmission rate, able to provide fast data communication services; and good global coverage capabilities, which can achieve global communication coverage.

[0030] A system for energy management of off-grid energy storage equipment, comprising:

[0031] A total task acquisition module, the total task acquisition module is used to read the total task set stored in the off-grid energy storage device, the total task set includes multiple tasks;

[0032] A calculation module, the calculation module is in communication with the overall task acquisition module and comprises the following submodules: a segmentation and calculation module, an adjustment module and an optimal generation module;

[0033] The segmentation and calculation module calculates the total energy consumption E corresponding to the segmentation ratio vector based on the segmentation ratio vector, wherein the total energy consumption E includes the total energy consumption E1 of the satellite and the total energy consumption E2 of the energy storage device, and the total energy consumption E1 of the satellite is the sum of the total energy consumption E11 of the satellite subtask processing and the total energy consumption E12 of the satellite subtask transmission;

[0034] The adjustment module is used to change the segmentation ratio vector;

[0035] The optimal generation module is used to screen out the minimum total energy consumption Emin, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector.

[0036] Preferably, it also includes a power protection module; the power protection module is communicatively connected with the general task acquisition module and the calculation module, and is used to monitor the input and output voltage and current of the energy storage device.

[0037] Power protection devices are used to monitor the input and output voltage and current of energy storage systems. When the voltage exceeds or falls below the preset safety threshold, the system is automatically disconnected or the power output is adjusted; the current level is monitored to prevent excessive or insufficient current. Excessive current may cause wire overheating and equipment damage, while too low current may indicate a system fault; the current is quickly cut off when a short circuit is detected to prevent damage to the energy storage device and the connected electrical system; when the system is started or the power suddenly changes, the output power is gradually increased or decreased to avoid the impact of instantaneous large current on the equipment.

[0038] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0039] 1. First, the division of computing tasks and the assistance of satellite computing can reduce the computing pressure of local equipment of off-grid energy storage equipment, greatly reduce the risk of out-of-control caused by insufficient computing power of off-grid energy storage equipment, and achieve stable data transmission and communication in remote areas, thereby improving the processing and response capabilities of the overall off-grid energy storage system. Secondly, by adjusting the calculation and changing the split ratio vector to obtain a better split ratio vector and minimum total energy consumption, energy waste is reduced and energy utilization is improved.

[0040] 2. The double-delayed deep deterministic policy gradient network training is used to calculate the minimum total energy consumption and the better segmentation ratio vector, which greatly reduces the energy consumption compared with the common deep learning neural network SoftActor-Critic (SAC) method and deep deterministic policy gradient (DDPG) method.

[0041] 3. The use of low-orbit earth satellites ensures high-speed, low-latency communications around the world, providing a stable and reliable data transmission channel for ground energy storage stations. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of the method steps of Example 1 of the present invention;

[0043] Figure 2 This is a schematic diagram of the system architecture of Example 2 of the present invention. DETAILED DESCRIPTION

[0044] In order to make the technical means, creative features, objectives and effects of the invention easier to understand, the invention is further described with reference to specific diagrams. However, the invention is not limited to the following implementation cases.

[0045] It should be noted that the structures, proportions, sizes, etc. illustrated in the drawings in this specification are only used to match the contents disclosed in the specification so as to facilitate understanding and reading by persons familiar with this technology. They are not used to limit the conditions under which the present invention can be implemented, and therefore have no substantive technical significance. Any structural modification, change in proportion or adjustment of size, without affecting the effects and purposes that can be achieved by the present invention, should still fall within the scope of the technical contents disclosed by the present invention.

[0046] Embodiment 1:

[0047] like Figure 1 A method for energy management of an off-grid energy storage device is shown, comprising:

[0048] Step S1: Total task acquisition; read the total task set stored in the off-grid energy storage device, the total task set includes multiple tasks; specifically, these tasks include but are not limited to battery state estimation, operation diagnosis, strategy scheduling, alarm analysis and other tasks, the tasks are collected, generated and stored by the energy storage device, and after being generated, they are fixed tasks to be processed and stored in the form of data. The generation and storage of computing tasks is a prior art, and the present invention does not make any special limitation or modification to this part.

[0049] The computing task model in this embodiment is as follows: K = {1, 2, ..., |K|} represents a task set, which contains multiple tasks, and a single task is represented in the form of a tuple. Specifically, the tuple (α k ,β k, χ k ) to represent the kth task. Among them, k is in the task set K, α k represents the size of the task, β k is the computing resource required to process each bit of the task (the computing resource referred to in this invention is the consumed CPU resource), is the maximum tolerable delay of the task, χ k It is the ratio of the processed task size to the original task size. These tasks are collected, generated and stored by energy storage devices. After they are generated, they become fixed tasks to be processed. Therefore, the data in the task tuple is known and fixed.

[0050] Step S2: Task segmentation: Each task in the total task set is segmented into subtasks according to the segmentation ratio vector. The segmentation ratio vector reflects the ratio of each task in the total task set that is segmented to the energy storage device for processing and to the satellite for processing. Under the optimal segmentation ratio vector, the total energy consumption value is minimized. The total task set contains multiple tasks, which are segmented to the energy storage device for processing and to the satellite for processing according to the segmentation ratio vector. The segmented tasks are subtasks, that is, each task is divided into two subtasks, one for the energy storage device for processing and the other for the satellite for processing. For example, the part of the k-th task that is segmented to the energy storage device for processing is subtask a, and the part that is segmented to the satellite for processing is subtask b.

[0051] The satellites used in this embodiment are low earth orbit satellites (hereinafter referred to as "LEO satellites"). The number of LEO satellites can be one, two, or more, depending on the actual network configuration. In this embodiment, multiple LEO satellites are used. The LEO satellite model used in this embodiment is as follows: Let S = {1, 2, ..., |S|} represent the LEO satellite set, Q s Represents the set of tasks that LEO satellites need to handle. There are constraints:

[0052]

[0053] The above constraints mean that each computing task in the task set is processed by only one LEO satellite with s∈S. The bandwidth allocated to the LEO satellite s for the subtask b of the kth task in the task set K, The computing resources allocated to the LEO satellite s for the subtask b of the kth task in the task set K, and It is set to calculate the total energy consumption of different allocation schemes under each split ratio vector.

[0054] is the coverage time of LEO satellites passing through the energy storage device in the sky. Assuming that each LEO satellite has a consistent operating speed and altitude, is a known constant.

[0055] ω k is the split ratio of the kth task, 0≤ω k ≤1, which means ω k Part of the subtask a is processed in the off-grid energy storage device, (1-ω k ) part of subtask b is offloaded to the LEO satellite for processing.

[0056] Step S3: energy consumption calculation; calculate the total energy consumption E under the split ratio vector, the total energy consumption E includes the total energy consumption of the satellite E1 and the total energy consumption of the energy storage device E2, E1 is the sum of the total energy consumption of the satellite subtask processing E11 and the total energy consumption of the satellite subtask transmission E12;

[0057] Step S4: adjusting the calculation; changing the split ratio vector, and calculating the total energy consumption E corresponding to different split ratio vectors;

[0058] Step S5: selecting the minimum total energy consumption; selecting the minimum total energy consumption Emin, the division ratio vector corresponding to the minimum total energy consumption Emin is the optimal division ratio vector.

[0059] Preferably, step S4 also includes optimizing the generation of a better segmentation ratio vector using a deep learning neural network. First, the deep learning neural network is initialized. This embodiment uses a double-delayed deep deterministic policy gradient network (hereinafter referred to as "TD3 neural network"). In the deep learning neural network, state, action, and reward are the core concepts in reinforcement learning, which define the basic framework of the interaction between the agent and the environment. Specifically, the agent selects an action based on the current state; after executing the action, the agent receives a new state and reward: this information is used to update the agent's strategy and value function (to help the agent evaluate the quality of the current strategy). Through the reward signal, the agent continuously learns and updates its strategy: in order to achieve the goal of maximizing long-term cumulative rewards. The training logic and basic framework of the TD3 neural network are prior art, and the above explanation is only used to understand the technical solution of the present invention.

[0060] The states, actions and rewards defined in this invention are as follows:

[0061] State: The state space is a one-dimensional vector containing the necessary data of each task tuple in the task set. Each task tuple in the task set K of this embodiment is as described in the task model. The task tuple is (α k ,β k , χ k ), and the bandwidth allocated by LEO satellite s to the kth mission subtask b and the computing resources allocated to the kth task subtask b

[0062] Action: Generate a split ratio vector, which represents the split ratio of each task in the task set; in this embodiment, the action is to generate a split ratio vector Ω = [ω1, ω2, …, ω |K| ], represents the split ratio of each task in the task set K. For example, the task set K includes task k1, task k2, task k3..., and the split ratio vector is Ω = [0.1, 0.3, 0.5...], which means that 10% of task k1 is processed by the off-grid energy storage device, and the rest is processed by the LEO satellite, 30% of task k2 is processed by the off-grid energy storage device, and the rest is processed by the LEO satellite, 50% of task k3 is processed by the off-grid energy storage device, and the rest is processed by the LEO satellite...

[0063] Reward: A reward function that is negatively correlated with total energy consumption. See below for details.

[0064] Since the energy storage device offloads the task to the LEO satellite for wireless communication, a certain data transmission time is required, that is, the delay of wireless communication. The delay model and communication model used in this embodiment are as follows:

[0065] Energy storage device delay model: The delay of the energy storage device is only the time it takes for the energy storage device to process data.

[0066]

[0067] set up (cycles / s) is the computing power provided by the energy storage device to the subtask a of the kth task. The processing time of the subtask a of the kth task on the energy storage device is:

[0068]

[0069] LEO satellite delay model: The transmission time of subtask b of the kth task to the LEO satellite is

[0070]

[0071] in, is the data rate between subtask b of the kth mission and its corresponding LEO satellite s. is the bandwidth of subtask b allocated by LEO satellite s to the kth task, P is the fixed transmission power of the energy storage device’s satellite-to-ground link, h is the channel gain between the energy storage device and its corresponding LEO satellite (the amplification or attenuation of the signal in the transmission path from the energy storage device to the LEO satellite), is the Gaussian white noise variance, i.e., the random interference from the energy storage device to the LEO satellite channel.

[0072] Subsequently, the LEO satellite transmits the processed subtask b of the kth task back to the energy storage device, and the transmission time is

[0073]

[0074] in, Indicates the data rate from LEO satellites to the energy storage device. The downlink bandwidth is fixed, that is, the data transmission rate from LEO satellites to the energy storage device is fixed. is a constant. The computing resources allocated to the LEO satellite s for the subtask b of the kth task are:

[0075]

[0076] Therefore, the total delay of offloading subtask b of the kth task to the LEO satellite s is

[0077]

[0078] Where c is the speed of light. is the physical distance from the energy storage device to the LEO satellite s. The total delay of the kth task is the time it takes the energy storage device to process the data and the delay of offloading to the satellite The maximum value of

[0079]

[0080] The split ratio vector and the bandwidth and computing resources allocated to each subtask processed by the energy storage device and the LEO satellite are brought into the linear programming model. Different allocation schemes result in different total energy consumption. The minimum total energy consumption under the split ratio vector is obtained through linear programming and constraints.

[0081] The nonlinear programming model used in this embodiment is as follows:

[0082]

[0083] Restrictions:

[0084]

[0085]

[0086]

[0087]

[0088]

[0089]

[0090] Specifically, the total energy consumption of the energy storage device is the cumulative sum of the energy consumption of processing each subtask and the cumulative energy consumption of unloading each subtask; the total energy consumption of the LEO satellite is the cumulative sum of the energy consumption of processing each subtask and the cumulative energy consumption of transmitting each subtask.

[0091] The total energy consumption model used in this embodiment is as follows:

[0092] The energy storage device calculates the energy consumption of subtask a of the kth task as:

[0093]

[0094] Where κ is the energy consumption coefficient. The energy consumption of unloading subtask b of the kth task to the LEO satellite is:

[0095]

[0096] The total energy consumption to complete the kth task is:

[0097]

[0098] The energy consumed by the LEO satellite to process subtask b of the kth task is:

[0099]

[0100] set up is the fixed transmission power of the LEO satellite. The transmission energy is:

[0101]

[0102] Therefore, the total energy consumption of LEO satellite s to complete subtask b of the kth mission is:

[0103]

[0104] The total energy consumption of the system is defined as the sum of the energy consumption of the energy storage device and the LEO satellite, represented by U:

[0105]

[0106] The reward function obtained by total energy consumption is used to train the TD3 neural network to generate a better segmentation ratio vector. The better segmentation ratio vector generates less total energy consumption than the segmentation ratio vector before training, and is therefore better.

[0107] In order to minimize the total energy consumption, this embodiment designs a reward function that is negatively correlated with the total energy consumption, namely:

[0108]

[0109] Where J is a constant to ensure that the reward function is positive; is the penalty when the total delay exceeds the actual maximum tolerable delay; if it does not exceed the actual maximum tolerable delay, The actual maximum tolerable delay is a fixed setting in actual application. is the energy consumed by the energy storage device to complete the kth task, is the total energy consumed by LEO satellite s to complete the kth mission; g is the scaling factor to ensure and energy consumption are of the same order of magnitude.

[0110] Embodiment 2:

[0111] Second, as Figure 2 The system for energy management of an off-grid energy storage device shown includes: a total task acquisition module, which is used to read the total task set stored in the off-grid energy storage device, and the total task set includes multiple tasks; a calculation module, which is communicated with the total task acquisition module and includes the following submodules: a segmentation and calculation module, an adjustment module, and an optimal generation module; the segmentation and calculation module calculates the total energy consumption E corresponding to the segmentation ratio vector based on the segmentation ratio vector, and the total energy consumption E includes the satellite total energy consumption E1 and the energy storage device total energy consumption E2, and the satellite total energy consumption E1 is the sum of the satellite subtask processing total energy consumption E11 and the satellite subtask transmission total energy consumption E12; the adjustment module is used to change the segmentation ratio vector; the optimal generation module is used to screen out the minimum total energy consumption Emin, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector. In actual use, the satellite can adopt a satellite network composed of one or more LEO satellites, and this embodiment adopts multiple LEO satellites.

[0112] The off-grid energy storage device energy management system also includes a power protection module; the power protection module is connected to the general task acquisition module and the calculation module to monitor the input and output voltage and current of the energy storage device. The power protection module is used to monitor the input and output voltage and current of the energy storage system. When the voltage exceeds or falls below the preset safety threshold, the system is automatically disconnected or the power output is adjusted; the current level is monitored to prevent the current from being too large or too small. Overcurrent may cause the wire to overheat and damage the equipment, while too low current may indicate a system fault; when a short circuit is detected, the current is quickly cut off to prevent damage to the energy storage device and the connected electrical system; when the system is started or the power suddenly changes, the output power is gradually increased or decreased to avoid the impact of instantaneous large current on the equipment.

[0113] The general task acquisition module is also used for data collection, including collecting necessary data of each group of batteries in the battery management system, such as total voltage, current, average temperature, charge and discharge current and power limit, single battery voltage, single battery temperature, balance status of each battery, fault and alarm information, historical charge and discharge power, historical charge and discharge energy, capacity, chargeable capacity, dischargeable capacity, charge and discharge times and other common information and displaying them; it also includes collecting and displaying relevant parameters of the PCS system (a digital mobile phone system that provides wireless voice and data communication services), including: voltage / current / power on the DC side, three-phase active power, reactive power, three-phase voltage, three-phase current, power factor, frequency, operating status, alarm and fault information and other common information of the PCS, as well as daily input power, daily output power, cumulative input power, cumulative output power, etc.; it also includes collecting various state quantities of the energy storage system, including the main circuit state (switch, accident trip signal, protection action signal, abnormal signal).

[0114] In this embodiment, the computing tasks of the computing module include but are not limited to the following aspects:

[0115] (1) Battery state estimation: Energy storage power stations need to accurately estimate the battery's future SOC (battery state of charge) and SOH (battery state of health) as a basis for battery replacement and modification of charging and discharging strategies. Currently, battery state estimation based on deep learning has achieved good results in specific scenarios.

[0116] (2) Power plant operation diagnosis: Use data analysis algorithms and machine learning models to conduct in-depth analysis of the collected data, identify abnormal behavior patterns, and quickly locate the cause of faults, such as overheating of battery cells, uneven voltage, short circuits, etc.

[0117] (3) Intelligent strategy scheduling: Calculate the optimal charging and discharging scheduling strategy based on the actual load curve of the power station and the policies and time-of-use electricity prices of the power station’s location. Introduce a large model to automatically capture and analyze information such as policies and time-of-use electricity prices to achieve fully automated intelligent scheduling.

[0118] (4) Log alarm analysis: Collect and display the event record information and alarm information (including historical information) of the entire system. When the analog quantity collected by the system exceeds the limit, the digital quantity changes position, or the network communication status self-diagnosis fault occurs, the energy management system will automatically pop up an alarm window to notify the operator that something has happened. At the same time, machine learning and data analysis technology are used to analyze historical alarm data, identify alarm patterns, predict potential faults, automatically perform preliminary fault diagnosis, and identify the root cause of the fault.

Claims

1. A method for energy management of off-grid energy storage equipment, characterized in that: include: Step S1: Obtaining the overall task; Reading a total task set stored in an off-grid energy storage device, wherein the total task set includes a plurality of tasks; Step S2: task segmentation; Splitting each of the tasks into subtasks according to a split ratio vector, wherein the split ratio vector reflects the ratio of each of the tasks split to the energy storage device for processing and to the satellite for processing; Step S3: energy consumption calculation; Calculate the total energy consumption E under the split ratio vector, wherein the total energy consumption E includes the total energy consumption E1 of the satellite and the total energy consumption E2 of the energy storage device, and the total energy consumption E1 of the satellite is the sum of the total energy consumption E11 of the satellite subtask processing and the total energy consumption E12 of the satellite subtask transmission; Step S4: adjusting the calculation; changing the split ratio vector, and calculating the total energy consumption E corresponding to different split ratio vectors; Step S5: selecting the minimum total energy consumption; selecting the minimum total energy consumption Emin from all the calculated total energy consumptions E, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector.

2. The method for energy management of off-grid energy storage equipment according to claim 1, characterized in that: In the step S4, a deep learning neural network is used to automatically change the segmentation ratio vector.

3. The method for energy management of off-grid energy storage equipment according to claim 2, characterized in that: The deep learning neural network is a double-delayed deep deterministic policy gradient network.

4. The method for energy management of off-grid energy storage equipment according to claim 2, characterized in that: In step S4, the deep learning neural network is initialized, and the initialization includes defining the state, action and reward of the deep learning neural network; the state includes necessary data of the task; the action is to generate a segmentation ratio vector; the reward is to set a reward function that is negatively correlated with the total energy consumption value E, which is used to train the deep learning neural network to generate a better segmentation ratio vector.

5. The method for energy management of off-grid energy storage equipment according to claim 4, characterized in that: The necessary data of each task include the task size, the computing resources required to process each bit of the task, the maximum tolerable delay of the task, and the ratio of the processed task size to the original task size.

6. The method for energy management of off-grid energy storage equipment according to any one of claims 1 to 5, characterized in that: In the step S4, the minimum total energy consumption E corresponding to the allocation of different bandwidths and computing resources under each of the split ratio vectors is calculated according to a linear programming model.

7. The method for energy management of off-grid energy storage equipment according to claim 6, characterized in that: The satellite is a low earth orbit satellite.

8. A system for energy management of off-grid energy storage equipment, characterized in that: Include: A total task acquisition module, the total task acquisition module is used to read the total task set stored in the off-grid energy storage device, the total task set includes multiple tasks; A calculation module, the calculation module is in communication with the overall task acquisition module and comprises the following submodules: a segmentation and calculation module, an adjustment module and an optimal generation module; The segmentation and calculation module calculates the total energy consumption E corresponding to the segmentation ratio vector based on the segmentation ratio vector, wherein the total energy consumption E includes the total energy consumption E1 of the satellite and the total energy consumption E2 of the energy storage device, and the total energy consumption E1 of the satellite is the sum of the total energy consumption E11 of the satellite subtask processing and the total energy consumption E12 of the satellite subtask transmission; The adjustment module is used to change the segmentation ratio vector; The optimal generation module is used to screen out the minimum total energy consumption Emin, and the segmentation ratio vector corresponding to the minimum total energy consumption Emin is the optimal segmentation ratio vector.

9. The off-grid energy storage device energy management system according to claim 8, characterized in that: It also includes a power protection module; the power protection module is communicatively connected with the general task acquisition module and the calculation module, and is used to monitor the input and output voltage and current of the energy storage device.